Swing question for people who play lots of hands

[sthief09]

this is something that I've suspected and tested


the way PT calculates SD is inaccurate. if you want a SD/100 hands, it should take hands in increments of 100 hands, and find a SD per 100-hand block. it does not do this.

it takes your sessions and cuts them by 100 and treats each "100-hand block" the same. so if you play 10, 10-hand sessions, it will put them together and count them as 1 100-hand block.

so why does it make a difference? well take 2 players. one player plays 4 tables for 24 hours straight without changing tables. one player never plays more than 100 hands per session. they both play the same exact way, and for the sake of the example, against the same players

so the player who plays long sessions probably has a low SD per session. the reason for this can be proven by an extreme case. say you play a series 100,000-hand sessions. clearly your SD between the sessions would be very small. in most of them you'd make +/- .5 bb/100. now, if you put those in PT, it would take each 100k hand session and cut it up 1,000 ways. so you have 1000 different 100-hand samples with very small variance. so a person who plays long sessions will have a smaller SD/100 hands because his 100 hand blocks typically represent a piece of a whole rather than actual 100-hand blocks

I'm sure you can see why it works the other way around.

so someone like you, who I suspect probably doesn't switch tables much (if I remember correctly you 8-table 30/60 so there's not much room for changing tables) but does play long sessions, will have a low SD, like yours.

someone who 1-tables with excellent table selection would have a higher SD because he's changing tables more.

the reason I'm pointing this out is because based on your post, I'm assuming you understand how SD can be used to find out how likely you are to go through these bad runs. the problem with this is your SD is probably closer to 20 than 15 which would suggest that these runs are a lot more likely than you realize, or that PT's SD calculation would indicate

[lefty rosen]

The tigher the game with less dead money in it. The worse the swings. I went 100BB on 2/4 party for the first time ever. I aknow its because that limit has become the Sahara desert compared to what it was 2 years ago..... [img]/images/graemlins/blush.gif[/img] [img]/images/graemlins/confused.gif[/img]

[sthief09]

[ QUOTE ]
The tigher the game with less dead money in it. The worse the swings. I went 100BB on 2/4 party for the first time ever. I aknow its because that limit has become the Sahara desert compared to what it was 2 years ago..... [img]/images/graemlins/blush.gif[/img] [img]/images/graemlins/confused.gif[/img]

[/ QUOTE ]



no, the opposite is true. this is what people mix up. a tight game will affect your winrate. a lower winrate will yield greater downswings. this is why the mediocre 0-.5 bb/100 winners have enormous downswings. a tight game corresponds with a lower SD, since there will be less big pots and you will be putting in a lot less bets postflop.

also, 100 BB in limit is nothing and it says nothing about the quality of the games.

[TimM]

[ QUOTE ]
if you want a SD/100 hands, it should take hands in increments of 100 hands, and find a SD per 100-hand block. it does not do this.

[/ QUOTE ]

I think the easiest way would be to calculate SD per hand, and then multiply by 10 to get SD/100.

( Because SD/(Y hands) = SQRT(Y/X) * SD/(X hands) )

I tried this in Excel, by going to game notes, clicking get all, then click the "p", then export to excel, etc.

Just take the whole BB_won column and do an =STDEV(firstcell:lastcell)

I got a SD/hand of 1.47, for an SD/100 of 14.7. PT gives my SD/100 as 15.6.

That was for a 4400 hand sample.

For a 31000 hand sample I get: Excel: 13.9, PT: 14.4

For a 55000 hand sample I get: Excel: 13.8, PT: 14.1

That was about the most it would let me export. Looks like the two methods converge as the sample gets larger and larger.

[sthief09]

the problem is it's using PT table sessions. if you play a 300-hand session, it should not be treated as 3 equal 100-hand blocks. you could've swung up and down 250 BB in those 300 hands.

[TimM]

[ QUOTE ]
the problem is it's using PT table sessions. if you play a 300-hand session, it should not be treated as 3 equal 100-hand blocks. you could've swung up and down 250 BB in those 300 hands.

[/ QUOTE ]

I see what you mean now. But over a large enough sample this will average out too. After enough sessions, your overall session results will reflect your variance just as well as if you broke them into 100 hand blocks, and the two calculation methods will converge.

I wrote an Excel file to show this. Say your results of a 100 hand block were like coin flips. Heads +100BB, tails -100BB.

Now if you play many 400 hand sessions, one sixteenth of your sessions will be +400, one sixteenth will be -400, one quarter will be +200, one quarter will be -200, etc.

As long as you play enough sessions so that this distribution is reflected well in your session results, it doesn't matter which method you use to calculate your SD/100.