Stats question

[btspider]

i had a healthy 6.1 BB/100 winrate, but i doubt either of us have enough hands to conclude anything. aggression is great, but at .50/1.00 you are drawing to two pair, gutshots, flushes and OESD's on the turn often enough that you do call a fair amount.

[Wahoo91]

bt-
Very true on the sample sizes...

I find that I usually raise my strong draws on the flop, and then it gets checked to me on all other streets, keeping my agression very high.

Maybe I should flat call those flops to get the overcalls your win rate may suggest I am missing.

[btspider]

my winrate is pretty meaningless. i pump strong draws as much as anyone. i just mean that at .50/1.00 you have weak draws that are worth playing alot.

[Blarg]

[ QUOTE ]
2) Why do folks say that 24k is not a good sample size. I am a professional research scientist, and to me 24,000 data points seems like an immense sample - how big could your standard deviation be on a sample that large?


[/ QUOTE ]

Two answers.

As Amirpc notes, some people go through 1,000 hands a day or more, so 24,000 can be well under a month's time. Put in that perspective, 24,000 turns into an iffy number regarding your long-term win rate. It's still good for many other things. Some things, like your PFR and VPIP, can come up in the stats fairly quickly and reliably. Win rate cannot.

The other reason is that disproportion in the distribution of hands needs a long time to iron out, as do unusual results when you do get certain hands. 24k in hands is functionally smaller than it seems.

Let me illustrate with some concrete examples from one of my own databases.

A great deal of your money in poker is made just off of AA, KK, QQ, and JJ. Yet even in tens of thousands of hands, you will not wind up getting, say, AA all that many times.

In 52,870 hands, I've gotten AA only 248 times, 245 KK, 212 QQ, and 239 JJ.

Look how few times I got QQ compared to AA, KK, and JJ. That disproportion affects my profitability and stats quite a bit at 53k, and would affect them even more so at 24. These are the money hands; having too many for your sample size or too few really changes up your numbers.

I had one 10k sample where KK was a loser. I've had stretches not quite as long where AA was a loser. Ouch!

And then, there's how you do on those hands. Getting outdrawn on the river happens to all of us, and happens plenty. But sizing my 248 times getting AA down to the level of 24k hands, let's say I could expect a generous 120 times getting them at 24k. If I averaged out to $10 profit every time I got AA, I would have made 120 x 10 = 1,200 off AA.

But what if I had even a very tiny swing, losing just 10% more of those 120 AA's? Say, instead of winning $10 each time, I lost $5 just 10% of the time? Now I've got 108 AA x $10 = 1080, along with (12 x -5) = -60, for a total of 1020 won. My average win has now plummeted 15% to $8.5 and I'm down $180 in cash from where I would have been. That's quite a drop! What if my losses per hand weren't so trivial on those hands? Or if I had even more losing hands than a tiny 10% negative swing would represent?

24,000 junk hands don't matter too much, but what happens with 120 AA matters a lot. That's the sample size you care about. You may find a lot of people with databases whose entire profit can come down to an amount of money equal to what they've won on their aces alone, or darn close to it.

[Blarg]

Your aggression number may be the highest I've ever seen anyone post here, if not THE highest. I doubt many players, winning or losing, come anywhere near your figures.

[Emperor]

Disclaimer I am no expert. I have an excel spreadsheet that crunches my numbers for me.

What I can tell you:

24K hands is a small sample considering... Your short term variance is huge, add multitabling and it is Huge^4th

Your SD is usually around 9 to 12 BB per hour at one live table.. short term that goes up with multitabling but QUICKLY comes down as you reach "the long term" a lot faster.

So even if you are making 4BB/100 hands then you can still break a 300BB bankroll easily multitabling. I've done it.

whats all this mean? Variance is INSANE if your SD is 9-12 BB/hour.

So long term stats are necessary to have any confidence of any particular stat (TG Excel can calculate confidence levels)

Do a search or read up on confidence levels and you will see you need about 40K hands to start bringing those confidence levels to a point that they aren't outside of your SD..

It like the election polls... They say they are accurate within 3% or 4%, but statistically they are only accurate to within 33% at best... :P

[detruncate]

My winrate varies a hell of a lot. I have 10,000 hand samples where it's high, and 10,000 hand samples where it's negative. Not long ago, I had 2 100 BB drops within 10,000 hands. Does this mean I suck? Could be. Did I rule when I was putting up better numbers? I doubt it.

The upshot is that I no longer pay much attention to my winrate. It's finally sunk in during the past few weeks that the whole thing is virtually meaningless. I'm constantly trying to improve. I am where I am, bankroll wise. There's nothing more to it.

[xxSHARKYxx]

Strong numbers IMO. VPIP (Vol. Put Dough In Pot) i'm 21.8%.
Some guys like 19-20 but no biggie.

Good aggression and excellent pre flop raisin! (att. to steal you 29 me 18) like your number more than mine.

Possible Leaks - W$WSF you 27% (won $ when see flop) really low, i'm 36% and I think this is pretty good. May be giving up on some winners that you can take at showdown w/ace high or scare lags off by bettin turn/riv aggressively. I remember the no foldem holdem game you are playin so be careful.

Also think you may play your big blind to much (to a steal)
I think most of the pros on this site are around 50-58%.

I think you are in the wrong game. Move up to at least 2/4. You will profit at 2/4 or 3/6 with these numbers IMO. (proper BR, 250 BB)

I play 2/4 to 5/10 mostly 6 max

onlineshark

[Wahoo91]

I could be overagressive, I demand to be leading each hand I participate in. Very rarely do I ever call a bet. I either bet, raise, reraise, or fold post-flop (except on the river for very large pots like Ed talks about I will call to go to SD even if I *think* I am beat). This seems to be working for me, but this is a smaller sample size (only a couple thousand hands since SSH where I have been focusing on that part of my game).

My combined aggression is significantly lower (pre- and post- SSH).

I probably should tone it down a bit to become a bit more "normal". That level of aggression is actaully a bit of an experiment for me, and it has been working out very well so far.

BTW, my VPIP is about 18 and PFR is about 7. But once I like a hand post flop I usually "burn rubber" or get out.