Standard deviation

[aslowjoe]

Is there any actual usefullness in knowing your standard deviation other then just for the sake of knowing.

[Grisgra]

[ QUOTE ]
Is there any actual usefullness in knowing your standard deviation other then just for the sake of knowing.

[/ QUOTE ]

If it's really high, like mine, then you're probably a fish.

If you know it's high then you know you can expect wider fluctuations. If you know it's low, and you're on a grim downswing, it's more likely that you suck than that it's just bad luck.

[Mig]

What is a high SD ? What is your SD ? Mine is below 100 and over 0... dunno if that's good ...

[Grisgra]

[ QUOTE ]
What is a high SD ? What is your SD ? Mine is below 100 and over 0... dunno if that's good ...

[/ QUOTE ]

My SD is around 20.5. Average SD for SH is around 16-17. Some tighter players have a lower SD, some looser ones, slightly higher. I have no freakin' idea why mine is so high. I'm not THAT LAGgy.

[Mig]

SD helps to know if you are a winning player or not right?

[sthief09]

you probably play shorter table sessions. PT's calculation is inaccurate. they're giving you a SD in BB/per table session, not per 100 hands. if you play 10 hand table sessions, they will be volatile but if you play 10 hour tble sessions they will be less volatile. if they're going to display it in terms of BB/100, it SHOULD take your $ won or lost after every block fo 100 hands you play, and find the standard deviationo f that

[RunDownHouse]

Wow, thanks for pointing that out. My SD is pretty high too, but I've also been following the Grisgra method of switching tables if I've taken a bit of a knock. Funny.

[sam h]

[ QUOTE ]
PT's calculation is inaccurate. they're giving you a SD in BB/per table session, not per 100 hands.

[/ QUOTE ]

Are you looking at the SD by going to the session tab and clicking "more detail"? Because it is labeled SD/100 there.

[Grisgra]

[ QUOTE ]
you probably play shorter table sessions. PT's calculation is inaccurate. they're giving you a SD in BB/per table session, not per 100 hands. if you play 10 hand table sessions, they will be volatile but if you play 10 hour tble sessions they will be less volatile. if they're going to display it in terms of BB/100, it SHOULD take your $ won or lost after every block fo 100 hands you play, and find the standard deviationo f that

[/ QUOTE ]

I'm almost positive you're wrong.

[Grisgra]

[ QUOTE ]
Wow, thanks for pointing that out. My SD is pretty high too, but I've also been following the Grisgra method of switching tables if I've taken a bit of a knock. Funny.

[/ QUOTE ]

It's a good system, one that Grisgra need to employ more often /images/graemlins/smile.gif.

[I Play 2 Ski]

Do you switch even if the table seems condusive to profiting?

[I Play 2 Ski]

its per 100 hands and also by hour

[sthief09]

[ QUOTE ]
[ QUOTE ]
you probably play shorter table sessions. PT's calculation is inaccurate. they're giving you a SD in BB/per table session, not per 100 hands. if you play 10 hand table sessions, they will be volatile but if you play 10 hour tble sessions they will be less volatile. if they're going to display it in terms of BB/100, it SHOULD take your $ won or lost after every block fo 100 hands you play, and find the standard deviationo f that

[/ QUOTE ]

I'm almost positive you're wrong.

[/ QUOTE ]


take a sample of hands in PT. export them to excel. figure out the standard deviation manually by using table sessions. compare it to the PT standard deviation. they are the same. I ahve proven this

[sthief09]

[ QUOTE ]
[ QUOTE ]
PT's calculation is inaccurate. they're giving you a SD in BB/per table session, not per 100 hands.

[/ QUOTE ]

Are you looking at the SD by going to the session tab and clicking "more detail"? Because it is labeled SD/100 there.

[/ QUOTE ]


that's what I'm saying. they giev you it in terms fo BB/100, but it's not a standard deviation based on a collection of 100 hand samples. it is based on a collection of table sessionswhich are of varying lengths

[sthief09]

[ QUOTE ]
its per 100 hands and also by hour

[/ QUOTE ]


you're not reading what I'm writing. in order to give SD in terms of BB/100 it should be based on 100-hand samples. it is not. it is based on table sessions. it is NOT the same as winrate, which is a simple, linear equation. there are squares and a square root in the SD formula, and attempting to "flatten" it into pretend 100 hand samples is inaccurate unless you play 100 hands everytime you sit down at a table.

[I Play 2 Ski]

I got ya now, it just represented as SD/100 in PT. If you say its calculated incorrectly, I believe ya /images/graemlins/laugh.gif

[Grisgra]

[ QUOTE ]
Do you switch even if the table seems condusive to profiting?

[/ QUOTE ]

If it's an obviously sweet table, probably not, but I have to face facts, if I'm down 30BB at a table and I can't put my finger on any specific huge bad beats . . . maybe they're not as bad as I thought. And my table image obviously sucks . . . and let's face it, they're probably not going to be geniuses at the next table I hit. So why not move?

[sthief09]

losing perpetuates losing. I am a big believer in this too. it makes them play more correctly than what's natural for them

[sam h]

[ QUOTE ]
that's what I'm saying. they giev you it in terms fo BB/100, but it's not a standard deviation based on a collection of 100 hand samples. it is based on a collection of table sessionswhich are of varying lengths

[/ QUOTE ]

Ahh. That is peculiar because it really shouldn't be that hard to simply arrange your hands chronologically, break them into 100 hand cases, and look at the SD in that data set. I just always assumed this is what the program was doing.

[sthief09]

me too, until I tested it.


fortunately, SD doesn't really matter. I always assume it's 18. just seems to make sense

[Grisgra]

[ QUOTE ]

Ahh. That is peculiar because it really shouldn't be that hard to simply arrange your hands chronologically, break them into 100 hand cases, and look at the SD in that data set. I just always assumed this is what the program was doing.

[/ QUOTE ]

I always assumed that it was calculating a weighted SD/100 using something akin (or exactly like) the maximum-likelihood equation Mason gives for said calculation in the back of Gambling Theory and Other Topics.

[sthief09]

[ QUOTE ]
[ QUOTE ]

Ahh. That is peculiar because it really shouldn't be that hard to simply arrange your hands chronologically, break them into 100 hand cases, and look at the SD in that data set. I just always assumed this is what the program was doing.

[/ QUOTE ]

I always assumed that it was calculating a weighted SD/100 using something akin (or exactly like) the maximum-likelihood equation Mason gives for said calculation in the back of Gambling Theory and Other Topics.

[/ QUOTE ]


when I used Excel I used the GTAOT formula and it came out exact. I'm just using common sense when I say that they yield different numbers, and haven't proved that. I don't know how different. but think of it this way. you have played 10k sesssions. is your SD going to be higher if they were 100-hand sessions or 10,000-hand sessions? even thoguh there is an attempt to flatten it by dividing by number of hands, it's not the same, as the 10,000 hand sessions will be a lot less turbulant than the 100-hand sessions.

[SomethingClever]

I play pretty short sessions, which means that my STDV should be high, right?

It's 14.2 this month for a 13k sample.

Last month it wasn't much higher for a 15k sample.

What gives?

[Grisgra]

[ QUOTE ]
I play pretty short sessions, which means that my STDV should be high, right?

It's 14.2 this month for a 13k sample.

Last month it wasn't much higher for a 15k sample.

What gives?

[/ QUOTE ]

No, sthief is on drugs, the equations used to calculate it standardize everything. Still, that's a really low SD. Your variance (SD**2) is literally only half what mine is. Jerk.

[sthief09]

[ QUOTE ]
[ QUOTE ]
I play pretty short sessions, which means that my STDV should be high, right?

It's 14.2 this month for a 13k sample.

Last month it wasn't much higher for a 15k sample.

What gives?

[/ QUOTE ]

No, sthief is on drugs, the equations used to calculate it standardize everything. Still, that's a really low SD. Your variance (SD**2) is literally only half what mine is. Jerk.

[/ QUOTE ]


if you have a winrate of 2 BB/100 and have a one-table, 1000-hand session where you win 2 BB/100, PT will give it a regression value of 0, or 10 zero-variance 100-hand sessions. however, it's possible that you lost 100 BB then won 100 BB in the same session, and if you mapped each 100-hand block sequentially, you'd have a huge standard deviation

conclusion: not on drugs

[sthief09]

[ QUOTE ]
I play pretty short sessions, which means that my STDV should be high, right?


[/ QUOTE ]


no, it means it should be higher than your actual rate, if your average table session is <100 hands with low table-session-length variance

[SomethingClever]

[ QUOTE ]
Your variance (SD**2) is literally only half what mine is. Jerk.

[/ QUOTE ]

It does seem remarkably low. In 28k hands, the total range of my variance is about 200 BB.

Which actually blows when you think about it.

[Grisgra]

[ QUOTE ]

if you have a winrate of 2 BB/100 and have a one-table, 1000-hand session where you win 2 BB/100, PT will give it a regression value of 0, or 10 zero-variance 100-hand sessions. however, it's possible that you lost 100 BB then won 100 BB in the same session, and if you mapped each 100-hand block sequentially, you'd have a huge standard deviation

conclusion: not on drugs

[/ QUOTE ]

Well yes, for one 1000-hand session.

But if somebody has 100 100-hand sessions vs 10 1000-hand sessions, technically they should end up with about the same SD/100 over the same number of hands. I agree though that you lose some detail in the 1000-hand sessions (basically, your estimate of the SD isn't as good).

[krishanleong]

[ QUOTE ]
Is there any actual usefullness in knowing your standard deviation other then just for the sake of knowing.

[/ QUOTE ]

Once in a while you'll have a real shitty day and you can look at your standard deviation and find out it's like 8/100. This might mean you received less playable hands then normal or were forced to fold the flop a ton resulting in a low stdv.

I pretty much never look at it.

Krishan