EV vs Players

[Schneids]

Progressing up through the limits and into games where I often play vs a small core group of players, one of the ways to table select is to evaluate whether I am +EV or -EV vs each individual player, and then consider as a whole and based on my seat position whether I am +EV or -EV in the game. However, I have no basis other than PT misc. tab and my own estimations whether I am in fact +EV or -EV vs each person.

So, saying that the PT misc tab will be my main indicator, how long do we expect it to take to be an accurate estimation if I'm winning or losing vs each individual player. And, in what ways should I be looking at this tab to be able to feel confident about my conclusions...?



For instance, player 1... Can I conclude I'm likely +EV vs him because of such a decisive difference (~1.75:1) in amt won vs amt lost, even though I've only had 289 times vs him. Or player 2, is the fact I have 1511 times vs him enough to conclude that the fact I'm up against him makes me +EV against him? Or should I be concerned that these results so far mean squat and I need a substantially larger sample size, as evidenced by the two #3's -- both 2p2ers who are well respected around here and people who I likely have no edge vs...or a small one at best.


Now, lets reverse the spectrum.


Using my DB of people I'm losing the most to, I see Boutros at the top. I do not find that surprising, and begins to lend itself some credibility until we go one further down and see Sansone #2, which then tells me this entire thing is worthless without huge sample size since I really doubt I'm -EV vs him.

Moreover, going to resder1 provides another interesting question: will the convergence speed depend on whether playing in games vs these people 6 handed vs 4 handed vs HU? Most of my playing vs resder1 has been HU, so, should I expect it to take longer to converge if we played eachother for a few thousand hands HU in an aggressive style?

And, not so shockingly, there are names of a few people on the list who I'm losing to but am almost positively not -EV against. When can I know for sure though?


And a final question: Lets say I can sit in a 6 handed table with one huge donater, but four people who I'm pretty sure I'm -EV vs. How big of a +EV do I have to be vs this donater to make up for me being -EV vs the other four (we'll say ranging from most of them being slightly -EV to maybe one player being largely -EV vs). Is it insurmountable? If not, how much do I need to factor in my seat in relation to the donater. Does the equation change substantially if I make it 3 I know I'm -EV vs, and one small +EV and one large +EV?


I know my questions are pretty vague but if anyone has a greater understanding of this topic or some educated guesses I'd appreciate some further insight. Also if anyone thinks it's in poor taste for me to not hide the names of the players I'm losing vs say something and I can go back and edit the picture.

[partygirluk]

I wouldn't use this to gauge your EV in a game, as it is near meaningless. An experienced player should be able to estimate whether players are +EV v. him or not MUCH MUCH better than by simply relying on results. Furthermore, you can never say "My EV v. player x is 1.7bb/100" even if you have a billion hands with him, as your EV will depend upon table position, who the other players are etc.

[DcifrThs]

schneids,

think of all the things that go into a calculation of expected value. if you were to model it, the error term would be huge, easily capturing more variance in the model than the parameters themselves. in english, this means that overall if you did tons of these models, the intangibles would win out in terms of explicative power.

basically, i dont think observed EV tells you a whole lot...as a good player you can see for yourself mistakes people are making att he time they are making them. if these mistakes are big enough that you feel you have an edge. have a seat.

when there are 3 good players in a game w/ two fish, id really only consider sitting if that one fish is just spewing and the other one is bad. you're paying basically a fee to the good players to sit and play against the fish (if you are -ev vs. the good players). if you're break even vs. them, you should sit to get the fishes money. either way. its a gamble.

Barron

[Schneids]

Honestly, you're sitting at a table with me, or BK, or James, or solid player XYZ. Can you really tell if you're winning or losing vs them just purely off observation? I know we can say "well I'm approximately breakeven" but the thing is we aren't, we're either winning or losing. I want to figure out which, since if I think I'm breaking even vs someone and in fact really I'm losing a small amount, then you multiply that times 3 or 4 others at the table, then we've suddenly got a problem unless there's some bigtime donaters.

[ QUOTE ]
when there are 3 good players in a game w/ two fish, id really only consider sitting if that one fish is just spewing and the other one is bad. you're paying basically a fee to the good players to sit and play against the fish (if you are -ev vs. the good players). if you're break even vs. them, you should sit to get the fishes money. either way. its a gamble.

[/ QUOTE ]

6 handed game with 3 good players and two fish sounds like a great game to me, I dunno what your game selection requirements are... I don't understand but it sounds like you are saying that basically the only playable games online are about $20/40 or smaller.

[ISF]

You could probably model this, but because of sample size issues you couldnt use standard statistical methods. If you took a bayesian approach and assigned everyone a prior ev against you either based on a guess from their play, or via their pokertracker stats and a prior probability. Then adjusted this using standard stats methods based on your ev against them in actual play you could probably get a resonable estimate of your actual ev vs them.
You could then pretty quickly figure out wether you were +EV vs any table on average. To take position into account would be much more difficult and probably not worth the effort.

[disjunction]

Isn't it a big deal whether you've been sitting to a player's left or his right? I would think that not only would you need to consider number of hands played, but you'd also need to consider number of sessions played before this number starts to become meaningful.

[SmileyEH]

A player sitting to your right will automatically be +EV against you unless he's beyond terrible...I guess if you averaged it out than it wouldn't matter, but that would take tons and tons of hands. Your idea is interesting but it seems really impractical to implement.

-SmileyEH

[bicyclekick]

A couple hundred hands or even a couple thousand hands vs one player is pretty meaningless. I mean, cmon...it's all talked about how little 30k hands means...or even 100k hands...and this isn't any different.

The more I play, the better of a guess I have of who I have an edge on and who I don't. I think with many top players it doesn't really matter much. Whether you or I have an edge on the other guy is so close it's not even funny. I don't think it's all that consistant either. I think some days I have an edge and some days you do. It has a lot to do with how we're running and some very recent hands.

Just worry about finding games with players you know are making decent sized mistakes.

[Nietzsche]

Just a comment about your result vs. the sansone guy. He won 187 while you only won 85. Obviously you are running bad vs. him but the amount he is ahead is remarkably small. He is not extracting as much from you as you are from him when winning hands. So it would be relatively safe to conclude that it is +EV for you to play this guy.

[elindauer]

Stick to finding particular weaknesses in your opponents games by reviewing hands. This kind of analysis is the only one that converges fast enough to give you any real confidence.

As far as your winrate vs other players, here's an idea: if you look but cannot identify a specific reason why you have an edge, it is best to assume you are -EV against them, and proceed accordingly.

my 2 cents.
eric