The biggest fish

[pottie]

One of my favourite statistics in Pokertracker is to check under "Summary" which players are the really big fish.

I've only got 3000 hands in Pokertracker. My biggest fish caught so far: I dont know if i should mention his name now... anyway he lost $84 in 368 hands at a whopping (22.9) BB/100.

I guess you guys with 100k hands + must have some really interesting stats here...

[TwoShedsJackson]

I have one who's lost over $800 in 1100 hands of .5/1.

[IsaacW]

[ QUOTE ]
One of my favourite statistics in Pokertracker is to check under "Summary" which players are the really big fish.

I've only got 3000 hands in Pokertracker. My biggest fish caught so far: I dont know if i should mention his name now... anyway he lost $84 in 368 hands at a whopping (22.9) BB/100.

I guess you guys with 100k hands + must have some really interesting stats here...

[/ QUOTE ]

I guess if you had recent stats on me you'd call me a fish too. Last night I played ~270 hands of $1/2 and lost 47.5 BB, or a GINORMOUS -17.6 BB/100. How did this happen? I caught a lot of draws that missed and I caught some huge hands that were up against monsters.

The winrate statistic does not begin to be significant until at least 100k hands, and you should be very careful when using it to classify someone as a fish. You are better to use VP$IP (Voluntarily Put $ In Pot), PFR (Preflop Raise %), and AF-TOT (total postflop aggression). My stats for those 270 hands were VP$IP 19, PFR 9, and AF-TOT 1.9. I'm not saying I'm any good ($1/2 has kicked my ass everytime I've tried it), but certainly not a fish.

Using winrate over a small sample to judge someone's ability is a good example of results-oriented thinking. Don't do it.

[__Q__]

[ QUOTE ]
The winrate statistic does not begin to be significant until at least 100k hands,

[/ QUOTE ]

The winrate statistic becomes significant long before 100k hands.

[k000k]

[ QUOTE ]
[ QUOTE ]
The winrate statistic does not begin to be significant until at least 100k hands,

[/ QUOTE ]

The winrate statistic becomes significant long before 100k hands.

[/ QUOTE ]

Not 'significant', just kind of an ok idea. And not 'long' before either. At 20k hands, I could have a great run and win 200bb quickly, bumping up my rate by a whole BB/100. I've have +200bb runs in 1000 hands before. +200 BB moves your winrate .2bb/100 even at 100k hands.

You can see how many hands you need to meet specified confidence intervals using bisinbison's WinRate spreadsheet. At 20k hands, you can't even be 97% sure you're a winning player with a winrate of 3bb/100 and a normal SD! (normal for .5/1 anyway)

[__Q__]

If jumped into the 30/60 and played 10k hands and your winrate was -10 BB/100 hands. Do you?

a)keep playing, you have to play 90k more hands before your win rate means anything

or

b) take a step back. reevaluate your skill level. try to find and fix holes in your game.


People always get carried away with this, "when have I played enough hands" question.

If you want to know some kind of magic number that is your "true win rate" exactly to 10 decimal places, then you are correct. You need a whole hell of a lot of hands. 10k or 20k won't cut it.

But your win rate at a level after your first 10k hands is significant peice of information. For example, if after 10k hands, your winrate is 3.0 BB/100, then you are probably capable of playing 100,000 hands at this level and turning a profit. If your BB/100 is -10.0, more likely than not you've got some major holes in your game. Also, you don't have to be 100 percent certain that a statistic is 100 percent accurate for it to be useful.

Finally, if you do have 100k hands, you probably aren't the same player that you were 90k ago. So should you really include the first 10k winrate calculations?

You can never know that theoretical "true winrate." Even after 100k hands, its still a guess. But, you can take valuable information away from such calculations "long before 100k hands."

[IsaacW]

[ QUOTE ]
If jumped into the 30/60 and played 10k hands and your winrate was -10 BB/100 hands. Do you?

a)keep playing, you have to play 90k more hands before your win rate means anything

or

b) take a step back. reevaluate your skill level. try to find and fix holes in your game.


People always get carried away with this, "when have I played enough hands" question.

If you want to know some kind of magic number that is your "true win rate" exactly to 10 decimal places, then you are correct. You need a whole hell of a lot of hands. 10k or 20k won't cut it.

But your win rate at a level after your first 10k hands is significant peice of information. For example, if after 10k hands, your winrate is 3.0 BB/100, then you are probably capable of playing 100,000 hands at this level and turning a profit. If your BB/100 is -10.0, more likely than not you've got some major holes in your game. Also, you don't have to be 100 percent certain that a statistic is 100 percent accurate for it to be useful.

Finally, if you do have 100k hands, you probably aren't the same player that you were 90k ago. So should you really include the first 10k winrate calculations?

You can never know that theoretical "true winrate." Even after 100k hands, its still a guess. But, you can take valuable information away from such calculations "long before 100k hands."

[/ QUOTE ]
Regardless, the point of my original post is that you cannot use winrate to categorize players. A sample size of a few hundred hands is far too small to get any read on a player's winrate. However, such a sample is large enough to get a fairly good idea of their VP$IP and PFR, so these are the statistics we should use when trying to classify a player with only a few hundred observed hands. Even better is to actually look at the kinds of hands they were playing and develop a strong read on them.

Unless you are playing at a very high level (where there aren't a huge number of players) or chase a particular fish around, you are unlikely to have enough hands on a single player to draw any conclusions from their winrate statistic.

[__Q__]

I completely agree.

[jskills]

Could you explain some of the values in this spreadsheet please? I assume we fill in EV/hr, SD/hr and CI but exactly where would I get these stats from? What is "UB" and "LB"?

EV/hr
SD/hr
CI
UB
LB

Sorry to be dense ...

[OrangeKing]

[ QUOTE ]
Could you explain some of the values in this spreadsheet please? I assume we fill in EV/hr, SD/hr and CI but exactly where would I get these stats from? What is "UB" and "LB"?

EV/hr
SD/hr
CI
UB
LB

Sorry to be dense ...

[/ QUOTE ]

I assumed UB and LB are "upper bound" and "lower bound." In other words, that's the range of your confidence interval. So if you put 98 in as your CI, you could be 98% sure your true win rate fell within the UB and LB (not exactly statistically true, but close enough).