SD in $/hand?

[JTG51]

Using data gathered from my online play using Poker Tracker I decided to calculate my SD terms of dollars per hand. I used the formula Mason provides in his essay here, substituting the number of hands played in each session for the number of hours. My result for 2/4 games (after close to 10K hands) is a little more than $6/hand. This seems very low to me since pots are routinely 10 times that much.

Am I misunderstanding something and that is in fact a reasonable number, or are my calculations wrong?

Thanks.

[Copernicus]

My guess is you are making a mistake, unless you are playing so tight that you almost never see a hand except from the blinds, and even in them you get lucky (unlikely with 10k hands).

[BruceZ]

It looks OK to me. Your SD for 40 hands would be $6*sqrt(40) = $37.95 = 9.5 bb. You typically want your SD around 10 bb for 40 hands, and lower is better.

[JTG51]

Thanks Bruce, I think I'm still a bit confused though.

If $6 is a reasonable SD/Hand, doesn't that mean that 95% of all hands should fall within $12 of my winrate and 99.7% within $18? Intuitively that just doesn't seem possible. It feels like I win or lose 3 big bets in more than 5% of the hands I play, and it really feels like I win or lose 5 big bets in more than 0.3% of the hands I play.

Am I misapplying a concept?

[ACPlayer]

Seems to me that the SD would include the plus minus on hands you dont play as well as those you do. But BZ will I am sure elaborate.

Also, I dont have MMs equation in front of me, but you should calculate the units of the equation to see if it actually $ per hand.

[JTG51]

I should have included a link to Mason's article in my original post. Here it is. (http://www.twoplustwo.com/mmessay8.html)

[ACPlayer]

MMs equation gives u SD in dollar per sq root of the time unit that you are calculating in. Not sure how it addresses your original question tho.

[JTG51]

Seems to me that the SD would include the plus minus on hands you dont play as well as those you do.

That makes a lot of sense. I suppose what I calculated would be similar to recording a session as 12 hours when I actually played for 3 and walked around the casino for the other 9.

However, as Bruce showed the number I came up with converts to a reasonable number per hour of play. Looks like I'm still a little confused.

Thanks.

[ACPlayer]

I would think your hands would include those you mucked when you counted up the wins and losses, rather than just using the hands where you invested some money. Correct? So in an hour you played 200 hands and won 300 dollars, rather than played 10 hands (the AA, KK that you actually played) and won 300 dollars -- and the former was the basis of your calculation.

If you do, walk around the casino or sit out online hands, those (times/count of hands) should not be included in your SD or EV calcs.

[BruceZ]

JTG51,

The SD you computed is fine. You want it to include hands you don't play, not just hands where you win or lose a pot. The hands where you win 0 will reduce your SD. Also, your actual fluctuations per hand will be larger than your SD/hand indicates because your results on individual hands are not distributed according to a normal distribution, so you can't use this number to predict the spread on a per hand basis. Even your hourly results won't be normally distributed. You can express the SD for 1 hand or for 1 hour, but it doesn't really have any statistical value until you multiply it by the square root of some larger number of hours or hands. That's why it has units of $/sqrt(hands) or $/sqrt(hrs). Your results over a longer period of time will be normally distributed with an SD of (SD/hand)*sqrt(hands). The benchmark value for 40 hands should be around 10 times your win rate for 40 hands.

-Bruce

[JTG51]

I would think your hands would include those you mucked when you counted up the wins and losses, rather than just using the hands where you invested some money. Correct?

Yes, that's what I was tying to say. I just made a bad analogy.

[JTG51]

Also, your actual fluctuations per hand will be larger than your SD/hand indicates because your results on individual hands are not distributed according to a normal distribution, so you can't use this number to predict the spread on a per hand basis.

This cleared everything up for me. Thanks for the help.