Data Scientist Explains How DUPR Algorithm Works

Поделиться
HTML-код
  • Опубликовано: 27 ноя 2024

Комментарии • 24

  • @ccbowers
    @ccbowers Месяц назад

    1:33 There is another video that contradicts this idea. Another DUPR data guy said that W-L comes first and you can't go up in DUPR if you lose, even if you exceed expectations and that you can't drop if you win, even if you really underperformed

    • @EddieAndWebby
      @EddieAndWebby  Месяц назад

      DUPR has evolved over the years. Any time a new video comes out about it, there is likely new features for the updated version of the DUPR app and algorithm. -Webby

    • @ccbowers
      @ccbowers Месяц назад

      ​@@EddieAndWebby I did look it up and did see the algorithm has changed in this exact way. Win vs loss comes first, and only after than does performance relative to expectations matter. I don't love this change, especially for newer players or players who play less frequently. For higher level players who play alot it probably makes sense

  • @omid2481
    @omid2481 8 месяцев назад +1

    Hi Tnx for this information. Has DUPR lookup table been made available? After a year from this video I would have expected it to have been out by now but cannot find it anywhere. Any tips is appreciated Tnx

    • @ccbowers
      @ccbowers Месяц назад

      @@omid2481 Algorithm changed so I wonder if that changed the release of that table

  • @davidallott5826
    @davidallott5826 Год назад +2

    7 months since this interview and I see no sign of the lookup tool. Would be very helpful to know the algorithms expectation before calling 0-0-2. Any update from the company on this?

  • @guymoretti1599
    @guymoretti1599 Год назад

    Why do Dupr scores lower when no activity or matches are being evaluated? My DUPR was 4.66 8 months ago and has been consistently dropping when no activity is being recorded. My current rating is now 4.19. Last week was 4.22.

    • @JB-by8tw
      @JB-by8tw Год назад +1

      Just a guess but maybe those players in your match history have had losing results recently (post your match against them).

    • @bryanyee7
      @bryanyee7 Год назад

      Maybe it assumes that if there are no recent matches, that a player hasn't been playing recently and may be a bit rusty and not as sharp?

    • @mattc3581
      @mattc3581 Год назад

      @@bryanyee7 I don't believe there is any natural decay built into the system. You shouldn't lose rating just over time.
      I also don't believe it takes subsequent matches your opponent's have played into account, your rating gain or loss should be based on what their rating was when you played them. That being said if they had outstanding results to go on from before you played them that altered their rating at the time you played them then when those results go on I would expect the system to rework your rating based on the new rating for the guys you played when you played them.
      Many tournament results seem to take months going on the system, so if they do backdate them into correct chronological order then this may be the problem. This could also happen if they consolidate players with multiple accounts into one.
      The other possibility is that they seem to have a system where your initial rating is based on some reference games against known players. These games don't seem to have any gain or loss of rating for the people involved since at least one player is a complete unknown. If they have retrospectively re-evaluated which games were reference games for some people then whole games may have been added or taken out of your rating history causing an adjustment.
      Honestly if it was just that prior games are still working through the system I think DUPR should tell people to live with it. Pandering to people not wanting their rating to change ever, even if it should, actually creates a lot more problems. But in reality it seems there are more issues than just this retrospective ratings adjustments. If you look at match history some people have ratings that just go up or down wildly between matches, for no apparent reason. I've seen one player who starts at 3.6, loses five matches in a row, losing rating points in every one, and ends up a 3.7.

  • @Thejoeb
    @Thejoeb 2 года назад

    Not sure why I lose rating when I dont log matches, but I started at 4.569 and now see 4.328 when I have logged zero matches (and apparently unsanctioned tournament wins dont count? I got 2 silvers at 4.5 since making my account...) not sure why a player would lose rating over time if they have logged zero official matches, I certainly havent gotten worse in the last month, so why does my rating go down for inactivity and not take into account non sanctioned tournament scores?
    makes it dubious at best for accuracy, and Im not interested in logging rec play matches.

    • @EddieAndWebby
      @EddieAndWebby  2 года назад +4

      Based on what I've learned so far, our rating can move even when we don't play because the ongoing performance and rating of others that we've played against is constantly being take into consideration. As time goes on and the algorithm kicks in and learns, our ratings are supposed to become more accurate. I've also found that a lot of the most frequent questions are covered on their support site:
      www.support.mydupr.com/
      -Webby

  • @pickleballtechniques
    @pickleballtechniques Год назад

    It is very weird that you can, and have to enter tournament data into DUPR system by youself, which should be closed to public. doing so as in golf hadicap system, it is so easy for players to do whatever they need so that doing sandbagging easily.

    • @EddieAndWebby
      @EddieAndWebby  Год назад

      Usually, the only tournament results you have to input yourself are smaller local tournaments. Any tournament in pickleball tournaments dot com gets input automatically. -Webby

    • @davidallott5826
      @davidallott5826 Год назад

      I'm not sure if it works differently for tournaments, but whenever I've used DUPR, all results have to be validated by an opponent.

    • @EddieAndWebby
      @EddieAndWebby  Год назад

      For sanctioned tournaments (or tournaments that have an arrangement with DUPR) it doesn't need to be validated by the players and will get added automatically. All rec play or smaller, local, tournaments matched will require player validation.

    • @davidallott5826
      @davidallott5826 Год назад

      @@EddieAndWebby Right, so depending if its a sanctioned tournament or not, either the tournament admins enter the results, or you do yourself and your opponent has to validate. Either way, its not exactly easy to sandbag your rating.

    • @EddieAndWebby
      @EddieAndWebby  Год назад

      Correct. It would require somebody with a reliable rating (not a provisional rating like everybody starts with) to sacrifice their own rating in order for any shenanigans to happen. I don't know too many people willing to do that to help alter somebody else's rating, lol. -Webby

  • @hermelyndapickleball
    @hermelyndapickleball 2 года назад

    Thanks!

  • @DannyBPlays
    @DannyBPlays 5 месяцев назад

    Good vid but guy looks ridiculous wearing sunglasses indoors and looking at a computer screen

    • @EddieAndWebby
      @EddieAndWebby  5 месяцев назад

      It used to be our thing. We don't do it as much these days. -Webby

  • @HardyHarz
    @HardyHarz 2 года назад +1

    Totally get what they're trying to do, but weighing point spread more so than wins/losses is a little counter-productive. It undervalues a player's/team's ability to get the win, which is ultimately the most important thing in any sport. With DUPR's method, the rating system is biased toward players who have perhaps one elite skill and play primarily in round-robin format matches. For example, if Team A has a player who has a wicked spin serve, or an exceptional speed-up shot, they may win games by a wide margin. In match play, Team B may be able to figure out the skill and actually win the best of 3, and ultimately be the better team/players. In this case, let's say Team A wins the first game 11-0, but loses the second 9-11, and the third 8-11. With the current system, Team A is rated as the better team because they scored a total of 28 pts to Team B's 22, for an overall point spread of 6, but Team B is actually the better team with the better players (and should have the higher rating).
    This sort of thing recently happened to my partner and me in a recent 4.5-level tournament. My partner and I won gold and our ratings went down, while the team that took silver (that we beat) had their ratings go up, This was due to them having a higher point spread in their matches, even though our record was better (and the fact that we beat them in the head to head match). At the end of the day, it doesn't REALLY matter how many points you win by, only if you win, so I think they should really weigh the wins/losses as a primary indicator and use the point spread as a secondary method of rating players.

    • @cadegirolamo2338
      @cadegirolamo2338 Год назад

      You are partially correct. But over as the number of games used to compute the DUPR rating increases, the more accurate it becomes based on the expected points v actual points logic. The converse of your point is if my opponents are a 4.0 and I am a 3.5, shouldn't my rating improve even though I lose 11-9?