Posted: Wed Jun 06, 2012 5:46 pm Post subject: Re: New Article by Alex Rousso - Luck Horizon
Quote:
We can elucidate further by considering what factors should affect luck horizon. First and foremost is obviously your skill level in the game you’re playing. Just to be clear here: if you have no edge, you have infinite variance – your results graph will just be a random walk.
I'm not sure I agree with this; at a win-rate of 0bb/100 variance will be similarly non-infinite to that at 0.5bb/100. In this case your "luck horizon" will be the time at which you can be reasonable sure you have exactly no edge, I think it applies similarly to winners, losers, and those on the border. _________________ http://www.alexbowler.com
Posted: Wed Jun 06, 2012 10:07 pm Post subject: Re: New Article by Alex Rousso - Luck Horizon
I agree with Alex on his point and I don't agree that this is a useful way to decide whether or not you should be playing or what games you should be playing. There is more to it than luck and working it out. There is the judgement of whether you are getting your ass kicked or whether you are kicking ass. Luck horizon ROI will never apply in any individual game it is statistical and it is also never constant. Nonetheless it was an interesting and informative article as ever.
Posted: Thu Jun 07, 2012 3:51 pm Post subject: Re: New Article by Alex Rousso - Luck Horizon
Poker players have taken to using the term variance incorrectly, they usually mean downswing or bad luck. I also don't like using the term 'negative variance' I prefer to keep the simpler downswing.
As far as variance goes nearly all of it is built into the choice of game as you mention but a player with a zero edge would not have infinite variance just the amount expected for a zero roi. If the player was so bad that even a numbskull could beat him/her the variance would actually approach zero due to nearly always losing. It is hard to avoid or reduce variance in any set game and trying too hard to do this is just a way of reducing your roi. There are times in tournaments when you may have such a great edge that you would be advised to avoid a high chip variance play but even here it is easy to give up too much of your starting edge. An interesting thing to note is in standard 9 seat SnG with a payout (50%,30%,20) and realistic results eg, fairly flat finish distribution you actually would like the variance of your results to rise, it would reach a peak at an ROI of about 50% and as this is greater than any non cheat so you should welcome the higher variance.
Posted: Tue Jun 19, 2012 5:22 am Post subject: Re: New Article by Alex Rousso - Luck Horizon
Yes, I was wrong about the 0% ROI = infinite variance. Well pointed out, Alex.
There is a special case where a player is so bad that they always lose, but since their results aren't normally distributed, they are not subject to variance in the same sense (put it another way, if you wanted to avoid variance at poker, you could deliberately lose (by folding!). But that's only a special case. _________________ Twitter: PicklemanPoker
Pickleman articles: http://www.bluffeurope.com/poker-news.../pickleman-poker/
Posted: Tue Jun 19, 2012 10:16 am Post subject: Re: New Article by Alex Rousso - Luck Horizon
with a non-finite set of events, a non-finite set of skill levels and knowable skill levels how do you arise at 96% +/-5 over 200,000. I hate SD to be honest but i still dont see how it can be anything but arbitrary in this instance(if not all).It does not seem to encompass any aspects of the 200,000 hands and almost hints at a million hands before to obtain a steady enough playing style to win 4bb/100 to make the distribution justifiable. _________________ The next best thing to gambling and winning is gambling and losing.
Posted: Tue Jun 19, 2012 6:34 pm Post subject: Re: New Article by Alex Rousso - Luck Horizon
darrensprengers wrote:
with a non-finite set of events, a non-finite set of skill levels and knowable skill levels how do you arise at 96% +/-5 over 200,000. I hate SD to be honest but i still dont see how it can be anything but arbitrary in this instance(if not all).It does not seem to encompass any aspects of the 200,000 hands and almost hints at a million hands before to obtain a steady enough playing style to win 4bb/100 to make the distribution justifiable.
Hi Darren,
Sorry I didn't really get this.
The comment in the article on this was a putative statement. That is to say, we could make statements of the ilk: "with a 200,000 sample and a 4bb/100 win rate over that sample, we could be (although this was just a guess - an example) 96% certain that this win rate was correct to within +/-10% (i.e. we can be 96% certain that the "actual" win rate of that player was between 3.6bb/100 and 4.4bb/100).
Once again, that was just a made up statement to give an example of the kind of thing we should be saying about variance in poker, rather than "I'm on a downswing at the moment" and "that game's really swingy". _________________ Twitter: PicklemanPoker
Pickleman articles: http://www.bluffeurope.com/poker-news.../pickleman-poker/
Posted: Thu Jun 21, 2012 9:07 am Post subject: Re: New Article by Alex Rousso - Luck Horizon
pickleman wrote:
darrensprengers wrote:
with a non-finite set of events, a non-finite set of skill levels and knowable skill levels how do you arise at 96% +/-5 over 200,000. I hate SD to be honest but i still dont see how it can be anything but arbitrary in this instance(if not all).It does not seem to encompass any aspects of the 200,000 hands and almost hints at a million hands before to obtain a steady enough playing style to win 4bb/100 to make the distribution justifiable.
Hi Darren,
Sorry I didn't really get this.
The comment in the article on this was a putative statement. That is to say, we could make statements of the ilk: "with a 200,000 sample and a 4bb/100 win rate over that sample, we could be (although this was just a guess - an example) 96% certain that this win rate was correct to within +/-10% (i.e. we can be 96% certain that the "actual" win rate of that player was between 3.6bb/100 and 4.4bb/100).
Once again, that was just a made up statement to give an example of the kind of thing we should be saying about variance in poker, rather than "I'm on a downswing at the moment" and "that game's really swingy".
i think i was thinking of 5 things at once but the main one is i don't think SD is necessarily the right way to think about distribution of hands as the amount of variable information is so great that it is more likely to show you how accurate your bb/100 in the 200k hands is rather than any future 200k hands. Considering you have the expected value of hands then surely a better mathematical analysis would be of the 200k hands and try and find trends that can be exploited in iteration. These trends, at least theoretically, are continually changing. Not necessarily improving, mathematically anyway, but would alter what is a successful strategy. Considering the uptake of poker recently this has coincided with a massive increase in poker ability but it is not exclusive to an increase of ability i.e.a 2.5x raise has many mathematical advantages but the vast majority of people do it force a trend and strategy change. _________________ The next best thing to gambling and winning is gambling and losing.
Joined: 23 Oct 2008 Posts: 1682 Location: Birmingham
Posted: Thu Jun 21, 2012 9:24 am Post subject: Re: New Article by Alex Rousso - Luck Horizon
darrensprengers wrote:
pickleman wrote:
darrensprengers wrote:
with a non-finite set of events, a non-finite set of skill levels and knowable skill levels how do you arise at 96% +/-5 over 200,000. I hate SD to be honest but i still dont see how it can be anything but arbitrary in this instance(if not all).It does not seem to encompass any aspects of the 200,000 hands and almost hints at a million hands before to obtain a steady enough playing style to win 4bb/100 to make the distribution justifiable.
Hi Darren,
Sorry I didn't really get this.
The comment in the article on this was a putative statement. That is to say, we could make statements of the ilk: "with a 200,000 sample and a 4bb/100 win rate over that sample, we could be (although this was just a guess - an example) 96% certain that this win rate was correct to within +/-10% (i.e. we can be 96% certain that the "actual" win rate of that player was between 3.6bb/100 and 4.4bb/100).
Once again, that was just a made up statement to give an example of the kind of thing we should be saying about variance in poker, rather than "I'm on a downswing at the moment" and "that game's really swingy".
i think i was thinking of 5 things at once but the main one is i don't think SD is necessarily the right way to think about distribution of hands as the amount of variable information is so great that it is more likely to show you how accurate your bb/100 in the 200k hands is rather than any future 200k hands. Considering you have the expected value of hands then surely a better mathematical analysis would be of the 200k hands and try and find trends that can be exploited in iteration. These trends, at least theoretically, are continually changing. Not necessarily improving, mathematically anyway, but would alter what is a successful strategy. Considering the uptake of poker recently this has coincided with a massive increase in poker ability but it is not exclusive to an increase of ability i.e.a 2.5x raise has many mathematical advantages but the vast majority of people do it force a trend and strategy change.
Joined: 12 May 2010 Posts: 2884 Location: nil fhios agam
Posted: Thu Jun 21, 2012 9:43 am Post subject: Re: New Article by Alex Rousso - Luck Horizon
Heading off to work now quietly laughing to myself, thanks SF. _________________ "I find it's best to start the day with no composure and then you've got nothing to lose." B. Boatman
Joined: 02 Feb 2007 Posts: 3146 Location: with position on you
Posted: Thu Jun 21, 2012 7:45 pm Post subject: Re: New Article by Alex Rousso - Luck Horizon
darrensprengers wrote:
pickleman wrote:
darrensprengers wrote:
with a non-finite set of events, a non-finite set of skill levels and knowable skill levels how do you arise at 96% +/-5 over 200,000. I hate SD to be honest but i still dont see how it can be anything but arbitrary in this instance(if not all).It does not seem to encompass any aspects of the 200,000 hands and almost hints at a million hands before to obtain a steady enough playing style to win 4bb/100 to make the distribution justifiable.
Hi Darren,
Sorry I didn't really get this.
The comment in the article on this was a putative statement. That is to say, we could make statements of the ilk: "with a 200,000 sample and a 4bb/100 win rate over that sample, we could be (although this was just a guess - an example) 96% certain that this win rate was correct to within +/-10% (i.e. we can be 96% certain that the "actual" win rate of that player was between 3.6bb/100 and 4.4bb/100).
Once again, that was just a made up statement to give an example of the kind of thing we should be saying about variance in poker, rather than "I'm on a downswing at the moment" and "that game's really swingy".
i think i was thinking of 5 things at once but the main one is i don't think SD is necessarily the right way to think about distribution of hands as the amount of variable information is so great that it is more likely to show you how accurate your bb/100 in the 200k hands is rather than any future 200k hands. Considering you have the expected value of hands then surely a better mathematical analysis would be of the 200k hands and try and find trends that can be exploited in iteration. These trends, at least theoretically, are continually changing. Not necessarily improving, mathematically anyway, but would alter what is a successful strategy. Considering the uptake of poker recently this has coincided with a massive increase in poker ability but it is not exclusive to an increase of ability i.e.a 2.5x raise has many mathematical advantages but the vast majority of people do it force a trend and strategy change.
I agree with this although we agree on so many things I am starting to think that I have a siamese twin somewhere that no bugger told me about and I don't want any more family so sod off Spengers.
I have always said that SD is too vague in poker because of the constantly shifting dynamic. This is where earn rates can look great on paper even over large sample sizes but it still only reflects an earn rate that was created over a certain number of hands that had certain playing dynamics. Even the slightest shift in that can totally alter bb/100 over future sample sizes.
Speaking hypothetically then let us assume that a player could play 1 million hands in a day and made 4ptbb/100 over that sample. Clearly you can attach a high level of confidence to that earn rate but only in so much that he earned 4ptbb/100 for that million hands. If that was done on say June 12th 2011 then how relevant would that 4ptbb/100 be should this player try to replicate that a full year into the future where the game dynamics would be different??
Posted: Fri Jun 22, 2012 9:47 pm Post subject: Re: New Article by Alex Rousso - Luck Horizon
The Dean wrote:
Speaking hypothetically then let us assume that a player could play 1 million hands in a day and made 4ptbb/100 over that sample. Clearly you can attach a high level of confidence to that earn rate but only in so much that he earned 4ptbb/100 for that million hands. If that was done on say June 12th 2011 then how relevant would that 4ptbb/100 be should this player try to replicate that a full year into the future where the game dynamics would be different??
You would need to also give the SD of these million hands to really say much about it.
SD is important and can be used to analyse some poker stuff quite well - it is very true that you often need lots of data to get a high % accuracy but that is really the point of it - how else would you know you need lots of data. Using stdev you can get a measure of how wide the range is likely to be.
If somebody asks for staking and they have a 10% roi over 10,000 games at SnGs over the previous year it might seem like a good bet. If these games were all 9 seat, 50,30,20 structure it is a good, reasonably safe bet (nb: as you mentioned so long as several other variables haven't changed) but if these are all at 180 seat SnGs it would be a very risky proposition. The reason is that the variance (ie, the SD squared) is so much higher that the chance of downswings in the future is much, much, higher and also the previous years results can mean it can be a 'true' -ve 10% roi player on a heater. One easy way to decide is to look at the SD of the situation but you do need to understand some maths to make sense of it.
I am sure there are other ways to analyse data more accurately but you would likely need the help of a maths expert and even in these other situations it is a good bet that the stdev is an important predictor variable.
(Incidentally 10,000 games at 10% for 50,30,20 structure with a typical finish distribution will give the roi as 10% +/- 3.2% for a 95% confidence interval, or it is likely to be outside this range 1 time in twenty)
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
We use cookies to make your experience on our website better. To comply with the new e-Privacy Directive, we need to let you know about this. By using our site with cookies enabled in your browser, you agree to their use. You can read more about this on the cookie information page. You can hide this message.
FAQ
Search
Memberlist
Profile
Log in to check your private messages
