Hands to determine skill level

[donkontilt]

On average how many hands does it take to determine weather or not you are a winning player? I think i say 50-100K around here somewhere. Is that correct?

[basementproject]

That's enough.

[SellingtheDrama]

Personally I think its infinite - you should always be learning and adjusting, thereby increasing your skill level. If you aren't, you are basically a robot.

[basementproject]

At some point you need to stop and evaluate your play, to determine whether you suck, whether you're good, or whether you suck and are running really hot.

50k hands is a good place to do that.

If you go by the 'infinite' theory, you could play crappy poker infinitely, and lose infinitely. Not a good plan.

[pzhon]

[ QUOTE ]
On average how many hands does it take to determine weather or not you are a winning player? I think i say 50-100K around here somewhere. Is that correct?

[/ QUOTE ]
No, generally it takes far fewer hands to determine whether you are winning or not from the results alone, and more hands to estimate your win rate accurately.

It can take more hands to determine whether you are a winner if your win rate is close to 0. It can take only a few thousand hands when your win rate is far from 0, or if you pay attention to more than just the results.

A common mistake is to imagine there is some magic threshold, below which no stats are meaningful, and above which all stats are meaningful. The reality is that as you gather more data, you gradually get better estimates of all of your statistics. Some come into focus sooner than others. The picture is never perfect. It takes different numbers of hands to answer different questions with different amounts of accuracy.

I define the long run to be the time it takes for breaking even to be 2 standard deviations away from the mean. If your win rate per period (per hour or per 100 hands) is WR, and your standard deviation per period is SD, then the long run is 4 * (SD/WR)^2 periods. If you play live limit games, and win 1 BB/hour with a typical standard deviation of 10 BB/hour, then the long run for you is 4 * (10/1)^2 hours = 400 hours ~ 10,000-12,000 hands.

If your win rate is half as much, the long run increases by a factor of 4.

If you play for 1 long run, there is about a 50% chance that your rough 95% confidence interval, results +- 2 standard deviations, will not contain 0, allowing you to determine (usually correctly) that you are a winning player, or that you are a losing player. If you play for 4 long runs, then the probability that 0 is contained in the rough 95% confidence interval is under 2.5%, so over 97.5% of the time, you can decide whether you are a winning player from the results alone.

[donkontilt]

[ QUOTE ]
[ QUOTE ]
On average how many hands does it take to determine weather or not you are a winning player? I think i say 50-100K around here somewhere. Is that correct?

[/ QUOTE ]
No, generally it takes far fewer hands to determine whether you are winning or not from the results alone, and more hands to estimate your win rate accurately.

It can take more hands to determine whether you are a winner if your win rate is close to 0. It can take only a few thousand hands when your win rate is far from 0, or if you pay attention to more than just the results.

A common mistake is to imagine there is some magic threshold, below which no stats are meaningful, and above which all stats are meaningful. The reality is that as you gather more data, you gradually get better estimates of all of your statistics. Some come into focus sooner than others. The picture is never perfect. It takes different numbers of hands to answer different questions with different amounts of accuracy.

I define the long run to be the time it takes for breaking even to be 2 standard deviations away from the mean. If your win rate per period (per hour or per 100 hands) is WR, and your standard deviation per period is SD, then the long run is 4 * (SD/WR)^2 periods. If you play live limit games, and win 1 BB/hour with a typical standard deviation of 10 BB/hour, then the long run for you is 4 * (10/1)^2 hours = 400 hours ~ 10,000-12,000 hands.

If your win rate is half as much, the long run increases by a factor of 4.

If you play for 1 long run, there is about a 50% chance that your rough 95% confidence interval, results +- 2 standard deviations, will not contain 0, allowing you to determine (usually correctly) that you are a winning player, or that you are a losing player. If you play for 4 long runs, then the probability that 0 is contained in the rough 95% confidence interval is under 2.5%, so over 97.5% of the time, you can decide whether you are a winning player from the results alone.

[/ QUOTE ]

great post thanks

[Boise123]

pzhon, That is great post, articulate and on the money. Thanks for input in the beginners forum.