What's my win rate? Sample graphs from 110K hands!

[Adde]

There has been some posts about win rates lately, so I thought I should investigate my own stats a little closer. Perhaps others are interested, so I post the results here.

When: Dec 1 2004 - Dec 1 2005
Site: PartyPoker
Game: $2/$4 Limit Holdem
Hands: 110K
Win rate: 2.17 BB/100

The problem with finding out your win rate is that the variance can be a bitch (and it will be!). To see what degree of bitchiness it can acheive, I took my PokerTracker data and split it up in chunks of 100 hands (PostgreSQL PG Admin is your friend). I then calculated the win rate for all subsets of 10K hands in the total sample size of 100K. That is, first sample of 10K hands contains hand 1-10000; second sample is hand 101-10100; ...; last sample is hand 100001-110000. This is repeated for all subsets of 20K up to 100K hands.

Each graph below shows how the win rate for a sample size fluctuates around the win rate calculated for all hands (110K). I'm a little sloppy and call the last value a "True" win rate, but only in the meaning of "true win rate for all 110K hands", not my acutal true win rate (which I don't know). There's also Min and Max values for win rate, and a Diff value for Max minus Min (even though the interesting Diff value is Min/Max versus True).

Some random note/s...

(1) Quite often people post or quote their Pokertracker stats after some 10K hands. I've had bad runs of (at least) 10K hands where I have lost money. 10K hands may sound quite much, but it isn't. It may sound strange to Average Joe that a winning player is down for that many hands, but it isn't. The numbers fool you.

(2) I'm tired and hungry.

Adde

this is very interesting stuff, thanks for tkaing the time do it!

[Adde]

More random notes...

(*) Anyone got way more hands (than my 110K), and want to share? If you have converted your PT database to SQL I can give you easy instructions to generate an export file, which I can concert to Excel graphs. It would be interesting to see how the stats for a sample of 100K hands converge for a larger total (say 200K+ hands).

Adde

[Sniper]

It would be useful to put the graphs in context by adding another graph of cumulative winnings in BB, ploted after each 100 hand block.

[Adde]

It would be useful to put the graphs in context by adding another graph of cumulative winnings in BB, ploted after each 100 hand block.

Like this?

("Linjär" = Linear; trend line)

[XChamp]

This is very interesting. Thanks a lot.

[MrEngenic]

You have played 110k 2/4 hands on party in december???
What is your screen name?

[Adde]

You have played 110k 2/4 hands on party in december???
What is your screen name?

110k hands from last december til now, one year, not one month.

Adde

[Sniper]

[ QUOTE ]
Like this?

[/ QUOTE ]

Yes, exactly what I meant... and what I expected it to look like [img]/images/graemlins/wink.gif[/img]

Have you given consideration to any further analysis to help evaluate the short term variations?

For example... how about an overlay of the 1st original chart (the 10K), with expected vs actual results based on your card distribution?

[Adde]

For example... how about an overlay of the 1st original chart (the 10K), with expected vs actual results based on your card distribution?

That's a very interesting thought. Pokerroom has extensive EV data for starting hands on their site which could be used for this. A bit more complicated to generate (edit: the data for) these graphs though, but I'll look into it.

Adde