Help me explain this Re: online winners/losers... (long)

UMTerp · #1 11-17-2005, 04:32 PM

The guy I'm having a conversation with on another board is a bright guy, but for some reason, he's not grasping this. How can I explain it more clearly? I'm trying to tell him that the overall percentage of online players that are longterm winners is in the neighborhood of 10%. Whether you agree with that statement or not, assume it's true. Can someone explain this more clearly than me (I'm in red, he's in blue). And FWIW, he is a profitable player:

umterp knows i think it's much higher than 10% or whatever. i've had two poker tracker databases in the past, and both show 40% winners 60% losers. that makes sense to me with the 10% difference (off of 50%) being people that would be net profitable but for the rake. yes it's not a complete sample size, but it's better tha random guesses. if someone has a theory as to why that 40% is so much higher than 10%, i'm open.

the other reason i think it's higher than 10% is simply by observation - i don't think i'm among the top 10% of people i play with and am a profitable player.

It's because you have such a small sample size (relatively speaking) of your opponents. If you have ten losing players sitting together at a table (which is probably typical at some tables, though not necessarily yours), somebody's going to win money in that particular session. I'm guessing the vast majority (98%+) of the opponents in your database had less than 5K or so hands played against you. Over a few hundred hands or whatever, sure, many losers will show a profit.

that's actually another way to think about it, so you think only 1 person at a 10 person table is ahead? i would say 3-4 is more accurate. we can go down the list of tables right now and figure out how many are up. it's definitely more than 1 at each table.

since i play shorthanded NL, there are always 2-3 people that are ahead and 3-4 that are losing. almost never is there one person beating everyone.

and sample size or not, that's two different databases as well as what other people's extensive databases on two plus two look like, all showing the exact same figure - 40%. put them all together and that's a large enough sample size. just because we don't have every hand played on line doesn't mean it's not representative once you get to a certain level of hands. and it certainly wouldn't show 40% over that many hands if it were truly 10%. that's too big a difference.

I'd agree with you regarding the bolded sentence, BUT...

...winners are more likely to play more (and keep playing) though - the losers keep revolving. I bet a lot of the "winning" players with a small number of hands in your pokertracker have gone bust, which is why you show them as winners and don't get any more data on them, don't you agree?

the point is every single person's database all shows the same thing - 40%. wouldn't some show 1% or even 10% if the latter number is the correct figure? combine all of our databases if you want to think of it that way, and that's a huge sample size.

you really think you're in the top 10% of players. i don't think i am and i could prolly pull 6 figures a year if i put in the time.

and that's not short term, that's consistently, every single day, it never changes always right around 40%. and i'm not talking about 5k hands. i played a lot over the last 3 years.

i'm not surprised by the results of this poll at all. i bet some of these are net losers but it doesn't seem like it cuz they aren't losing much, and you'd have a number pretty close to 40% if they really got their accurate numbers.

He makes his living with stocks. so I tried to go this route:

No, not at all. I don't understand why you're not grasping this concept... argh - I know you have a good analytical mind too, I'm just trying to think how to explain it to you... ummm...

I'm totally unfamiliar with the relative success of stocks, but I'll try to equate it to that... Question - is there some sample section that has like a 10% success (or failure) rate or something?

If there is, look at that sample group and see how they perform in any one given week? 40%ish? Look at any other given week - probably right around 40%. Those are like individual player's pokertracker databases. Every week, 40% will be up/down, even though longterm only 10% will be. Same principle.

yes, i even took several semesters of statistics. they don't tell you in statistics that you need every hand to come up with a good guess. and if the number were truly 10%, then my databases would've showed a gradual drop to the true number. so if you saw a database with 10 million hands, which we could prolly get by combining everyone's databases, it would show 40%. there is no sample size big enough for you. you don't need every hand ever played to get statistically valid result. i have yet to see a database that is anywhere close to the number you are thinking and they all include the rake.

everytime i sit at a table at NL1000, i know for a fact that at least 2 are profitable players, at a least a third of the table, and usually half. i know from playing them. 

HELP!!

SunOfBeach · #2 11-17-2005, 06:28 PM

if he has a database full of guys who play 1000NL, im not surprised that more than 10% are profitable. he should try datamining the 1/2 limit game for awhile, then see what he comes up with.

anyhow, the mistake hes making is related to a concept in statistical analysis referred to as 'survivorship bias'. google it and give him a lesson, but the cliffs notes version is that he keeps looking only at those players who are still around, and not at all of those who have already gone broke and quit - particularly if hes playing NL ring games with 5-10 blinds...

SumZero · #3 11-17-2005, 06:44 PM

Simulation.

The best way to show him you are right is to simulate it. Create a population of 1,000,000 players with various bankrolls. Make only 10% of them winning players, but have the typical random outcomes when people play. Give people bankrolls that once they lose they stop. Simulate the population for a while. Subsample the population of players (looking only at a small % of the hands played) to simulate the types of looks your tracking player gets. Show that if you only look at the subsamples you think a lot of players are profitable that if you look at the full data and/or the actual a priori settings you realize aren't.

That may be more work than you are willing to put into it.

Allinlife · #4 11-17-2005, 06:55 PM

calculate -ev of rake, proving that this isn't a zero sum game.

prove to him only few have winrates that surpass the -ev generated by rake.