He also has a 100% rate of success when it comes to casting black magic and causing a team to completely throw their lead and lose bo5 simply by uttering the phrase "this is unthrowable, easy win"
Unthrowable (R): Sally casts a curse when he says 'unthrowable'. This will apply a whole team debuff that will cause them to throw the game from 10k gold lead to getting 3-0'd. This debuff cannot be cleansed.
Sjokz still gets flak to this day for that one time in 2020 and yet Caedrel never gets recognition for all the teams he cursed and he is still cursing.
Yeah, well most pros and teams have the exact same ways and thought processes of approaching drafts, in that they have their 20-30 champions or so and they won't even consider picking anything outside of that bubble. With a few exceptions, most teams draft like that so if you have a lot of insight into these processes and then you also know the comfort picks of the players and teams, then guessing which champs will be picked quickly becomes very easy because you only have to consider these 20-30 champs and don't have to think outside the box at all because most teams won't even try to be creative or innovative in their drafts.
@@MefoWho I'm just saying that him being part of the scene, having played as a pro himself and everything makes it very easy for him to guess what pros are thinking during draft... I'm not saying anyone can do it like it's nothing, but I'm saying that almost anyone with a multiple year background as a pro player or coach would be able to consistently predict pro drafts because they know how pros and coaches think. It's not a crazy difficult skill for someone with insight into the thought processes of pro teams.
@MefoWho he also once said that he needs to watch more games so he generally starts predicting picks mid season, also lck,lpl rarely go off meta. There are also picks that every big player is recognized for like Keria pyke so that helps
@lytom. Yea exactly, because the pro meta is narrow and only ever included a handful of picks per role once you figure out what these champs are and which champ is good against which it becomes easy it to guess what the pros will pick
Combining this with the strength of a team either the official riot power ranking or there rank within a region. would have weighted the how much he like a draft against the strength of a team.
you could weight the predictions by power difference in teams. As an exaple. GenG vs Bro, is weighted low. Because the skill difference is too big. Weighting could as mentioned be done by regional standing. The further apart two teams are. the less weight is applied.
Great video! As someone with a stats degree and an engineering career in developing analytics tools, I really appreciated how you broke down your methodology into understandable yet accurate explanations with respected to linear regression and r-squared. Unfortunately, I do think your model is missing too many variables. What you probably would want is a multiple regression model that factors in team strength as well. Team strength is a very difficult variable to quantify, but from traditional sports analytics a very decent variable to look at is Money Lines from from esports bookies (gambling websites). Bookies set and constantly adjust their odds such that an equilibrium is hit with the money on each side, so typically it gets pretty close to the perceived win probability in a match (before any drafting or game begins). You should translate money lines into numerical odds as well (for example, +100 translates to Even and +200 translates to 33%, while -200 translates into 66%). With this variable included, it will help isolate your results from being confounded with overall team difference. I would also consider whether you need to normalize the gold diff by match time (perhaps you can create two separate models).
This was my thought as well. It is a common occurrence for a team to "win draft" but be expected to lose due to team strength difference. That fact alone is problematic for these results.
This is a great suggestion. One thing I thought would be interesting to see, without doing much more work is splitting the three favour variables into their own graphs and looking at the R values for each. If Caedrels analysis is good, we should expect the R value to increase from Slight favour to strong favour. Clearly we're still missing a lot of information but I think that might represent the data better than what was shown in the video.
@@torr3nt_loading according to this dataset, which as you point out yourself isn't super great due to things like basing it off word choice. no strength of team included and strength of players on those champs can be fit under how cadrel feels about a draft but would also be an interesting data point to include (each players W-L on a champ in their carrier) edit also: which draft cadrel likes better isn't necessarily who he thinks will win either.
@@torr3nt_loadingGreat video but this statement is inherently wrong. Caedrel does not say that the team with the favored draft will win (for the most part), he just says their draft is better or likely to win. For instance, BRO could cook up the draft of a lifetime but we all know GenG will stomp them. This was the case last year when caedrel would say T1 have a less favored draft but says they’ll win anyways except for when facing GenG. Reality is , he’s an analysts (although I do think some are better). There’s objectively a better draft but that does not mean that the draft will win. A better way to analyze caedrels draft would be 1. How often he correctly predicts the champions picked on draft 2. Evaluate his drafts based on equal teams 3. How likely his prediction about the early , mid game and late game are often correct
Hey, data science student here, still learning a lot but this was super cool to watch and see applications of stuff I’ve learned on hobbies I have. Well done!
I think the binary of just win loss would be better. If I see a team go behind a bunch of gold and still win, it often means they had a draft advantage because they were allowed to misplay in the early game but had the tools in their comp to come back. A lot of times when a team wins with a worse draft, it was because they managed to stomp the early game so hard that the strengths of the champions didn't come into play.
Wow! Great content! really liked it. I think that in the discussion about "Bias" from Cadreal you should probably mention that as a streamer he MIGHT have an incentive to use stronger language that his actual opinion varrants. I Wish Cadreal would take a split where you two designed an experiment, with procentage chances and stuff. OFC the Hands, Team etc. factor would probably still be 90+ percent.
I mean, he is considered a “draft analyst.” Which I understand as, being able to predict what certain teams will do in draft, which champs are currently strong in the meta and will be contention points for certain teams, and over all when nameplates don’t exist which draft should be objectively stronger (whether it be through raw numbers or champion synergy). Which I think caedrel does a pretty damn good job with. The main takeaway I have from this video is that draft analysis is just that. It analyzes the draft, it doesn’t predict which team will win or lose. There are too many variables that are not related to the draft that also affect the outcome of the game. I don’t think that draft analysis is useless though. It allows for viewers to understand why a champ is being picked, why certain players will gravitate to certain picks, or why champs aren’t being picked for a team comp even though they’re considered “meta.” I don’t gamble, but I would never place a bet based off draft analysis alone. It doesn’t and quite simply can’t paint the full picture.
Oh my god this is so good!! If you have the time to pull up this Worlds data when Caedrel favors the other's team draft vs T1, I know T1 is going to win. That's why when he said he favors BLG draft over T1 in Game 4/5 I know T1 is winning Worlds. I don't know why I feel like T1 is the team he can't feel the draft most of the time and in the most weird way T1 pull off their draft Might be interesting for you to look at!
Worlds alone would likely be a very small data set as I can only record data for games where he commetns on draft, which is not as much as you'd think. This data set is just the summer split and only those leagues Ceadrel watches, so LPL, LCK and LEC plus the occasional LCK challenger games. I could add worlds to the data, but I doubt it would change the results very much.
The problem with those games is that the other team choked. It also happened in last year's semis game 1 against JDG. He said JDG had the better draft but 369 was piss useless with Rumble.
imo, "gold diff" is not enough to define the stomp tho. There are teams that can end quick with one or two teamfights and some that cant even end the game with barons or massive gold diff. And there are much more other factors like time, comp, player skills so to define stompness I think it's quite complicated. But overall a fun video fyi: t1 mdk 1647 is 8k diff, t1 blg game5 worlds final is 7k diff, with gold diff criteria, both games are in the game group tho lol
Fair point, a metric that controls for time might be more useful, though that runs into the problem of disadvantaging late game comps that only really start "stomping" after like min 25. Any ideas?
@@torr3nt_loading may be one could be used as a weight for the other (the less the game takes to finish the more weighted the gold diff). There are also some metrics that are usually used already like looking at gold diff, objectives taken...etc at specific times of the game. But this second approach seems a bit complicated. But it can tell you who won lane, transitioned better to the midgame and had better team fights in the lategame. But to keep it simple: just weights on game duration and may be which team is favored beforehand (using history of matches for exemple). The complicated part is the weights but a Bayesian approach could take care of that
Agree that a gold diff is not necessarily a good metric, especially if we talk about drafts out of all stuff. Although every team would like to have more gold than the other, it does not mean that every team's win condition is having maximum possible gold differential at the end of the game. You can have have 0 gold difference at 40 minutes, when team A has Smolder/Corki/Kassadin/etc teamcomp and they have 40% of their gold on their hypercarry, while team B has some kind of Ahri/Nocturne midgame-esque teamcomp without much of lategame potential. Even though there is no gold difference between two teams, it is clear that team A achieved their win conditions with much more success that team B, but wouldn't be reflected in the final gold diff if game ends quickly enough. At the same time if team B would be able to snowball and close out at 25 minutes, they would probably have a huge gold diff at this moment. Although much more demanding, an alternative approach would be to assign a "gold-hungriness" coefficient to each champion (big for hypercarries, small for most tanks) and calculate the "weighted gold differential", which would measure which team has succeeded more in feeding gold to the team members who need it more. If LoL stats would be accessible in a nice format without the need of manual work, you could make it more precise. One could even make this coefficient time-dependent (because you would like a lot of your gold on someone like Nidalee in the early game to get the snowball rolling, but at 30 minutes it's much less pertinent). This metric would reflect the success of a team in feeding a gold to the right team member at the right time. One could also use a less data-hungry approach by discretizing the time-averaging process : calculate time-champion-weighted gold diff at 15 min (to make justice for early game comps), 25 (midgame comps) and at the end (late game comps and overall). Another significant factor that I disagree with is "analyzing draft without nameplates". If you want to be able to predict game outcome from the draft, you SHOULD have the nameplates! Choosing is draft is kinda similar in choosing a champ you play in your soloQ game (although drafting a teamcomp is much more complex, there are still some similarities). If without considering your personal skills, picking Irelia would be 10% better than playing Riven, it doesn't mean you should be playing Irelia if you are a Riven OTP who have played 3 Irelia games in ARAM in your life. Similarly, if on this patch tank-heavy teamfight-focused teamcomps are slightly better performing than the other ones, it does not mean team A should draft it if they thrive in lane-snowball siege-centered teamcomps and their teamfights suck. Fun video in any case!
@@torr3nt_loadingyou prolly have to take more different metrics into account, weighing them vs each other (e.g the factor Gold difference gains more importance the longer the game goes, thus taking early wins into account but lategame comps not losing "stomp factor" cause theyll have a big Gold lead)
Great video! I think what this shows is that draft although an important starting point, is just not all the battle of league. League much like football or soccer is a game of execution, even with the perfect setup and right players, you need to make sure people are buying the right items and playing the right way.
I really love your work, but as you know the model it's too simple. It doesn't take in the concideration strength of the team, which can win games with bad drafts or weak teams might not understand win conditions of comps and lose. Second problem it's that predicting contionous variable with just 3 level variables is optimistic. However I like the metrics of gold diff and I think that the main problem is just lack of others explaining variables. And I think it's now possibile to compare strength of teams with AWS power rankings. For now i give sub and like because it's great work and i hope for more P.S. Its better to use caedrel prediction as discrete value rather than continous, because it can predict non-linear correlations
Ye, I also dislike the 3 point scale for draft_eval and it is (mathematically speaking) an abuse of the statistics to run a linear regression with an ordinal predictor variable, but between the two of us: who cares^^
Hi torr3nt, First off, good job on recording and regressing Caedrel's draft sentiment on the results of the game-that’s pretty cool! Your regression would likely improve if you divided the gold difference by the game length, as some of the variance in gold difference stems from game duration, and the relationship is probably close to linear. However, I would also strongly advocate showing the results of regressing the win/loss dummy variable on your draft sentiment metric. Are you willing to share your dataset? I would also recommend recording additional features, such as game length, team identities (even if you don’t use that variable, as it would require scraping team elo from the day prior), gold difference, tower counts, dragons, and barons for each team (to calculate differences later), each team's KDA (later adjusted by game length), inhibitors taken, First Blood, and nexus towers. Maybe even vision control, levels and cancelled spells, but these are likely more time consuming to manually scrape. The added time to record all these features wouldn't be too significant once you have the games loaded. The hardest part is probably listening to Caedrel for 100+ games, though his streaming-casting is enjoyable! With these variables, you could perform your original regression and additionally analyze how much impact features like First Blood have on Caedrel's draft signals. Standardizing variables is crucial here, and your gold metric should now also include towers. I believe a lot of the gold difference variation is tied to game length, so dividing by game length is important. Also goes for the other feature variables. Once again, good job on the video. Kind regards, Mads
Hey Mads, thanks for the feedback. I'd love to do a more complex analysis, but I'm afraid I'm kinda limited by my lazy data collection. Didn't even record which team won or was favoured or when the game was played. If I do this again in teh future, I'll do better. docs.google.com/spreadsheets/d/1uuA30j5P7nYlBND8v4jXfpzYSN_JlmQxD124qPj10TY/edit?gid=0#gid=0
Man it's really funny, im a biology student and your video is a really good example of scientific method, especially on data interpretation (like explaning what's R²) and discussion. Idk why but seeing a video put these concepts in the exact same way i learnt them is a pretty funny experience, especially on LoL where anecdotal evidence is king (cough cough losers queue)
it's awesome it really is, he went step by step through every "introduction to quantitative research" guide with it you could mouth what he would say next i was dying laughing half the time. he even did a little "proof of relevancy" in the first section and the full "limitations, outlook and further research prop" dance every uni student does to justify why their variables don't predict shit
Thanks! I have to give the caviat that this is not very serious scientific work, regressions technically don't allow for categorial variables (like my draft_eval) as independent variables. This is however often done: you essentially assume that the categorical variable represents an unobserved continuous variable. There are also more sophisticated ways around that, but tbh they are beyond my understanding of statistics^^
@@torr3nt_loading I mean sure, but you explained extremely boring and somewhat complex stuff (most people don't even see these kind of statistics) and you made it very fun ! And you also adressed in the video it wasn't serious so if you ask me you're the goat still
Draft can help win game, but we've seen what happens when team like KC at spring gaps in comp but ints the game many times. It also depends how the team can play/is comfortable on picks and their skill. I like more Pedros predictions what will be picked than if it's strong. Btw. there aint no way that he only ate 6 times in 100 games. He eats like 6 times in one bo3. Also it would be interesting if his predictions are better in different regions like LCK or LEC.
There is a lot of variance involved. Like an early game comp for an early game team should be a good draft in a vaccuum. But lets say a disastrous level 1 happens it might end in a stomp but that doesnt mean its a bad draft
Yo I feel like I’m in a stats class or listening to a research paper. hope you do this again next year with more details, would be super cool to watch and maybe even get him to help haha
If only I had not just finished my bachelor thesis where I had for some unholy reason decided to do an empirical study and thus have PTSD from terms like "statistical significance" I could have really enjoyed this video. Great video anyway!
Cool video, altho I think that it would be way more appropriate to use gold ratio of the two teams rather than an absolute difference, since winning a game in 18 min with 2k ahead should not be one fifth as impactuful as winning a game in 50 min with 10 k ahead
Interesting approach, I like that you went further than the binary correct/incorrect results, but I think there are other factors that could be interesting to look at, like game duration. For example, a game can be an absolute stomp and end at min. 22 but "only" have a gold diff of 5 or 6k gold. Perhaps it would be more representative to look at the gold difference per minute rather than total gold diff. Also would be interesting to take into account objectives that give you no gold like dragons. Another interesting factor would be the expected outcome, so for example if you play rogue vs. fnatic, the game is expected to go to fnatic regardless of whether rogue has a favorable draft. I understand this is in part what the R2 value means, but it might make caedrel's take seem bad if he said he liked the worse team's draft better but then they got stomped by a better team. Either way, great video, very interesting take and love to see some numbres on whether all these people claiming to know a lot about league actually know their shit hahaha
While I commend you for actually doing this because this is an area we sorely need data in I think the methodology is too flawed to draw any conclusions apart from that draft influences the game, and I don't think anyone would contest that statement. Some other thoughts of mine: 1. Could you make the data public? Not only would it allow for other people to verify your analysis, but also do their own without having to collect the same data. If you have more data than shown in the videon even better. 2. I think %gold would be just better than gold dif, but there might be even better criteria (pls comment your ideas) but these would probably have to be different depending on how the analysis is done. The winning teams gold lead seems to be around 10k regardless of the prediction which should tell us that this metric is not useful. Gold lead might be dependent on the type of game, but the type of game is not dependent on the gold lead. 3. As pointed out by other comments, this should be combined with a power ranking for the teams. Ideally you would do your own, since riots doesn't account for roster changes and values internationals differently to nationals. However an accurate team elo system is probably completely impossible given the amount of variations in any span of time you would have to carry out the analysis over to get any significance. Therefore I would just use the official ratings which are based on a modified elo system as explained here lolesports.com/en-GB/news/dev-diary-unveiling-the-global-power-rankings 4. To calculate a prior probability for the win and then calculating how much Caedrel's prediction affects the result. 5. As pointed out using 5 categories instead of 3 might yield better results, however each new category increases the influence of bias in the data collection since that would already mean having, for example, 50-50, 60-40, 70-30, 80-20 and (90-100)-(10-0) 6. Having multiple groups of different attitudes and familiarities with Caedrel doing the categorization, then analyzing those categories separately and comparing the results should yield the most interesting results. 7. As you point out this bias could be reduced by having multiple people do the categorization, though here we run into a question of what we want to measure. 8. If we want to analyze purely the accuracy of the words he is saying then no access to information such as teams and champions should be given. On the other hand if we want to analyze how much his comments help to understand the game at hand, then it might be better to give access to champs and teams. This would run into problems of plausible foreknowledge of how the game ended and older games would suffer from recency bias. 9. Important distinction: we are not measuring the effect of draft, we are trying to measure CAEDREL'S GUESS based on the draft. To even get at the effect of draft we would need multiple people to evaluate the drafts and even then there are multiple problems. 10. When analyzing the results we cannot be sure that Caedrel evaluated all the drafts based on the same criteria, especially those on the lower end might be based more on comfort/signature for the players.
Sure, but I'm afrtaid it won't be of much use, because I was a little lazy during data collection. Didn't even record which team won or was favoured or when the game was played. docs.google.com/spreadsheets/d/1uuA30j5P7nYlBND8v4jXfpzYSN_JlmQxD124qPj10TY/edit?gid=0#gid=0
Before I saw the results, I predicted that it would be near zero (indifference); it seems that was correct. Your work here is a good clue into the hidden problem with draft forecasting and analysis in League as a whole. Analysts think they are making predicitions about game outcomes and focus on this because they think this is the only game in town (unforunately this is the case). But what they are really doing for the most part, when you really listen to them, is determining the capabilities of champions and team compositions in the context of specific phases of the game and doctrinal adherence, not the outcome of the game itself. For example, one may say that composition A is favored over composition Z during the laning phase, and composition Z beats composition A in teamfights, but these inferences do not generalize to answering the question, "which team will win the game, and by how much?" This is related to problems in, for example, military science, where people may have good predicitions about the outcome of a specific battle, but it's much harder to predict the outcome of a long war. Some talented analysts in the LoL community are good at answering compositional favorability for specific operational moments, but this does not mean they are now oracles which predict game outcomes from draft alone. What does this mean? League analysts (I believe) have legitimately useful game knowledge, but this does not translate to forecasting ability. Why? Because players are actually primarily concerned with winning battles, not winning wars--these are not the same thing. However, players and analysts confuse themselves into thinking that they are the same thing. I have not seen anyone recognize this key distinction. Moreover, the theories in this discipline (if you can call it that) are very underdeveloped to nonexistent. Thus, every analyst resorts to forecasting because there are no real models in League, nor any attempt to make one.
I know this isn't really serious but a few things I thought about improving the approach: Gold diff as a measurement varies on length of the game, and some compositions favour game state and items in a way that's disproportionate to items, so that you might be barely ahead in gold but hugely advantaged because your champs scale better with the gold they have. Also, as you mentioned, indicating favour of a draft is not in a void. Teams are favoured over others, and so if Caedrel favours a draft of say BDS vs T1, it's hardly fair to treat that equally to him favouring T1's draft vs BDS's. In cases like that the difference in team skill might be large enough that the analysis was correct, but it wasn't enough to overcome skill difference.
there are quite a few covariances that affect the dependent variable, in here, the gold diff. i like the gold diff, but gold diff as a metric diminishes over time. a dependent value calculated by gold diff/time will take time into account. there is also scaling comps that their value is not reflected in gold diff, i.e. veigar, smolder. of course, personal bias, i.e. the rat’s preferences for dk and wbg, sometimes the underdogs like kt vs geng, might also account for his takes. players’ ability to get the most out of the comp can also confound this results, so a r-square of 0.06 where 6% of variance is explained by draft is actually quite good. anyhow, good video
The problem with predicting games is that, individual skill and cohesiveness of a team is more indicative of how well a team will perform than draft is. There are only a handful of truly great teams in league, if those teams draft well, then the gap between them and everyone else only gets bigger.
His analysis to drafts are actually really analytical and makes sense...the only factor that makes the predictions and analysis wrong are in the players and team playing the comp... Even if fox got the better draft, they will still get stomped by GenG...
Love the mathematical approach and the hard work, also it seems like 6% is not a lot, but imagine if every team had the same exact level and you were able to make your team win 6% more games than the others, it is not that bad altogether!
as a stats nerd, i would propose one more "uni level" bit of advice to making the testing more robust, which is using a broader 5-point scale, you could up it to 7 too, but there's diminished returns past 5, it allows for a bit more variation and a closer to accurate analysis after the fact.
The obvs problem with this is that many times the team that gets outdrafted still win because the other team doesn't execute the right plan/plays required for draft to work or just gets outplayed.
really sick content mate! love that you explain in detail your methodology and outline it's shortcoming even when its just for a ytb video lol. Do you consider sharing the data you collected, or making it available somewhere?
Sure, but I'm afrtaid it won't be of much use, because I was a little lazy during data collection. Didn't even record which team won or was favoured or when the game was played. docs.google.com/spreadsheets/d/1uuA30j5P7nYlBND8v4jXfpzYSN_JlmQxD124qPj10TY/edit?gid=0#gid=0
Covariance is 1780. But interpreting that beyond saying that draft_eval and gold_dif tend to incease together(which is already shown by R^2) is iffy: www.sciencedirect.com/topics/mathematics/positive-covariance
Sure, but I'm afrtaid it won't be of much use, because I was a little lazy during data collection. Didn't even record which team won or was favoured or when the game was played. docs.google.com/spreadsheets/d/1uuA30j5P7nYlBND8v4jXfpzYSN_JlmQxD124qPj10TY/edit?gid=0#gid=0
To an extent I guess the results confirm the common knowledge of "better team wins". There's a chance that, as LS often implies (though not obviously), that if the game played perfectly (i.e. if other factors are equalized) draft matters more.
I think that for this simple of a model, linear regretpssion is like shooting from a cannon to a bird, you could parametrical or non-parametrical correlation, and the result would be the same. I think that you should model things such as team strentgh (for example probability of win based on mmr on average winrate from last seasons, possibly weighted), or how good the people actually are on selected champions. Also im not sure if game time as a measure of stompiness isnt better, but its up to discussion
I disagree with using gold diff as a good evaluator for which comp is better, for example in a scaling vs tempo scenario, the tempo comp could still perform well in gold diff, but the scaling comp uses the gold more efficiently. Instead of a linear regression, you could use a logistic regression for whether a team won or lost. I think you also focused too much on the R^2 value when there are more important factors to a games outcome such as the head to head record of a team with each other, perhaps you could have looked at head to head win% of the two teams as another variable in the regression. R^2 looks at the variability explained by the model, which is no surprised you can't 100% predict the outcome of a game based on draft evaluation alone. I think it would be interesting to observe the difference between the R^2 value between a model with draft evaluation and a model without draft evaluation. Overall though, interesting video - I admire the amount of hours looking through Caedrel VoDs to collect this data
It’s really hard to do an analysis like this, as you note. The one idea I have to get past the “hands diff” and “build inting” is to get access to all league analytics. When he identifies a comp thats good, look for all games played with that same draft, hopefully on the same patch. It’s essentially multiple simulations of how the two comps do against each other with enough random variables that the “correct” answer pops out. Though I’m not sure how you’d get access to that data.
I think the 6% kind of relates to skill. As some who's never been to Challenger or GM, and barely reached Master, I believe no matter the comp, a group of 5 Masters can't beat a professional team with any team comp (even with pocket picks). The other 90-94% is micro/macro skill bs draft. On a even playing field, draft can potentially swing into ones favor. Also some cheese snowballs can push your comp into your favor if you aren't favored.
I think that looking at the outcome of the game to conclude on the strength of the daft just doesn't work at all. It's quite likely that good teams with good players can pick confort against bad players to secure the win. This would mean that the better comp loses more that the good comp on average and would defeat this type of analysis. It would be possible to integrate the game odds by using the twitter and analysts prediction from before the game and looking how much they outperformed or under performed. It would also be possible to power rank teams (though there might be some data leakage because the game outcome might be used in the power rankings). Even there, it wouldn't be optimal because the analysts can be rating the game outcome based on teams picking good drafts. I also think gold diff is quite a bad indicator because it mostly informs on game length than on it being a stomp. Game time would be slightly better but Caedrel was ranking aurelion sol smolder comp as busted early in the year and those comps scale and win late game. This can somewhat be mitigated by going for more complex analysis of stomp or not stomp (based on unchallenged gold graph, classifying comp types, quick victories, ...) The idea is really good and it's really good how you gathered the data though.
Nice analysis! Just out of interest: Did you also run the regression of prediction on pure outcome? I get the intuition behind your argument that this looses variance in the outcome, but after all only win/loss "really matters"
Ye, if you see it that way: in a high level competition every fraction of percent counts, then I agree, I might have undervalued the 6% in my interpretation. I guess the point I was making is, that other factors seem to be MUCH more important and Ceadrel seems to put a bit too much emphasis on it. But then again he is an analyst, so what do you expect. He's not gonna tell you how T1's chef is a huge advantage on the rift because Faker got his broccoli today.
The thing is that the outcome of the game is decided by so many factors and not just draft.And Caedrel is a guy that usually that does a lot of mistake cause the outcome of the game is completely random . Especially in LEC where good teams get rolled by bad teams .
Sure, but I'm afrtaid it won't be of much use, because I was a little lazy during data collection. Didn't even record which team won or was favoured or when the game was played. docs.google.com/spreadsheets/d/1uuA30j5P7nYlBND8v4jXfpzYSN_JlmQxD124qPj10TY/edit?gid=0#gid=0
Gold diff is useless as a comparison though. Different comps have different needs for for gold and are designed to come online at different times. Any gold diff at full build is useless. You also need to look at relative team strength. DFM vs T1 is not going to be determined by draft. Probably should scale the influence of these results in the analysis. I would take betting odds or analyst predictions before the draft as a baseline. And then check if favourable draft corelates to the team winning more than it should based on odds given by betting sites. Could be your next project. Interesting baseline, would be interesting to refine the data.
Sure, but I'm afrtaid it won't be of much use, because I was a little lazy during data collection. Didn't even record which team won or was favoured or when the game was played. docs.google.com/spreadsheets/d/1uuA30j5P7nYlBND8v4jXfpzYSN_JlmQxD124qPj10TY/edit?gid=0#gid=0
(1) You didn't take into account prior win rates (i.e. which team is more likely to win, e.g. by Riots power rankings). A better methodology would be comparing prediction power of only using that prior, and prediction power of prior + Bayesian update on draft (2) Gold is not a good indicator of a stomp for several reasons: (a) Total stomps end earlier, potentially leading to less gold lead since there's less time to build it (b) scaling comps don't need huge gold wins to win. The malphite draft is a good example: the malphite team can be down 5k gold but still be favored against a full AD comp (c) gold distribution os important. If your Alistar gets all the kills your gold leas is meaningless. Also champions like TF and Pyke skew gold statistics In general I think just using a binary variable is fine. If you use anything else you shift the goalpost which inevitably devalues a certain way of winning the game.
I think gold difference might not be the right stat to determine how much better a comp is. A lot of strong comps during the season were scaling comps with smolder ziggs that don't need a gold lead to be completly untouchable in games. Smolders stacks mean the actual power difference between comps is vastly different then gold difference. Using gold difference also unfairly overvalues snowballing comps large gold leads are more common due to the champions usually falling off later on and champions being balanced around being gold inflated.
im at 4:00 and I was thinking how can someone say he predicted it right or wrong. There is so much possibility to alter end of game gold like what if i had a shit draft but i took soul 4 barons and needed 15k gold diff to end. what if i picked super scaling and won in 40 min with 1 fight and in the end it was 3k gold lead but the enemy couldn't do nothing had 0 chance because of the draft and we just sit back and cleared waves traded obj and win in 40 min. I just remember DK at worlds started playing scaling for no reason was it a good draft maybe but for that team it wasnt clearly they needed early game skirmish drafts so we have to look at teams again what if we had a superteam against a last place team and the superteam win with a bad draft just because they are better. Looking only at end of game gold diff will not give a good representation of good or bad draft, sometimes it does but it wont always.
I agree, that gold dif wil not always be a good representtion of how well a comp did, but I think it is a decent proxy. Maybe, as some here have suggested, a gold dif weigted by game length would have been better. That account at least for some of the scenarios you mentioned.
@torr3nt_loading I would first of all check the playing teams they playstyle and the comp playstyle each mu win rate(proffessional) +weight player win rate on the champ + add soloq winrate if more data needed. That gives you multiple information about the comps you can create data based on this like winning lanes early comp type like skirmish scaling sidelane teamfight etc and check the actual early mid and lategame gold diff xp diff/lane obj diff . Then you have to compare the factual data to the one u were expecting (a team plays full early game vs full scaling expected to get gold lead early but if they didnt create much we can say the scaling team comps draft was good because they shut down the early of the enemy) if we evaluate the skill diff of the players then we can get this a bit closer to reality. TLDR check gold,xp,obj diff each lane multiple times in the game and calculate with skill, matchup and playstyle differences.
Gold lead is a misleading variable, as a team could have a major advantage in draft by for example playing passively, and winning a game by 3k gold was a stomp. Ex certain teams with Kalista being even or slightly behind at a certain point in the game they are even further behind than a gold lead can suggest.
Honestly, im not sure the gold is the best value since its the outcome who matters. For instance you could prefer a late-game comp that gets itself 10k behind, but still wins and finish the game at -5k. Also, looking at the graph (and supposing the - gold is a loss and the +gold is a win), I think the pattern is interesting. - When he thinks its slightly favored for a team, he is more often wrong. Maybe meaning the teamcomp is a bit worse but they picked confort, etc. - Seems pretty equal for the "favored" - Highly favored tho is where its more clear. About 20-7 ratio. So the draft, when clearly won does have a huge impact, according to those few samples.
Would percentage gold diff improve the data set over total gold diff? If you have a 20k gold lead, but one team has 90k gold and one has 110k gold, is that 20x more significant a factor as a 1k gold lead, when one team has 5500 and the other 4500. (Definitely extreme ends both ways, but that 1k gold lead should at LEAST be considered just as significant as the 20k gold lead.) If the draft picks were amazing, you might expect an early stomp, which might have a similar percentage gold diff, but a vastly underperforming total gold diff.
i really enjoyed the video and your hypothesis, but i feel like something like this is so hard to objectively evaluate, the human element is also something to take into account and because we are not robots or consistent machines, i often feel like he can be right about a draft and they still lose, or he can wrong and they win, everything has to coexist
Thing is, when his predictions and takes does miss, he can recognize that the players can play it differently than he imagined, or if there's other factors and variable that he didn't account for, and when confronted about it, he does address it and took it with grace(as much grace as a rat can have at least). Meanwhile some other streamers, especially *cough*LS*cough*, are so tunnel focused on their own takes and ideas of what the meta is, they don't think they can be wrong. They can't comprehend that the players and teams have their own preference, strategies and playstyle regardless of the meta, and that League is a game with thousands if not millions of variables than can't possibly be perfectly accounted for by just a draft. And when confronted about it, they just don't address it, evades it with something else, or just makes excuses.
Sure, but I'm afrtaid it won't be of much use, because I was a little lazy during data collection. Didn't even record which team won or was favoured or when the game was played. docs.google.com/spreadsheets/d/1uuA30j5P7nYlBND8v4jXfpzYSN_JlmQxD124qPj10TY/edit?gid=0#gid=0
7:00 the draft_eval variable is categorical, ordinal variable, not quantitative. Running a linear regression model on this data for fit is not going to give good results for regression variance/fit.
I m not sure taking the final gold diff is right though, a game might be close until it explodes and they take baron ,3 towers ... But honestly I don't know what other metric you can use (probably something like average gold difference per minute, though it's hard to get this data), another thing no fucking way he only ate 8 times during drafts XD.
Think your entire premise that gold difference is a good predictor of how good a draft was is just entirely flawed. Often times when we are talking about good drafts we mean drafts that can win the game even when several thousands of gold behind. Just the same if we were to think of a way to win despite having inferior drafting, it would probably be by snowballing early and having an insane gold lead. I would even venture a guess that it's a worse indicator of the draft being good than win/loss percentage for the reasons stated above.
Well, winning a game with a negative gold dif is really rare just because the towers during your final push and the team fight you likely just won give some much gold. But you are right, a good "comeback comp" that (in theory) would often end games with a close gold lead is not something the gold dif represents very well. Still, I don't agree that the entire premise is wong, gold dif is higher in onesided games. And if you say that the comp will roll the enemy team, you expect a onesided game and a win with a high gold dif.
5:10 i mean t1 draft is just objectively better If they didnt get stomped in 20 minutes they wouldve won at 30minutes+ with a 3k gold defecit Thats also kindof the problem with your evaluation metric Gold lead doesnt indicate how good a draft is. If your team is down 5k gold and can still win, thats draft diff
Fair, I still think gold_dif a good proxy for how well a team (and with that draft) did. As some have pointed out weighting by game length can offer a more accurate measure.
It was 6%, right? We talk about the highest level of competition. +6% Winrate is HUGE in that case. (Only by the draft diff) Not in my soloQ😂... in the Lck, LCS, MSI or Wolrds matches. Statistics is a very interesting subject because we can saw the SAME results way diffrent.
Its not 6% winrate. Favoured comps (all 3 categories) had a roughly 51% winrate. High favour comps had 66.67 WR, but the data set for those are only 30 games.
gold diff is not a good dependent variable. different comps have different gameplans and wincons. the goal of draft is not to win with the highest gold difference, the goal is to win period. I don't buy your justification to use gold diff. even a simple approach with W/L would've been better.
I dont think using gold diff is the correct way to evaluate draft predictions. What i mean by this is lets say that have one team with a better comp and they end up losing, then the fact that they have a good comp probably means that the enemy team requires a larger gold lead to end the game. Whereas the gold diff they themselves require to win is smaller
Gold difference at the end of the game is such a terrible metric. It is so composition dependent that it is basically worthless. Not scaling it with game length makes it even more silly. A 50 Minute slugfest that ends with the scaling draft barely surviving mid game and then slowly choking out the enemy team until they have a 15k gold lead is the same as a 20-0 kill 19 minute win in this metric.
Some drafts with a 70% win rate might actually produce 30% absolute stomp loses and 70% wins with slight gold advantages. There is no logic behind the idea that a greater gold lead and the average chance of winning have any kind of logical connection of equivalence.
You are discounting the bias that might come from being a favored in the game. If T1 plays FLY and T1 is a 5:1 favorite you don't think there is a chance this might play a role in impacting Caedrels prediction? For example he says T1 won draft. And they win. Was it because T1 had the better draft or because they were better players? And did this info influence Faker. The easiest way around it would be to only use games between teams that were even odds in your analysis.
i think this research means close to nothing, firstly a team comp is just one small variable amongst hubdreds of others secondly the mechanic of adding and subtracting game gold is very very stupid for many reasons
He also has a 100% rate of success when it comes to casting black magic and causing a team to completely throw their lead and lose bo5 simply by uttering the phrase "this is unthrowable, easy win"
Unthrowable (R): Sally casts a curse when he says 'unthrowable'. This will apply a whole team debuff that will cause them to throw the game from 10k gold lead to getting 3-0'd. This debuff cannot be cleansed.
Sjokz still gets flak to this day for that one time in 2020 and yet Caedrel never gets recognition for all the teams he cursed and he is still cursing.
@@valmiro4164 The " It all starts with" incident
somebody pull out the stats on how many times did the team jersey's pedro was wearing, lost the game, i think it tanked last year lmao
WizardCastsBlackCurse
He's alot better at predicting the champs drafted than who wins. Our rat
Yeah, well most pros and teams have the exact same ways and thought processes of approaching drafts, in that they have their 20-30 champions or so and they won't even consider picking anything outside of that bubble. With a few exceptions, most teams draft like that so if you have a lot of insight into these processes and then you also know the comfort picks of the players and teams, then guessing which champs will be picked quickly becomes very easy because you only have to consider these 20-30 champs and don't have to think outside the box at all because most teams won't even try to be creative or innovative in their drafts.
@@mojin7470 its like saying anyone can be a doctor, just learn 4Head
@@MefoWho I'm just saying that him being part of the scene, having played as a pro himself and everything makes it very easy for him to guess what pros are thinking during draft... I'm not saying anyone can do it like it's nothing, but I'm saying that almost anyone with a multiple year background as a pro player or coach would be able to consistently predict pro drafts because they know how pros and coaches think. It's not a crazy difficult skill for someone with insight into the thought processes of pro teams.
@MefoWho he also once said that he needs to watch more games so he generally starts predicting picks mid season, also lck,lpl rarely go off meta. There are also picks that every big player is recognized for like Keria pyke so that helps
@lytom. Yea exactly, because the pro meta is narrow and only ever included a handful of picks per role once you figure out what these champs are and which champ is good against which it becomes easy it to guess what the pros will pick
ending the conclusion with a variant of "more research is needed", classic
It's the only move
Combining this with the strength of a team either the official riot power ranking or there rank within a region. would have weighted the how much he like a draft against the strength of a team.
That's a nice idea, not sure how you'd go about this though.
@@torr3nt_loadingmaybe normalize the rating and put it as multiplier / another variable
@@torr3nt_loading you'd probobly have to complete an elo calculaton for each team
you could weight the predictions by power difference in teams. As an exaple. GenG vs Bro, is weighted low. Because the skill difference is too big. Weighting could as mentioned be done by regional standing. The further apart two teams are. the less weight is applied.
This weighting would also create a continous variable instead of 0,1,2
Great video! As someone with a stats degree and an engineering career in developing analytics tools, I really appreciated how you broke down your methodology into understandable yet accurate explanations with respected to linear regression and r-squared. Unfortunately, I do think your model is missing too many variables. What you probably would want is a multiple regression model that factors in team strength as well. Team strength is a very difficult variable to quantify, but from traditional sports analytics a very decent variable to look at is Money Lines from from esports bookies (gambling websites). Bookies set and constantly adjust their odds such that an equilibrium is hit with the money on each side, so typically it gets pretty close to the perceived win probability in a match (before any drafting or game begins). You should translate money lines into numerical odds as well (for example, +100 translates to Even and +200 translates to 33%, while -200 translates into 66%). With this variable included, it will help isolate your results from being confounded with overall team difference.
I would also consider whether you need to normalize the gold diff by match time (perhaps you can create two separate models).
This was my thought as well. It is a common occurrence for a team to "win draft" but be expected to lose due to team strength difference. That fact alone is problematic for these results.
This is a great suggestion. One thing I thought would be interesting to see, without doing much more work is splitting the three favour variables into their own graphs and looking at the R values for each. If Caedrels analysis is good, we should expect the R value to increase from Slight favour to strong favour. Clearly we're still missing a lot of information but I think that might represent the data better than what was shown in the video.
Also, teams can win with negative gold difference...
he kinda predicted finals drafts i dont remember his exact takes but while watching i was matching memories and draft and thought he was quite right
Sure, occasionally, Caedrel is spot on, but sometimes he's way off. On average he is better than flipping a coin, but not very much.
The one iteration he didn't do was T1 banning Yone and that just eats Skarner ban for BLG
Depends how many rats are in the game. More rats. More accurate.
@@torr3nt_loading according to this dataset, which as you point out yourself isn't super great due to things like basing it off word choice. no strength of team included and strength of players on those champs can be fit under how cadrel feels about a draft but would also be an interesting data point to include (each players W-L on a champ in their carrier)
edit also: which draft cadrel likes better isn't necessarily who he thinks will win either.
@@torr3nt_loadingGreat video but this statement is inherently wrong. Caedrel does not say that the team with the favored draft will win (for the most part), he just says their draft is better or likely to win. For instance, BRO could cook up the draft of a lifetime but we all know GenG will stomp them. This was the case last year when caedrel would say T1 have a less favored draft but says they’ll win anyways except for when facing GenG.
Reality is , he’s an analysts (although I do think some are better). There’s objectively a better draft but that does not mean that the draft will win. A better way to analyze caedrels draft would be
1. How often he correctly predicts the champions picked on draft
2. Evaluate his drafts based on equal teams
3. How likely his prediction about the early , mid game and late game are often correct
I’ve just started learning linear regression in stats and this is unironically very helpful
Hey, data science student here, still learning a lot but this was super cool to watch and see applications of stuff I’ve learned on hobbies I have. Well done!
I think the binary of just win loss would be better. If I see a team go behind a bunch of gold and still win, it often means they had a draft advantage because they were allowed to misplay in the early game but had the tools in their comp to come back. A lot of times when a team wins with a worse draft, it was because they managed to stomp the early game so hard that the strengths of the champions didn't come into play.
With a binary variable (favoured draft lost=1; favoured draft won=1) you get an even worse R^2 of .05.
@@torr3nt_loading not looking good for our Rat King then, but I still stand by that being a more accurate measurement
I have been wondering about this since earlier in the year, thanks for taking the time to do it!
Wow! Great content! really liked it. I think that in the discussion about "Bias" from Cadreal you should probably mention that as a streamer he MIGHT have an incentive to use stronger language that his actual opinion varrants. I Wish Cadreal would take a split where you two designed an experiment, with procentage chances and stuff. OFC the Hands, Team etc. factor would probably still be 90+ percent.
You're right, hyping up things is entertaining, hadn't thought of that.
I mean, he is considered a “draft analyst.” Which I understand as, being able to predict what certain teams will do in draft, which champs are currently strong in the meta and will be contention points for certain teams, and over all when nameplates don’t exist which draft should be objectively stronger (whether it be through raw numbers or champion synergy). Which I think caedrel does a pretty damn good job with.
The main takeaway I have from this video is that draft analysis is just that. It analyzes the draft, it doesn’t predict which team will win or lose. There are too many variables that are not related to the draft that also affect the outcome of the game. I don’t think that draft analysis is useless though. It allows for viewers to understand why a champ is being picked, why certain players will gravitate to certain picks, or why champs aren’t being picked for a team comp even though they’re considered “meta.” I don’t gamble, but I would never place a bet based off draft analysis alone. It doesn’t and quite simply can’t paint the full picture.
Love this video, very well done. Editing, commentary and discussion is great
Oh my god this is so good!! If you have the time to pull up this Worlds data when Caedrel favors the other's team draft vs T1, I know T1 is going to win. That's why when he said he favors BLG draft over T1 in Game 4/5 I know T1 is winning Worlds. I don't know why I feel like T1 is the team he can't feel the draft most of the time and in the most weird way T1 pull off their draft
Might be interesting for you to look at!
Worlds alone would likely be a very small data set as I can only record data for games where he commetns on draft, which is not as much as you'd think. This data set is just the summer split and only those leagues Ceadrel watches, so LPL, LCK and LEC plus the occasional LCK challenger games. I could add worlds to the data, but I doubt it would change the results very much.
The problem with those games is that the other team choked. It also happened in last year's semis game 1 against JDG. He said JDG had the better draft but 369 was piss useless with Rumble.
Just wanna say this is really fantastic content! Would love to see you do more stats breakdown of Caedrels content!!
Great video man! Respect for the work you put in gathering data. Keep it up!
imo, "gold diff" is not enough to define the stomp tho. There are teams that can end quick with one or two teamfights and some that cant even end the game with barons or massive gold diff. And there are much more other factors like time, comp, player skills so to define stompness I think it's quite complicated. But overall a fun video
fyi: t1 mdk 1647 is 8k diff, t1 blg game5 worlds final is 7k diff, with gold diff criteria, both games are in the game group tho lol
Fair point, a metric that controls for time might be more useful, though that runs into the problem of disadvantaging late game comps that only really start "stomping" after like min 25. Any ideas?
@@torr3nt_loading may be one could be used as a weight for the other (the less the game takes to finish the more weighted the gold diff). There are also some metrics that are usually used already like looking at gold diff, objectives taken...etc at specific times of the game. But this second approach seems a bit complicated. But it can tell you who won lane, transitioned better to the midgame and had better team fights in the lategame. But to keep it simple: just weights on game duration and may be which team is favored beforehand (using history of matches for exemple). The complicated part is the weights but a Bayesian approach could take care of that
Agree that a gold diff is not necessarily a good metric, especially if we talk about drafts out of all stuff. Although every team would like to have more gold than the other, it does not mean that every team's win condition is having maximum possible gold differential at the end of the game. You can have have 0 gold difference at 40 minutes, when team A has Smolder/Corki/Kassadin/etc teamcomp and they have 40% of their gold on their hypercarry, while team B has some kind of Ahri/Nocturne midgame-esque teamcomp without much of lategame potential. Even though there is no gold difference between two teams, it is clear that team A achieved their win conditions with much more success that team B, but wouldn't be reflected in the final gold diff if game ends quickly enough. At the same time if team B would be able to snowball and close out at 25 minutes, they would probably have a huge gold diff at this moment.
Although much more demanding, an alternative approach would be to assign a "gold-hungriness" coefficient to each champion (big for hypercarries, small for most tanks) and calculate the "weighted gold differential", which would measure which team has succeeded more in feeding gold to the team members who need it more.
If LoL stats would be accessible in a nice format without the need of manual work, you could make it more precise. One could even make this coefficient time-dependent (because you would like a lot of your gold on someone like Nidalee in the early game to get the snowball rolling, but at 30 minutes it's much less pertinent). This metric would reflect the success of a team in feeding a gold to the right team member at the right time. One could also use a less data-hungry approach by discretizing the time-averaging process : calculate time-champion-weighted gold diff at 15 min (to make justice for early game comps), 25 (midgame comps) and at the end (late game comps and overall).
Another significant factor that I disagree with is "analyzing draft without nameplates". If you want to be able to predict game outcome from the draft, you SHOULD have the nameplates! Choosing is draft is kinda similar in choosing a champ you play in your soloQ game (although drafting a teamcomp is much more complex, there are still some similarities). If without considering your personal skills, picking Irelia would be 10% better than playing Riven, it doesn't mean you should be playing Irelia if you are a Riven OTP who have played 3 Irelia games in ARAM in your life. Similarly, if on this patch tank-heavy teamfight-focused teamcomps are slightly better performing than the other ones, it does not mean team A should draft it if they thrive in lane-snowball siege-centered teamcomps and their teamfights suck.
Fun video in any case!
@@torr3nt_loadingyou prolly have to take more different metrics into account, weighing them vs each other (e.g the factor Gold difference gains more importance the longer the game goes, thus taking early wins into account but lategame comps not losing "stomp factor" cause theyll have a big Gold lead)
@@torr3nt_loading What about using the % total gold stat instead of the absolute difference?
Great video man! Good analysis, and enough memes to be perfect!
Great video!
I think what this shows is that draft although an important starting point, is just not all the battle of league. League much like football or soccer is a game of execution, even with the perfect setup and right players, you need to make sure people are buying the right items and playing the right way.
I really love your work, but as you know the model it's too simple. It doesn't take in the concideration strength of the team, which can win games with bad drafts or weak teams might not understand win conditions of comps and lose. Second problem it's that predicting contionous variable with just 3 level variables is optimistic. However I like the metrics of gold diff and I think that the main problem is just lack of others explaining variables. And I think it's now possibile to compare strength of teams with AWS power rankings. For now i give sub and like because it's great work and i hope for more
P.S. Its better to use caedrel prediction as discrete value rather than continous, because it can predict non-linear correlations
Ye, I also dislike the 3 point scale for draft_eval and it is (mathematically speaking) an abuse of the statistics to run a linear regression with an ordinal predictor variable, but between the two of us: who cares^^
@@torr3nt_loadingI care a little bit 😂😂
Hi torr3nt,
First off, good job on recording and regressing Caedrel's draft sentiment on the results of the game-that’s pretty cool!
Your regression would likely improve if you divided the gold difference by the game length, as some of the variance in gold difference stems from game duration, and the relationship is probably close to linear. However, I would also strongly advocate showing the results of regressing the win/loss dummy variable on your draft sentiment metric.
Are you willing to share your dataset?
I would also recommend recording additional features, such as game length, team identities (even if you don’t use that variable, as it would require scraping team elo from the day prior), gold difference, tower counts, dragons, and barons for each team (to calculate differences later), each team's KDA (later adjusted by game length), inhibitors taken, First Blood, and nexus towers. Maybe even vision control, levels and cancelled spells, but these are likely more time consuming to manually scrape.
The added time to record all these features wouldn't be too significant once you have the games loaded. The hardest part is probably listening to Caedrel for 100+ games, though his streaming-casting is enjoyable!
With these variables, you could perform your original regression and additionally analyze how much impact features like First Blood have on Caedrel's draft signals. Standardizing variables is crucial here, and your gold metric should now also include towers. I believe a lot of the gold difference variation is tied to game length, so dividing by game length is important. Also goes for the other feature variables.
Once again, good job on the video.
Kind regards,
Mads
Hey Mads, thanks for the feedback. I'd love to do a more complex analysis, but I'm afraid I'm kinda limited by my lazy data collection. Didn't even record which team won or was favoured or when the game was played. If I do this again in teh future, I'll do better.
docs.google.com/spreadsheets/d/1uuA30j5P7nYlBND8v4jXfpzYSN_JlmQxD124qPj10TY/edit?gid=0#gid=0
Man it's really funny, im a biology student and your video is a really good example of scientific method, especially on data interpretation (like explaning what's R²) and discussion.
Idk why but seeing a video put these concepts in the exact same way i learnt them is a pretty funny experience, especially on LoL where anecdotal evidence is king (cough cough losers queue)
it's awesome it really is, he went step by step through every "introduction to quantitative research" guide with it you could mouth what he would say next i was dying laughing half the time. he even did a little "proof of relevancy" in the first section and the full "limitations, outlook and further research prop" dance every uni student does to justify why their variables don't predict shit
Thanks! I have to give the caviat that this is not very serious scientific work, regressions technically don't allow for categorial variables (like my draft_eval) as independent variables. This is however often done: you essentially assume that the categorical variable represents an unobserved continuous variable. There are also more sophisticated ways around that, but tbh they are beyond my understanding of statistics^^
@@torr3nt_loading I mean sure, but you explained extremely boring and somewhat complex stuff (most people don't even see these kind of statistics) and you made it very fun ! And you also adressed in the video it wasn't serious so if you ask me you're the goat still
I agree it's actually kinda crazy
I predicted Sylas would be a mega pick this worlds, TY king rat.
Great Vid
would be interesting to see if others, e.g. LS have better results
Draft can help win game, but we've seen what happens when team like KC at spring gaps in comp but ints the game many times. It also depends how the team can play/is comfortable on picks and their skill. I like more Pedros predictions what will be picked than if it's strong. Btw. there aint no way that he only ate 6 times in 100 games. He eats like 6 times in one bo3. Also it would be interesting if his predictions are better in different regions like LCK or LEC.
Procastinating from reviewing for a stats exam I have in around 3 hours and watching this vid unknowingly what will be in it is an experience.
crazy good video bro, an inspiration thanks
There is a lot of variance involved. Like an early game comp for an early game team should be a good draft in a vaccuum. But lets say a disastrous level 1 happens it might end in a stomp but that doesnt mean its a bad draft
More of this content as someone who is pursuing an analytics degree! Doing analytics with something you find personally fun is much more rewarding :)
Yo I feel like I’m in a stats class or listening to a research paper. hope you do this again next year with more details, would be super cool to watch and maybe even get him to help haha
If only I had not just finished my bachelor thesis where I had for some unholy reason decided to do an empirical study and thus have PTSD from terms like "statistical significance" I could have really enjoyed this video.
Great video anyway!
Cool video, altho I think that it would be way more appropriate to use gold ratio of the two teams rather than an absolute difference, since winning a game in 18 min with 2k ahead should not be one fifth as impactuful as winning a game in 50 min with 10 k ahead
Interesting approach, I like that you went further than the binary correct/incorrect results, but I think there are other factors that could be interesting to look at, like game duration. For example, a game can be an absolute stomp and end at min. 22 but "only" have a gold diff of 5 or 6k gold. Perhaps it would be more representative to look at the gold difference per minute rather than total gold diff. Also would be interesting to take into account objectives that give you no gold like dragons.
Another interesting factor would be the expected outcome, so for example if you play rogue vs. fnatic, the game is expected to go to fnatic regardless of whether rogue has a favorable draft. I understand this is in part what the R2 value means, but it might make caedrel's take seem bad if he said he liked the worse team's draft better but then they got stomped by a better team.
Either way, great video, very interesting take and love to see some numbres on whether all these people claiming to know a lot about league actually know their shit hahaha
Very nice video, thank you for that!
While I commend you for actually doing this because this is an area we sorely need data in I think the methodology is too flawed to draw any conclusions apart from that draft influences the game, and I don't think anyone would contest that statement. Some other thoughts of mine:
1. Could you make the data public? Not only would it allow for other people to verify your analysis, but also do their own without having to collect the same data. If you have more data than shown in the videon even better.
2. I think %gold would be just better than gold dif, but there might be even better criteria (pls comment your ideas) but these would probably have to be different depending on how the analysis is done. The winning teams gold lead seems to be around 10k regardless of the prediction which should tell us that this metric is not useful. Gold lead might be dependent on the type of game, but the type of game is not dependent on the gold lead.
3. As pointed out by other comments, this should be combined with a power ranking for the teams. Ideally you would do your own, since riots doesn't account for roster changes and values internationals differently to nationals. However an accurate team elo system is probably completely impossible given the amount of variations in any span of time you would have to carry out the analysis over to get any significance. Therefore I would just use the official ratings which are based on a modified elo system as explained here lolesports.com/en-GB/news/dev-diary-unveiling-the-global-power-rankings
4. To calculate a prior probability for the win and then calculating how much Caedrel's prediction affects the result.
5. As pointed out using 5 categories instead of 3 might yield better results, however each new category increases the influence of bias in the data collection since that would already mean having, for example,
50-50, 60-40, 70-30, 80-20 and (90-100)-(10-0)
6. Having multiple groups of different attitudes and familiarities with Caedrel doing the categorization, then analyzing those categories separately and comparing the results should yield the most interesting results.
7. As you point out this bias could be reduced by having multiple people do the categorization, though here we run into a question of what we want to measure.
8. If we want to analyze purely the accuracy of the words he is saying then no access to information such as teams and champions should be given. On the other hand if we want to analyze how much his comments help to understand the game at hand, then it might be better to give access to champs and teams. This would run into problems of plausible foreknowledge of how the game ended and older games would suffer from recency bias.
9. Important distinction: we are not measuring the effect of draft, we are trying to measure CAEDREL'S GUESS based on the draft. To even get at the effect of draft we would need multiple people to evaluate the drafts and even then there are multiple problems.
10. When analyzing the results we cannot be sure that Caedrel evaluated all the drafts based on the same criteria, especially those on the lower end might be based more on comfort/signature for the players.
Sure, but I'm afrtaid it won't be of much use, because I was a little lazy during data collection. Didn't even record which team won or was favoured or when the game was played.
docs.google.com/spreadsheets/d/1uuA30j5P7nYlBND8v4jXfpzYSN_JlmQxD124qPj10TY/edit?gid=0#gid=0
Before I saw the results, I predicted that it would be near zero (indifference); it seems that was correct.
Your work here is a good clue into the hidden problem with draft forecasting and analysis in League as a whole. Analysts think they are making predicitions about game outcomes and focus on this because they think this is the only game in town (unforunately this is the case). But what they are really doing for the most part, when you really listen to them, is determining the capabilities of champions and team compositions in the context of specific phases of the game and doctrinal adherence, not the outcome of the game itself. For example, one may say that composition A is favored over composition Z during the laning phase, and composition Z beats composition A in teamfights, but these inferences do not generalize to answering the question, "which team will win the game, and by how much?" This is related to problems in, for example, military science, where people may have good predicitions about the outcome of a specific battle, but it's much harder to predict the outcome of a long war. Some talented analysts in the LoL community are good at answering compositional favorability for specific operational moments, but this does not mean they are now oracles which predict game outcomes from draft alone.
What does this mean? League analysts (I believe) have legitimately useful game knowledge, but this does not translate to forecasting ability. Why? Because players are actually primarily concerned with winning battles, not winning wars--these are not the same thing. However, players and analysts confuse themselves into thinking that they are the same thing. I have not seen anyone recognize this key distinction. Moreover, the theories in this discipline (if you can call it that) are very underdeveloped to nonexistent. Thus, every analyst resorts to forecasting because there are no real models in League, nor any attempt to make one.
That's a pretty good analytics
I know this isn't really serious but a few things I thought about improving the approach:
Gold diff as a measurement varies on length of the game, and some compositions favour game state and items in a way that's disproportionate to items, so that you might be barely ahead in gold but hugely advantaged because your champs scale better with the gold they have.
Also, as you mentioned, indicating favour of a draft is not in a void. Teams are favoured over others, and so if Caedrel favours a draft of say BDS vs T1, it's hardly fair to treat that equally to him favouring T1's draft vs BDS's. In cases like that the difference in team skill might be large enough that the analysis was correct, but it wasn't enough to overcome skill difference.
there are quite a few covariances that affect the dependent variable, in here, the gold diff. i like the gold diff, but gold diff as a metric diminishes over time. a dependent value calculated by gold diff/time will take time into account. there is also scaling comps that their value is not reflected in gold diff, i.e. veigar, smolder. of course, personal bias, i.e. the rat’s preferences for dk and wbg, sometimes the underdogs like kt vs geng, might also account for his takes. players’ ability to get the most out of the comp can also confound this results, so a r-square of 0.06 where 6% of variance is explained by draft is actually quite good. anyhow, good video
The problem with predicting games is that, individual skill and cohesiveness of a team is more indicative of how well a team will perform than draft is. There are only a handful of truly great teams in league, if those teams draft well, then the gap between them and everyone else only gets bigger.
His analysis to drafts are actually really analytical and makes sense...the only factor that makes the predictions and analysis wrong are in the players and team playing the comp... Even if fox got the better draft, they will still get stomped by GenG...
Love the mathematical approach and the hard work, also it seems like 6% is not a lot, but imagine if every team had the same exact level and you were able to make your team win 6% more games than the others, it is not that bad altogether!
as a stats nerd, i would propose one more "uni level" bit of advice to making the testing more robust, which is using a broader 5-point scale, you could up it to 7 too, but there's diminished returns past 5, it allows for a bit more variation and a closer to accurate analysis after the fact.
Would have liked to do that, but it gets really hard to separate Caedrel's comments into so many categories.
@@torr3nt_loading honestly i get it, it's more time spent ofc so no shame for it either.
The obvs problem with this is that many times the team that gets outdrafted still win because the other team doesn't execute the right plan/plays required for draft to work or just gets outplayed.
really sick content mate! love that you explain in detail your methodology and outline it's shortcoming even when its just for a ytb video lol. Do you consider sharing the data you collected, or making it available somewhere?
Sure, but I'm afrtaid it won't be of much use, because I was a little lazy during data collection. Didn't even record which team won or was favoured or when the game was played.
docs.google.com/spreadsheets/d/1uuA30j5P7nYlBND8v4jXfpzYSN_JlmQxD124qPj10TY/edit?gid=0#gid=0
i think you should have perhaps evaluated the covariance between gold_diff and the 1-3 score caedrel gave the drafts
Covariance is 1780. But interpreting that beyond saying that draft_eval and gold_dif tend to incease together(which is already shown by R^2) is iffy: www.sciencedirect.com/topics/mathematics/positive-covariance
Super interested in the dataset to do some work with! Wonder if you are willing to open source the work behind it. Really cool video!
Sure, but I'm afrtaid it won't be of much use, because I was a little lazy during data collection. Didn't even record which team won or was favoured or when the game was played.
docs.google.com/spreadsheets/d/1uuA30j5P7nYlBND8v4jXfpzYSN_JlmQxD124qPj10TY/edit?gid=0#gid=0
a few of his achievements:
sally ult
mikyx over keria
keria out of top 5 supports
doran's previous lawyer
zeus current lawyer
To an extent I guess the results confirm the common knowledge of "better team wins". There's a chance that, as LS often implies (though not obviously), that if the game played perfectly (i.e. if other factors are equalized) draft matters more.
I think that for this simple of a model, linear regretpssion is like shooting from a cannon to a bird, you could parametrical or non-parametrical correlation, and the result would be the same.
I think that you should model things such as team strentgh (for example probability of win based on mmr on average winrate from last seasons, possibly weighted), or how good the people actually are on selected champions. Also im not sure if game time as a measure of stompiness isnt better, but its up to discussion
Expecting slight, but not overwhelming correlation.
I disagree with using gold diff as a good evaluator for which comp is better, for example in a scaling vs tempo scenario, the tempo comp could still perform well in gold diff, but the scaling comp uses the gold more efficiently. Instead of a linear regression, you could use a logistic regression for whether a team won or lost.
I think you also focused too much on the R^2 value when there are more important factors to a games outcome such as the head to head record of a team with each other, perhaps you could have looked at head to head win% of the two teams as another variable in the regression.
R^2 looks at the variability explained by the model, which is no surprised you can't 100% predict the outcome of a game based on draft evaluation alone. I think it would be interesting to observe the difference between the R^2 value between a model with draft evaluation and a model without draft evaluation.
Overall though, interesting video - I admire the amount of hours looking through Caedrel VoDs to collect this data
It’s really hard to do an analysis like this, as you note. The one idea I have to get past the “hands diff” and “build inting” is to get access to all league analytics. When he identifies a comp thats good, look for all games played with that same draft, hopefully on the same patch. It’s essentially multiple simulations of how the two comps do against each other with enough random variables that the “correct” answer pops out. Though I’m not sure how you’d get access to that data.
I think the 6% kind of relates to skill. As some who's never been to Challenger or GM, and barely reached Master, I believe no matter the comp, a group of 5 Masters can't beat a professional team with any team comp (even with pocket picks). The other 90-94% is micro/macro skill bs draft.
On a even playing field, draft can potentially swing into ones favor.
Also some cheese snowballs can push your comp into your favor if you aren't favored.
I think that looking at the outcome of the game to conclude on the strength of the daft just doesn't work at all. It's quite likely that good teams with good players can pick confort against bad players to secure the win. This would mean that the better comp loses more that the good comp on average and would defeat this type of analysis.
It would be possible to integrate the game odds by using the twitter and analysts prediction from before the game and looking how much they outperformed or under performed. It would also be possible to power rank teams (though there might be some data leakage because the game outcome might be used in the power rankings).
Even there, it wouldn't be optimal because the analysts can be rating the game outcome based on teams picking good drafts.
I also think gold diff is quite a bad indicator because it mostly informs on game length than on it being a stomp. Game time would be slightly better but Caedrel was ranking aurelion sol smolder comp as busted early in the year and those comps scale and win late game. This can somewhat be mitigated by going for more complex analysis of stomp or not stomp (based on unchallenged gold graph, classifying comp types, quick victories, ...)
The idea is really good and it's really good how you gathered the data though.
Nice analysis! Just out of interest: Did you also run the regression of prediction on pure outcome? I get the intuition behind your argument that this looses variance in the outcome, but after all only win/loss "really matters"
Just using the binary variable as DV leaves you with R^2 = .05, if I remember correctly.
Considering that LoL is team/indivudual micro/macro/momentum based game, starting with a 6% edge just by picks is huge
Also Caedrel loves to analyse drafts so he can tell us how the game strats will play out, and how efficient is one vs the other ( i'm his lawyer)
Ye, if you see it that way: in a high level competition every fraction of percent counts, then I agree, I might have undervalued the 6% in my interpretation. I guess the point I was making is, that other factors seem to be MUCH more important and Ceadrel seems to put a bit too much emphasis on it. But then again he is an analyst, so what do you expect. He's not gonna tell you how T1's chef is a huge advantage on the rift because Faker got his broccoli today.
The thing is that the outcome of the game is decided by so many factors and not just draft.And Caedrel is a guy that usually that does a lot of mistake cause the outcome of the game is completely random . Especially in LEC where good teams get rolled by bad teams .
Could you put those predictions over a period of time like a tournament to show adaptation?
yo the random stray at caedrel's ori mid i love it
Are you able to release the data? I'd love to see how the R^2 changes with the outliers removed
Sure, but I'm afrtaid it won't be of much use, because I was a little lazy during data collection. Didn't even record which team won or was favoured or when the game was played.
docs.google.com/spreadsheets/d/1uuA30j5P7nYlBND8v4jXfpzYSN_JlmQxD124qPj10TY/edit?gid=0#gid=0
Gold diff is useless as a comparison though. Different comps have different needs for for gold and are designed to come online at different times. Any gold diff at full build is useless.
You also need to look at relative team strength. DFM vs T1 is not going to be determined by draft. Probably should scale the influence of these results in the analysis.
I would take betting odds or analyst predictions before the draft as a baseline. And then check if favourable draft corelates to the team winning more than it should based on odds given by betting sites. Could be your next project. Interesting baseline, would be interesting to refine the data.
Can we have the excel sheet of the data used? Somewhat interested in analyzing it as well!
Sure, but I'm afrtaid it won't be of much use, because I was a little lazy during data collection. Didn't even record which team won or was favoured or when the game was played.
docs.google.com/spreadsheets/d/1uuA30j5P7nYlBND8v4jXfpzYSN_JlmQxD124qPj10TY/edit?gid=0#gid=0
“Malphite would be a mega pick here no?” -our rat in any draft where top isn’t early rotated xdd
Commenting for algorithm
Great video!!
(1) You didn't take into account prior win rates (i.e. which team is more likely to win, e.g. by Riots power rankings). A better methodology would be comparing prediction power of only using that prior, and prediction power of prior + Bayesian update on draft
(2) Gold is not a good indicator of a stomp for several reasons: (a) Total stomps end earlier, potentially leading to less gold lead since there's less time to build it (b) scaling comps don't need huge gold wins to win. The malphite draft is a good example: the malphite team can be down 5k gold but still be favored against a full AD comp (c) gold distribution os important. If your Alistar gets all the kills your gold leas is meaningless. Also champions like TF and Pyke skew gold statistics
In general I think just using a binary variable is fine. If you use anything else you shift the goalpost which inevitably devalues a certain way of winning the game.
Just using the binary variable as DV leaves you with R^2 = .05, if I remember correctly.
I think gold difference might not be the right stat to determine how much better a comp is. A lot of strong comps during the season were scaling comps with smolder ziggs that don't need a gold lead to be completly untouchable in games. Smolders stacks mean the actual power difference between comps is vastly different then gold difference. Using gold difference also unfairly overvalues snowballing comps large gold leads are more common due to the champions usually falling off later on and champions being balanced around being gold inflated.
Im here for the sally content
im at 4:00 and I was thinking how can someone say he predicted it right or wrong. There is so much possibility to alter end of game gold like what if i had a shit draft but i took soul 4 barons and needed 15k gold diff to end. what if i picked super scaling and won in 40 min with 1 fight and in the end it was 3k gold lead but the enemy couldn't do nothing had 0 chance because of the draft and we just sit back and cleared waves traded obj and win in 40 min. I just remember DK at worlds started playing scaling for no reason was it a good draft maybe but for that team it wasnt clearly they needed early game skirmish drafts so we have to look at teams again what if we had a superteam against a last place team and the superteam win with a bad draft just because they are better. Looking only at end of game gold diff will not give a good representation of good or bad draft, sometimes it does but it wont always.
I agree, that gold dif wil not always be a good representtion of how well a comp did, but I think it is a decent proxy. Maybe, as some here have suggested, a gold dif weigted by game length would have been better. That account at least for some of the scenarios you mentioned.
@torr3nt_loading I would first of all check the playing teams they playstyle and the comp playstyle each mu win rate(proffessional) +weight player win rate on the champ + add soloq winrate if more data needed. That gives you multiple information about the comps you can create data based on this like winning lanes early comp type like skirmish scaling sidelane teamfight etc and check the actual early mid and lategame gold diff xp diff/lane obj diff . Then you have to compare the factual data to the one u were expecting (a team plays full early game vs full scaling expected to get gold lead early but if they didnt create much we can say the scaling team comps draft was good because they shut down the early of the enemy) if we evaluate the skill diff of the players then we can get this a bit closer to reality. TLDR check gold,xp,obj diff each lane multiple times in the game and calculate with skill, matchup and playstyle differences.
Gold lead is a misleading variable, as a team could have a major advantage in draft by for example playing passively, and winning a game by 3k gold was a stomp. Ex certain teams with Kalista being even or slightly behind at a certain point in the game they are even further behind than a gold lead can suggest.
Honestly, im not sure the gold is the best value since its the outcome who matters. For instance you could prefer a late-game comp that gets itself 10k behind, but still wins and finish the game at -5k.
Also, looking at the graph (and supposing the - gold is a loss and the +gold is a win), I think the pattern is interesting.
- When he thinks its slightly favored for a team, he is more often wrong. Maybe meaning the teamcomp is a bit worse but they picked confort, etc.
- Seems pretty equal for the "favored"
- Highly favored tho is where its more clear. About 20-7 ratio. So the draft, when clearly won does have a huge impact, according to those few samples.
Would percentage gold diff improve the data set over total gold diff?
If you have a 20k gold lead, but one team has 90k gold and one has 110k gold, is that 20x more significant a factor as a 1k gold lead, when one team has 5500 and the other 4500. (Definitely extreme ends both ways, but that 1k gold lead should at LEAST be considered just as significant as the 20k gold lead.)
If the draft picks were amazing, you might expect an early stomp, which might have a similar percentage gold diff, but a vastly underperforming total gold diff.
i really enjoyed the video and your hypothesis, but i feel like something like this is so hard to objectively evaluate, the human element is also something to take into account and because we are not robots or consistent machines, i often feel like he can be right about a draft and they still lose, or he can wrong and they win, everything has to coexist
How are the stats for gold_diff/gametime?
Thing is, when his predictions and takes does miss, he can recognize that the players can play it differently than he imagined, or if there's other factors and variable that he didn't account for, and when confronted about it, he does address it and took it with grace(as much grace as a rat can have at least). Meanwhile some other streamers, especially *cough*LS*cough*, are so tunnel focused on their own takes and ideas of what the meta is, they don't think they can be wrong. They can't comprehend that the players and teams have their own preference, strategies and playstyle regardless of the meta, and that League is a game with thousands if not millions of variables than can't possibly be perfectly accounted for by just a draft. And when confronted about it, they just don't address it, evades it with something else, or just makes excuses.
Any chance you could chuck he data you used in a google sheet and share it?
Sure, but I'm afrtaid it won't be of much use, because I was a little lazy during data collection. Didn't even record which team won or was favoured or when the game was played.
docs.google.com/spreadsheets/d/1uuA30j5P7nYlBND8v4jXfpzYSN_JlmQxD124qPj10TY/edit?gid=0#gid=0
@@torr3nt_loading
all 1 2 3
correct 59 11 28 20
incorrect 57 20 27 10
wr 0.50862069 0.35483871 0.509090909 0.666666667
p value 0.390369564 0.925193608 0.393853095 0.021386973
basically only meaningful if he expresses strong favour
Not me watching this before my methodology test
Cool Video!
7:00 the draft_eval variable is categorical, ordinal variable, not quantitative. Running a linear regression model on this data for fit is not going to give good results for regression variance/fit.
Wow, uni classes coming in handy
A rat essay? Highly fucking approved, liked and subbed LETSGO
Do LS next?
I m not sure taking the final gold diff is right though, a game might be close until it explodes and they take baron ,3 towers ... But honestly I don't know what other metric you can use (probably something like average gold difference per minute, though it's hard to get this data), another thing no fucking way he only ate 8 times during drafts XD.
hes good at analyzing draft
but theres a curse of the rat king debuff on him
Gamba: even with a slight favor of wrong xdd
Edit: did.. did u just big kiss me without my consent?
Sorry :*
I wonder what the results would be, if he analyzed LS's predictions. XD
Think your entire premise that gold difference is a good predictor of how good a draft was is just entirely flawed. Often times when we are talking about good drafts we mean drafts that can win the game even when several thousands of gold behind. Just the same if we were to think of a way to win despite having inferior drafting, it would probably be by snowballing early and having an insane gold lead. I would even venture a guess that it's a worse indicator of the draft being good than win/loss percentage for the reasons stated above.
Well, winning a game with a negative gold dif is really rare just because the towers during your final push and the team fight you likely just won give some much gold. But you are right, a good "comeback comp" that (in theory) would often end games with a close gold lead is not something the gold dif represents very well. Still, I don't agree that the entire premise is wong, gold dif is higher in onesided games. And if you say that the comp will roll the enemy team, you expect a onesided game and a win with a high gold dif.
@torr3nt_loading I mostly agree, but that always depends on the comp. Simply put this analysis is so reductive as to be nearly useless.
@@LordKeram Entertainment is the main use of this ;)
Did I just watch a league research presentation
Hi great video
5:10 i mean t1 draft is just objectively better
If they didnt get stomped in 20 minutes they wouldve won at 30minutes+ with a 3k gold defecit
Thats also kindof the problem with your evaluation metric
Gold lead doesnt indicate how good a draft is. If your team is down 5k gold and can still win, thats draft diff
Fair, I still think gold_dif a good proxy for how well a team (and with that draft) did. As some have pointed out weighting by game length can offer a more accurate measure.
It was 6%, right?
We talk about the highest level of competition.
+6% Winrate is HUGE in that case.
(Only by the draft diff)
Not in my soloQ😂... in the Lck, LCS, MSI or Wolrds matches.
Statistics is a very interesting subject because we can saw the SAME results way diffrent.
Its not 6% winrate. Favoured comps (all 3 categories) had a roughly 51% winrate. High favour comps had 66.67 WR, but the data set for those are only 30 games.
gold diff is not a good dependent variable. different comps have different gameplans and wincons. the goal of draft is not to win with the highest gold difference, the goal is to win period. I don't buy your justification to use gold diff. even a simple approach with W/L would've been better.
also, Caedrel probably thinks about Win/Loss more than Gold Diff when winning
Doesnt matter gold diff if the item are the same or when enemy team hit their item power spike and its stronger than other
I wonder if he is ever going to become a coach?
Prediction: Its good.
I dont think using gold diff is the correct way to evaluate draft predictions. What i mean by this is lets say that have one team with a better comp and they end up losing, then the fact that they have a good comp probably means that the enemy team requires a larger gold lead to end the game. Whereas the gold diff they themselves require to win is smaller
Gold difference at the end of the game is such a terrible metric.
It is so composition dependent that it is basically worthless.
Not scaling it with game length makes it even more silly.
A 50 Minute slugfest that ends with the scaling draft barely surviving mid game and then slowly choking out the enemy team until they have a 15k gold lead is the same as a 20-0 kill 19 minute win in this metric.
Some drafts with a 70% win rate might actually produce 30% absolute stomp loses and 70% wins with slight gold advantages.
There is no logic behind the idea that a greater gold lead and the average chance of winning have any kind of logical connection of equivalence.
You are discounting the bias that might come from being a favored in the game. If T1 plays FLY and T1 is a 5:1 favorite you don't think there is a chance this might play a role in impacting Caedrels prediction? For example he says T1 won draft. And they win. Was it because T1 had the better draft or because they were better players? And did this info influence Faker.
The easiest way around it would be to only use games between teams that were even odds in your analysis.
Caedrel is very good at predicting picks / drafts based on experience, he's worse at picking the winners since copium exists.
He picks the team based on statistic but teams can change
MODS PAYOUT GAMBA 🦍
i think this research means close to nothing, firstly a team comp is just one small variable amongst hubdreds of others secondly the mechanic of adding and subtracting game gold is very very stupid for many reasons