I'm interested in the 3bet stats, as mine are similar until you get to the CO and BTN where they are like 6% lower. I'm wondering how a 20% btn 3 bet range varies with the position of the initial preflop raiser. I can see how you could 3 bet a CO raise of say A8, A6s, but not an UTG raise. So in order to get up to this sort of 3 bet percentage on the button, you must be reasonably tight against an UTG raise then have a very wide range indeed versus a CO raise?

t-test which can be done with any standard stat software. For example say your results in a given spot are +1, -3, +6, then the following R code will work:

t.test(c(1,-3,6), alternative="less")

which results in:

One Sample t-test

data: c(1, -3, 6)
t = 0.5121, df = 2, p-value = 0.6703
alternative hypothesis: true mean is less than 0
95 percent confidence interval:
-Inf 8.935272
sample estimates:
mean of x
1.333333

The bolded value tells you that there is a 67% chance you would get data this favorable or more favorable to the hypothesis that the mean win is 0 or worse.

If you're filtering for hands that you see the flop with, the SD is going to be much higher because you just filtered out a bunch of hands that all have the same value. i.e. data of -3, 4, 5, 0, 0, 0 has a lower SD then -3, 4, 5. Also the better the hand the higher the SD because you will be putting in more bets on average and thus you will either have a large positive result or a large negative result.

Apologies in advance if my question is inane but just how accurate would a winrate be over 55k hands?

I realize that this is perhaps a question that can be explained through statistics, but i am a layperson when it comes to stats and probably wouldn't understand it unless it was overly simplified.

There is a good reason i ask. Over the last 55k hands i've played, my BB/100 is stupidly high. I realise poker has a serious amount of variance, but just how much? I'm starting to get high off my winrate, feeling that it is sustainable etc.

I guess i want somebody to say either-

1. Boyo, your winrate is sustainable, you can beat the pokers if you keep doing what you're doing.

2. Wow this is variance, serious variance, enjoy it but ignore it, it's an illusion.

Perhaps there is a #3 i'm missing? Perhaps there are people with huge sample sizes or something that can tell this young upstart about the variance in the long run.

I think where some people get confused here is that they think about hands that have close to 50% equity long term, which would have higher standard deviation in win%, but not necessarily in bets (as it's actually measured) as you pointed out.

We can say with 95% confidence that your winrate is between your winrate+1.7 and your winrate-1.7.

ty leader

Leader, if you don't mind, would you be able to calculate for 150k hands? I'm having some similar feelings as devildances.

In In general, a 95% confidence interval will be:

(winrate-1.96*sqrt(hands/100)*20/(hands/100) , winrate+1.96*sqrt(hands/100)*20/(hands/100))

assuming a standard deviation of ~20BB/100. So for 150K

(winrate-1.012140,winrate+1.012140)

Thanks Leader!

I just got Pokertacker but I can't find standard deviation anywhere.

Sessions Tab
Display : Sessions -Details (upper panel dropdown)
Display : Details- Details (Lower panel dropdown)

Thanks!

This is for 2/4 filtered 5-6 handed. I know this is a small sample size, but I just want to know if I am on the right path. Can anybody tell me if there are any big noticeable leaks?

Thanks!

Maybe a bit passive postflop, but not too bad. Other than that your stats seem fine to me. If you are looking to improve it may be something beyond the stats.

been working on my game and trying to open more on btn and co.
doing pretty bad...
have a general question - figured this thread will do - i table select very well, finding the fish to my right. do u ever consider not sitting in a game when a good tough player sit to your left? i have been loosing a lot in spots when i get 3! lightly and i either fold of call my A high too often... maybe just running bad but still started to think i need to table select my left players as well?

Maybe I will get bashed for this, but I would rather have them directly on my left rather than two or three to me left because of stealing. BvB doesn't come into play at the level I am playing, but Button vs Blind and CO vs Blind comes up a lot.

well this is the problem i have been dealing with lately - a strong player to my immediate left 3 betting my btn open from the sb, 3betting my CO open, or 3betting my iso raise when the fish limps. while trying to open my game i just feel that i need to tighten up in these circumstances...

sorry for the delayed response.
in short, apparently, i'm a wizard postflop (not my words, see earlier posts).
at length, yes but it depends on who is UTG. I'm not 3-betting A6s OTB or even in the CO vs UTG unless UTG sucks ass at poker.

having tighties in the blinds helps. they're going to fold if you raise anyway. there are times when I muck certain hands OTB vs a certain CO open that I would otherwise 3-bet or cap vs other COs.

not sure if that helps but...yeah, you pretty much answered your own question.

I've made a few adjustments and am concerned about my big blind stats (mostly 5-6 handed lhe). Over my last 2200 (big blind) hands I've noticed I'm losing .259bb/100!! Obviously, this is a small sample size, but I'm still a little concerned. I think it should be closer to .21/100 or lower.

Filtering for big blind hands only, my standard deviation is 18bb/100. I calculated that 1 standard deviation to be 83bb or .259bb/100. Is this correct? If so, then I definitely can't worry about it. Still, my bb/100 from the bb doesn't seem like it should take this long to converge. Normally, it seems I'm almost always within .04bb/100 in a couple thousand hands, which tells me that I'm miscalculating something.

FWIW- vpip (from bb) - 43.1 which I know is high. I'd appreciate any help. Thanks.

(winrate-1.96*sqrt(Hands/100)*20/(Hands/100) , winrate+1.96*sqrt(Hands/100)*20/(Hands/100))

winrate=-25.9
Hands=2200
SD=18

(-34.25747,-17.54253)

Without looking I'm pretty suspicious of that SD number for this situation. Seems low.

Excuse my ineptness at math. I follow what you wrote, but am not sure what it means. What I did was take 2200/100. Then took the sqr of that and multiplied it by 18 to get 1 standard deviation. Then divided by 100 to get bb/100 std Dev. I'm not sure what the winrate = 25.9 or the (-34.25747,-17.54253) means in your example above.

I was too. But I do have a lower standard deviation than most. My overall std Dev is only 15.5bb/100. I think most tags are closer to 17. I filtered for only big blind hands and HEM gives me 18.08 under the std Dev BB column. I assume it's right? I appreciate your help.

Well you're losing at a rate of .259/hand I would assume. So SD is measured in BB/100 and winrate here is measured per hand so we need a consistent scale. I chose the BB/100 scale because of theoretical concerns.

That's the general way a confidence interval is expressed. So in words, we are 95% confident that the true value of your winrate in the BB falls between -34.25747 BB/100 and -17.54253 BB/100.

Yeah that is all very odd imo. I'd be concerned if my SD was 15.5BB/100, but idk it could work.

Jesus, what a careless and stupid mistake! That's why I suck at math.

So it appears very likely (too likely), that I'm closer to -.25/100 than the -.20/100 that I feel it should be. So I take it that it's safe to assume there IS a problem.

I like to think it's because I read hands pretty well, but that's wishful thinking. I'm probably too foldy in some spots and pass up some of the thinnest edges that others go for. My overall vpip is probably lower than most at just slightly higher than 28%. But I still have a pretty respectable win rate. It's not world class or anything, but from what I see it's definitely above average. Thanks again!

Just a matter of practice. I make mistakes all the time too



This is not the conclusion I would draw from the CI. It is true, but what you're saying can be deduced from the mean. You're more likely to be at .25 then .20 because your mean value is .259*. What I would say is that the evidence is inconclusive. You're true value may be .20; it may be .25.

Yeah it's just sort of strange. I certainly wouldn't recommend any radical changes. You might just want to look for spots in terms of looseness and aggression that you might be missing. It's really about improving in specific spots though.

*and the distribution of the mean is symmetric.

Right, and I'm quite sure my big blind defense play can improve. From what I've seen ITT and on PTR win rates of -.19 to -.21 seem to be the norm for good winning players. I have never managed to lose less than -.24 from the bb. Shaving even .03 off that is huge!

I wish I new math well enough to play around with stove. I've never been comfortable with how to proceed after defending against a button steal with a hand like A4 off or dry kings on a QJ7 two tone flop. I know A4 has around 1/3 equity against a typical button range , but it's terribly reversed implied too and so much depends on how cbet frequency affects range, etc.

Speaking just for myself, I think the most room for improvement exists in play from the bb. I'm comfortable in most other spots, but I find myself absolutely lost and not knowing how to proceed from the bb.