Yeah, as bip! mentions, my Std Dev (according to PokerJournal) is actually slightly less than Dizzy's in terms of bb/hr.

GcluelessStdDevnoobG

Mpethy may be used to stdev in /100

90bb/100 ~= 52bb/hr

If I just did it right, I came back with a standard deviation of $301.60 per hour for my last 1250 hours. That yields a CI of +/- $8.53 centered on my achieved WR of $20.70 for that sample, or 95% confidence of $12.17 to $29.23.

Look right? It basically makes sense to me, as my EV adjusted WR is right near the top of that range.

Where I am currently at, due to lack of game election, I primarily have to play 1/2NL ($300 cap buy in) and regardless of the game I am playing, length or the session or how it went, I input the results.

Since being here this time, I have logged 387.75 hours, maintained an hourly of over $25 and according to Poker Income, my Std Dev is $111.73.

Also, much of this time has been during "non-prime" days/hours due to my schedule. I believe my earn rate would go up but so too would my Std Dev if I were able to put in more hours on the weekends and play more during evenings.

Are you playing 1/3 mostly?

If it is 1/2 - I think that would sound high.

It's all 1/2; I pulled out the 1/3.

I have since gone back and rechecked it, and it seems to be correct.

Don't know if anyone here is a stats genius, but there must be some type of calc to incorporate the knowledge that we are not making above a certain amount (say 15bbs). Seems that this would lower the CI for a lot of people. Anyone know how to some fancy stuff like that?

What if your graph looks like a family of giraffes?

Eatsomemoregruyerelilfatty.gfy





Not quite sure what you're trying to get at here. You can't (statistically) KNOW you're capped at some wr, even though there's always some probability you're below it.

You could do a 95, 96,97, etc ci to see the likelihood of e.g. 15bb being a likely cap.

I suppose there should be a way to calculate a ci given an upper limit. Don't think it would affect the lower end though.

Stfu Zoltan!

My standard deviation over 958 hours of 1/2 is 34.44bb per hour.

No idea how to put that number in context. High? Low?

Sent from my LG-D801 using 2+2 Forums

We can't statistically know we are capped, but we can intellectually know. It would have to effect the lower limit. Say you play an hour of 1/2, and you estimate your SD at $200/hr. You run like god and win $1000 per hour. Well your 99% confidence interval is going to be like $400-1600 per hour. If we say we are realistically capped at $20 per hour, we should be able to see that we might well be a loser in the game. Our lower interval should not be $20, but actually well into the negative. That's my understanding anyway.

No worries, that's mostly why I asked. My range seemed so wide it seemed useless to calculate. Was curious around what amount of hours the range began to tighten to give you a better indication of your win rate.

I use Pokercharts.com to track my sessions, I am using the Win Rate Std Dev they calculate for me. How would I go about calculating it on my own/checking if they calculate it correctly?

Oh look who's a bayesian now!

The thing is, when we use CI, we're using a frequentist worldview... Which means we're naive to all information that might be available. Which means there's no theoretical cap on wr. (absurd, but this is the frequentist approach taught in stats 101.)

Sometimes (not always) CI and credible interval converge (ie when sampling distribution is normally distributed). I sort of doubt that's the case very often though.

Bip!, what's your take on how the typical players results are distributed?

This result is very surprising to me, as I think my biggest leak is over aggression, particularly post flop.

I'm admittedly quite tight PF, but I play big pots with TPTK pretty often, since I play in games where Vs will often call three streets with TPmehK. Often I value-cut myself against their weird 2p hands that they never raise. While I am confident I more than make up for it in value, I would expect this to make for a pretty high standard dev.

Also, bip!'s observation of my playing style might be a bit off, based on the fact that when we played together I had a known maniac on my immediate right, so I adjusted by calling down a lot to keep his range wide and let him do the betting for me while he imagined FE.

So I talked to a couple of other players in my game about their stdev last night. They are both using poker journal and it seems poker journal spits out an errant number in their cases. More work to do - I guess they have to pay $4.99 to export the data and work in excel - TBD if all the poker journal numbers are mis reported low.

. I did not mean nitty in a bad way - more TAG which is top strat for LLSNL. Also, are you using poker journal?

Anyone who uses the app and has an excel sheet with the same records can confirm my suspicion about a miscalculation.

Why don't you post this very question in poker journal thread on 2+2.

Well - my interest was more in helping people ITT calculate their poker results, not really going for improving someone else's commercial software.

/ I might be wrong and the app may be right. Also high potential for garbage in = garbage out, PEBKAC, etc.

App is dead, so you won't have to worry about helping a commercial software.

Just a guess, but I would expect a distribution like the blue one below:



^the mode of that chart would be in losing territory.

However, I have no evidence at all to support that guess.

Also see "paretonormal" below



Some online poker room would know very well how these are distributed.

When the guys busted the ultimate bet cheaters - didn't they post winrate distribution charts that showed the cheaters as outliers? Maybe dig those up.

Another point is that the attrition of losing players (and even break even players) is likely much much higher than that of the top winners.. so even though the population distribution may be X - the sample any given night might have an over-representation of the top players. That could lead to all sorts of distribution possibilities including bimodal where there is a good portion of winners(most likely to be there), and a good portion of slight losers (most likely population).

I think the "smoking gun" they had was a plot of VPIP to winrate against several known winners.

And potripper was crushing playing 35% of hands in full ring or something like that.

Not insulted, just surprised. I def feel like I play a pretty textbook TAG style most of the time. It's just that when I overdo something, it's usually the "AG."

I use a freeware excel spreadsheet called "Session Log." No app, as I'm too lazy to re-enter all of my results.