SD and variance

[brian0729]

I have questions concerning standard deviation and variance. I know what my SD is (about 12BB per hour), what I would like to know is what does this say about my game if anything. Should I try hard to lower this number? Is this number ok where it is or should it go down or does it vary from style to style? Is SD and variance the same thing?

[pudley4]

We also need to know what your win rate is, and approx how many hands per hour you play.

Example:

Win rate = 3 BB/hr, played 60 hands/hr. SD of 12 is very good
Win rate = .1 BB/hr, played 60 hands/hr. SD of 12 is very high.

[brian0729]

Win rate is about 1.4 BB's per hour right now and am playing one online table at a time so about 60 hands per hour. From your example I would think this is a touch high but not horrible??

[pudley4]

No, I think your results are good.

In a B&M, a SD of around 10-12 with a win rate of 1 BB/hr is pretty standard. You're winning at a higher rate, so your SD is lower in relation to your win rate. Also, playing online gets you more hands per hour, so you'd expect your SD to be even higher online.

[Homer]

Your numbers seem about right for a skilled player. Good numbers for live play are an EV of 1 BB/hr and an SD of 9-10 BB/hr. Playing online you get about 1.5 times as many hands in per hour (60 versus 40), so your numbers are equivalent to an EV of .93 BB/hr (1.4/1.5) and an SD of 9.8 BB/hr (12/sqrt(1.5)) in a live game.

-- Homer

[daryn]

now that we're talking about this kind of stuff, and you numbers guys are commenting, how about these stats?

SD/hr : 17.19 BB

WR/hr : 3.02 BB

this is over 111.25 hours.

anyone want to make something of this data? i need help..

[Glenn]

the only thing you can get from that small a sample is that you have, in fact, played poker

[Nottom]

You are 90% likely to have a winrate over .5BB/hour?

[daryn]

ok so how long does a sample have to be to extract useful info?

[Glenn]

That is not necessarily true. Since he won 3BB/hr, we already almost definitely know that he was running good (since almost no one, if anyone, can sustain that win rate, except 1-1--he will likely regress to the mean). Therefore, his win rate will not be normally distributed around the sample averge. The distribution is skewed to the left. When people post their win rates, they are always running very well...so this concept is very important. I'm sure Daryn is a decent poker player and wins in the long run, but those stats don't mean much.

[Glenn]

Search the archives and find Bruce Z's post explaining SD and such. It is very good, and it won't take you long to understand.

Bascially, if you have a sample average and a sample SD, you can calculate a range of possible win rates with a certain confidence. For an SD of 17BB/hr, it takes a while to get any confidence in your result. If you want 95% confidence within 1BB/hr (which still isn't great since your win rate will probably be something like 1BB/hr and there is a big difference between 0bb/hr and 2BB/hr) it will take:

drumroll....

1156 hours...

and to get 95% within .5 BB/hr it will take...

4624 hours!

pretty sad huh? this is why some more experienced players are constantly annoyed by people who come on the scene bragging about how much they've won (not saying you have done this, you just asked a reasonable question...but people do all of the time).

The best thing to do...and I always tell people this, but no one listens...is to not assume anything about your win rate and just play as best you can. If you win money, be happy, and if not, get a new hobby. That's all that matters if you're not a pro.

[Nottom]

I'm not so sure that it matters if he was running good or not. BruceZ has mentioned a few times that while playing IRC poker he had acheived a winrate of 10BB/hour over some small amount of time. He knew this was a case of running good, but could still maintain that it was enough to show that he was very likely at least a 1BB/hour winner.

[Glenn]

You can still show something. Obviously, a 1BB/hr winner is more likely to win 3BB/hr for 100hrs than a 1BB/hr loser with the same SD. However, if you use the normal curve around a point that you know is an unreasonably high average the following problem arises: it will show that there is a 50% chance that his win rate is *higher* than the sample average, when, in fact, there is basically no chance it is higher. That 50% has to go somewhere! So he still might be 3 times more likely to win at least 1BB/hr than .5BB/hr, but his overall chance of winning less than .5BB/hr is greater than the normal curve will show. While the effect is probably impossible to quantify in this case, it is still important to note.

[BruceZ]

Search the archives and find Bruce Z's post explaining SD and such. It is very good, and it won't take you long to understand.

Thanks. Here's the link.

Standard deviation (http://forumserver.twoplustwo.com/showthreaded.php?Cat=&Number=344985&page=&view=&sb =5&o=&vc=1)

[Redhotman]

[ QUOTE ]
Search the archives and find Bruce Z's post explaining SD and such. It is very good, and it won't take you long to understand.

Thanks. Here's the link.

Standard deviation (http://forumserver.twoplustwo.com/showthreaded.php?Cat=&Number=344985&page=&view=&sb =5&o=&vc=1)

[/ QUOTE ]
Bruce,
I play multiple tables. Lets say I play 5 tables for an hour. Does that count as 1 hour or 5 hours?

[ChipWrecked]

Do these sessions have to be in the same game? I play different games at different levels.

My overall SD (as computed at pokercharts) is 57.19 after 536.75 hours. Win rate
-$0.66/hr. Pokercharts allows filtering by game or location.

(oh well, negative win rate /images/graemlins/frown.gif I figure I won't feel bad unless I'm still behind after 1k hours... I tell myself that losing less than $1/hr means I'm slightly ahead of the players /images/graemlins/smile.gif but behind the rake /images/graemlins/mad.gif

And one or two good sessions would put me in the black /images/graemlins/cool.gif)