Questions about winrate and sample size

[joker122]

My situation is this: for ~85k hands I was winning at Xbb/100 hands in the 5/10 6max games at party. During my last 48k hands, I am winning at (X/2)bb/100 hands. In other words, over my last 48k hands my win rate is half of what it was during my first 85k hands.

I was wondering if anyone else has had a similar experience...that is, has anyone seen a big drop in thier winrate during a stretch of hands that's significant (i.e. at least 35k hands)?

I guess I'm just trying to make sense of it and figure out what to expect. Thanks.

[dave44]

Well first off, the fact that it halved means nothing in just that context. Dropping from 6 BB/100 to 3 BB/100 is much more significant than dropping from 1 BB/100 to .5 BB/100.

With that said, the standard deviation for 85K hands is simply your SD/100 divided by sqrt(850).

I don't know what your SD is, but mine is about 16.5. Thus, after 85K hands the SD of your winrate is .566 and after 48K hands the SD of your winrate is .753.

Suppose your actual true winrate is 2 BB/100. If you ran 1 SD above your true winrate for the first 85K, you'd be at 2.5. If you ran 1 SD below your winrate for the next 48K, you'd be at 1.25, exactly half. Running outside of 1 SD will happen about 1/3 of the time so yes your situation is certainly plausible and likely not worth much concern.

[sethypooh21]

I think OP's concerns really illustrate the fact that people still have trouble grappling with the consequences of a situation where even the best players have a SD of at least 5x their win rate (3 bb/100, 15 SD is a godlike player, if you ask me...), and the implications that has for fluctuation in win rates. When people respond to 10k stat posts with "sample size" they aren't just being dicks, the math is pretty simple, even relatively low SD player like the example above needs about 360k hands to cabin the true WR withing .5bb in either direction with 95% certainty.

[sthief09]

maybe it's because you didn't play or read for like 4 months

[mperich]

I ran 60k hands at about 4.5/100, my next 30k have been at 1.1/100

I'd say you are perfectly normal, variance sucks =P

-Mike

[Zygote]

[ QUOTE ]
My situation is this: for ~85k hands I was winning at Xbb/100 hands in the 5/10 6max games at party. During my last 48k hands, I am winning at (X/2)bb/100 hands. In other words, over my last 48k hands my win rate is half of what it was during my first 85k hands.

I was wondering if anyone else has had a similar experience...that is, has anyone seen a big drop in thier winrate during a stretch of hands that's significant (i.e. at least 35k hands)?

I guess I'm just trying to make sense of it and figure out what to expect. Thanks.

[/ QUOTE ]

You are more than likely just experiencing regression. Your results are continually approaching your true earn, which may be much lower or higher than your first 85K will illude you to believe. Your play might have changed, i don't know, but if it hasn't, then, the change in your winrate is due to regression alone. BTW, the amount you regressed is easily concievable and your winrate is still likley to make new significant changes.

[sthief09]

[ QUOTE ]
I ran 60k hands at about 4.5/100, my next 30k have been at 1.1/100

I'd say you are perfectly normal, variance sucks =P

-Mike

[/ QUOTE ]


this all at 10/20?

[mperich]

correct, all at the 10/20. Maybe I just suck now =P

-Mike

[sthief09]

wow, how many tables do you play

I'm good at reading my datamines. I had you pegged as possibly the best 10/20 regular before I knew who you were.

[mperich]

4 tables pretty much exclusively. I like to read 2+2 and chat on msn too much to play 6 tables /images/graemlins/wink.gif

-Mike