% Winners and Losers

Mr. Flibble · 19-01-2008 4:29pm #1

Where would you prefer to play? At a site with 80% losers and 20% winners or a site with a 50/50 distribution?

Would you think there are more likely worse players at a 80/20 site?
I know that this data is in theory independent of how good the players are but if all the 2+2 MSNLers (ie. reasonable players) played on one site I would imagine that there would be a nearer 50/50 distribution of winners and losers. Perhaps this is wrong though.

iPoker is about 37%/62% winners/losers and seems to have the same distribution at all levels, 20NL to 1000NL.
I don't have data on other sites, are they about the same does anyone know?

Mellor · 19-01-2008 4:44pm

My mine hands on ipoker are at 58/42 or there about

digiman · 19-01-2008 5:29pm

On ipoker at 50NL I have 12.5k players and a total of 1687k hands, the dist is 38.01% winners to 61.99% losers

spectre · 19-01-2008 6:14pm

The percentage of winners and losers isn't really that relevant. The distribution of winnings is a more accurate picture of how a poker ecosystem behaves. If someone could post a graph of this distribution, it would be interesting.

Mr. Flibble · 19-01-2008 6:24pm

You mean each player on the x axis and how much they won on the y axis? A bit like this showing 6 players where each column is a different player showing how much they won/lost?

w
ww
ww
www
www
www
___LLL
___LLL
___LL
___LL
___L

jayhawk · 19-01-2008 7:23pm

Although the citation is too ancient for me to resurrect, I’ve accepted that the conventional, historical wisdom was that 92% of online poker players were long-term losers and that the bottom 10% were problem gamblers. Of the 8% who were reported to be lifetime winners, a quarter of that group were said only to be casual players. (I would guess these assumptions are no longer fully valid on sites that continue to allow Americans access -- narrower shark to fish ratio there.)

There is a vast disconnect between this assertion and a 65/35 Poker Tracker distribution.

It would seem to be the most logical to play at a table that seated the greatest number of players inferior to you, since exploitive play is fundamentally based on avoiding the strong and feeding on the weak. (The advantage of Poker Tracker and HUD data in cash games is that one can identify these opponents more quickly.)

My personal view is that any distribution of opponents’ relative skills is irrelevant after a history of play has been established. I’m going to play the site, and the game, and the level where I’ve been the most profitable, regardless of how the other lads appear to be doing. And if I’m not, I’m going elsewhere. It hardly matters if my table is full of fish, if I’m a fish, too.

Zab · 19-01-2008 7:25pm

I'm not sure that a typical PT database is going to give a good indication of this. I'd say a typical non-maniacal losing player might book several small wins and then have a blowup where they lose everything (they may even move up first). If you miss the blowup then he appears to be a winner.

spectre · 19-01-2008 7:37pm

Mr. Flibble wrote: »

You mean each player on the x axis and how much they won on the y axis? A bit like this showing 6 players where each column is a different player showing how much they won/lost?

w
ww
ww
www
www
www
___LLL
___LLL
___LL
___LL
___L

You would need to do a level by level analysis and you would need a fairly huge mined dataset (If all hands include the same player [i.e. you] the results will be biased)

Only include players with a certain amount of hands played, say 400

x-axis: winrates ranges (small increments, say 0.1BB/100 hands)
y-axis: number of players within those ranges

It would be in interesting graph I'm sure

hotspur · 19-01-2008 7:55pm

You might want % of players on the X axis and bb/100 won/lost on the y axis.

It is not easy to get especially meaningful information out of the data. You obviously cannot use $ won/lost across the limits as losing $5k at $25/50 means nothing compared to losing it at .50/$1. And using it per limit has the problem of risking characterising players as losing players but they may only be losing players a specific limit.

So I guess you have to use bb/100 won/lost across all limits per player. It should also be filtered for a certain minimum number of hands for a player to qualify for inclusion as otherwise it will skew the results.

Your results are interesting, you know I really don't think there is much public information of this type at all out there as mass access to data mining is the only way for it to emerge in the absence of poker sites revealing accurate stats on it (which they won't). And how many others have your mass access to data mining? Though somehow I have come across similar figures before.

Looking at it on a micro single 10 person table level, I know from experiments done that a poor player at a table with 2 expert players has a better chance of making money by playing the casino games than by playing in that poker game. If the winning/losing % were really a reflection of expertise then that would mean it is a really really bad idea for poor players to ever play poker as they would be sitting with 3.7 experts on average- they should be playing roulette! But it isn't really an accurate reflection of course.

The question of whether all players of the same starting skill level would produce a 50/50 split after a period of time is actually a hell of an interesting question. Poker is dynamic and adaptationist. If all the players were solid experienced players to begin with then the % split would be determined by the distribution of the ability to adapt.

Firstly let's assume that the 50% of losers are not going to become winners by adapting, so we are still going to have a minimum of 50% losers. Of the winning 50% - 50% will be good at adapting and 50% not. We now have a 25% of winning and good adapting players, a 25% of previously winning and not good at adapting players, and our 50% of previously losing players.

The 25% of winning and good adapting player continue to win and the 50% of losers continue to lose. So now we have to split up the 25% of previously winning but poor adapting players into those who continue to win from the losers and those who now lose to the new supergood players. Assume that they split off 50-50.

That leaves us with an end result of:
25% of winning and good adapting players, 12.5% of winning but poor adapting players, 12.5% of previous winning poor adapting player who now lose, and the 50% of players who always lost. This leave 37.5% winning players and 62.5% losing players. And this is exactly what the real figures show.

So you see the figures are precisely in line with what one could predict from an evolutionary theory of what should happen.

spectre · 19-01-2008 8:23pm

Interesting post hotspur. It's true, you really need to have access to a poker site's database to have all the information.

However, it's still possible to extract somewhat meaningful data. For instance, take an unbiased sample of all players at 200NL who have played >= 6000 hands in the last 3 months. This gives a pretty good idea of the distribution of win rates of regular players. It would also be interesting to compare this distribution to a 3 month interval 18 months ago. It should give a good idea how the games have changed since then.

EDIT: Can't emphasise enough that you need a truly massive sample of data to render useful results

Macspower · 19-01-2008 8:36pm

I would have thought that most sites we 80% loosers and 20% winners? maybe not on the ones with USA players now thought... thy could be closer to 50/50 i guess but i dont play there

spectre · 19-01-2008 8:41pm

I think it's funny how people assume that US players are more likely to be sh!t than their Irish counterparts

Mr. Flibble · 20-01-2008 5:52pm

hotspur wrote: »

Firstly let's assume that the 50% of losers are not going to become winners by adapting, so we are still going to have a minimum of 50% losers. Of the winning 50% - 50% will be good at adapting and 50% not. We now have a 25% of winning and good adapting players, a 25% of previously winning and not good at adapting players, and our 50% of previously losing players.

The 25% of winning and good adapting player continue to win and the 50% of losers continue to lose. So now we have to split up the 25% of previously winning but poor adapting players into those who continue to win from the losers and those who now lose to the new supergood players. Assume that they split off 50-50.

That leaves us with an end result of:
25% of winning and good adapting players, 12.5% of winning but poor adapting players, 12.5% of previous winning poor adapting player who now lose, and the 50% of players who always lost. This leave 37.5% winning players and 62.5% losing players. And this is exactly what the real figures show.

So you see the figures are precisely in line with what one could predict from an evolutionary theory of what should happen.

Is there an evolutionary theory saying there are 3 50/50 splits like that? Does it take rake into account...

I posted in the postgresql poker tracker forum to see if anyone has any ideas how to go about doing this. There are no postgres gurus here are there?

jayhawk, if you had no other information about two sites other than one has a 50/50 winner/loser split and the other a 80/20 split which would you play at? Or would not affect your decision?

Macspower · 20-01-2008 6:19pm

spectre wrote: »

I think it's funny how people assume that US players are more likely to be sh!t than their Irish counterparts

quite the opposite in fact if you read it again... the US player like the irish players have 20% good player and 80% fish (at a guess) the 20% good players seem to have made their way onto the 2 sites that still allow US players but they have to do a bit of hoop jumping to get there I think.... the 80% bad players are just not playing anymore.... the US players that seem to be online now ..to me.. seem to be the cream of the crop...

maybe I have it wrong but my understanding of it

TommyGunne · 20-01-2008 7:29pm

jayhawk wrote: »

Although the citation is too ancient for me to resurrect, I’ve accepted that the conventional, historical wisdom was that 92% of online poker players were long-term losers and that the bottom 10% were problem gamblers. Of the 8% who were reported to be lifetime winners, a quarter of that group were said only to be casual players. (I would guess these assumptions are no longer fully valid on sites that continue to allow Americans access -- narrower shark to fish ratio there.)

There is a vast disconnect between this assertion and a 65/35 Poker Tracker distribution.

I don't think the PT stats include rake.... this would reallign the stats along with the old assertation. I am willing to be corrected on this though.

NickyOD · 20-01-2008 10:21pm

TommyGunne wrote: »

I don't think the PT stats include rake.... this would reallign the stats along with the old assertation. I am willing to be corrected on this though.

PT includes rake

Imposter · 21-01-2008 7:42am

You can't really ignore the players who play only a few hands either though. Chances are they are the fish that sit with 1 buy-in and proceed to lose it. Ignoring them would skew the results quite substantially I would think.

Mr. Flibble · 21-01-2008 9:41am

Imposter wrote: »

You can't really ignore the players who play only a few hands either though. Chances are they are the fish that sit with 1 buy-in and proceed to lose it. Ignoring them would skew the results quite substantially I would think.

Given the PT stats I suppose its actually quite possible that at any given time the majority of players playing at a site are winning players if there are a lot of losers who only play for a short period then leave.

% Winners and Losers

Comments