Learning about Ergodicity via Ole Peters and Taleb was the most interesting thing I learned in a long time. I loved the interesting examples from Incerto the blog: "6 people playing Russian roulette is very different from 1 person playing Russian roulette 6 times" and "The cost-benefit analysis of smoking a single cigarette is positive" (so let that be the only time you ever take a risk :))
What you mentioned about survival reminds me of the dichotomy of finite and infinite games; with a finite game, you're playing to win, but in an infinite game, you're playing to keep playing.
Watch it without trying to understand the math and listen well to the warnings about absorption. The rest is showing that his explanation is provable via mathematics which is only necessary for those who need deeper proof/understanding.
Non-ergodicity is so prevalent in trading that actually launching 2 versions of the same automatic trading strategy now and then 5 minutes later will end up with drastic different results after several months of trading, like one of version of the strategy can do 50 % and then other lose 30 %. I mean nothing changes, same parameters, just when the trading It started can change drastically the final results after compounding. Something that I never see discussed in papers, videos or books on trading and it is such a fundamental fact about trading.
This probably implies the strategy is not robust. It may be that it is an overfit backtest or it could be one that is sensitive to timing luck. You might be able to lower timing luck by trading half your bet size now and the other half 5 minutes later. An ensemble of trading strategies each initiated at different times. I think there are one or two articles on this topic.
@@lukedali4999 did you try to test this yourself with what you consider a robust strategy? I think what I described is typical of almost any strategy if it is compounding the returns, it is a property of any process based on the product of the returns. Please share a particular well known strategy that doesn't behave like this and I will test it. What you suggest to deal with this issue is what I actually do that is starting n instances of the same strategy at subsequent times, with 1/n of the total capital for each instance and then rebalancing between the instances. But this issue is never discussed in trading books or journals as far as I know.
it was missed? I, as non statistician, always thought it weird you would "expect to be at zero" no matter the deviation, while also assuming that you can't converge towards zero from that point. if that makes sense.
I would like examples of ergodic games/ investments . And how to turn a non ergodic investment into a ergodic one. This Law of Large Numbers was discovered by Komolgorov , wasn't it?
Learning about Ergodicity via Ole Peters and Taleb was the most interesting thing I learned in a long time. I loved the interesting examples from Incerto the blog: "6 people playing Russian roulette is very different from 1 person playing Russian roulette 6 times" and "The cost-benefit analysis of smoking a single cigarette is positive" (so let that be the only time you ever take a risk :))
nice, where have you read this excellent example?
What you mentioned about survival reminds me of the dichotomy of finite and infinite games; with a finite game, you're playing to win, but in an infinite game, you're playing to keep playing.
Needed to watch this 20x over
Watch it without trying to understand the math and listen well to the warnings about absorption. The rest is showing that his explanation is provable via mathematics which is only necessary for those who need deeper proof/understanding.
Never thought about the difference between long time periods and multiple worlds. Thanks for the enlightenment :)
So is there a number or a measure that tells us how much non-ergodic a system is?
Non-ergodicity is so prevalent in trading that actually launching 2 versions of the same automatic trading strategy now and then 5 minutes later will end up with drastic different results after several months of trading, like one of version of the strategy can do 50 % and then other lose 30 %. I mean nothing changes, same parameters, just when the trading It started can change drastically the final results after compounding. Something that I never see discussed in papers, videos or books on trading and it is such a fundamental fact about trading.
This probably implies the strategy is not robust. It may be that it is an overfit backtest or it could be one that is sensitive to timing luck. You might be able to lower timing luck by trading half your bet size now and the other half 5 minutes later. An ensemble of trading strategies each initiated at different times. I think there are one or two articles on this topic.
@@lukedali4999 did you try to test this yourself with what you consider a robust strategy?
I think what I described is typical of almost any strategy if it is compounding the returns, it is a property of any process based on the product of the returns. Please share a particular well known strategy that doesn't behave like this and I will test it.
What you suggest to deal with this issue is what I actually do that is starting n instances of the same strategy at subsequent times, with 1/n of the total capital for each instance and then rebalancing between the instances. But this issue is never discussed in trading books or journals as far as I know.
If there are articles on the subject and you know the reference I would appreciate that, I searched and I could not find anything relevant.
really deep insight
it was missed? I, as non statistician, always thought it weird you would "expect to be at zero" no matter the deviation, while also assuming that you can't converge towards zero from that point. if that makes sense.
Thanks 🙏🏻
Interesting. Thanks for sharing.
I would like examples of ergodic games/ investments . And how to turn a non ergodic investment into a ergodic one.
This Law of Large Numbers was discovered by Komolgorov , wasn't it?
Thank you
Little confused on this but I'll investigate further. Thanks for breaking this down though nassim.
For anyone studying this topic, I found this to be a handy introduction (less technical):
taylorpearson.me/ergodicity/
Thank you for this! This is the first article that really helped me grasp this concept