| Author |
Message |
|
Posted: Wed, 29 Jun 2005, 3:41pm Post subject: Calculate your winrate |
|
|
4-of-a-Kind
Joined: 14 Oct 2004
Posts: 1791
WPP: 58
Location: PEANUT BUTTER JELLY TIME
|
|
|
|
|
|
|
|
Posted: Wed, 29 Jun 2005, 6:17pm Post subject: |
|
|
4-of-a-Kind
Joined: 12 Aug 2004
Posts: 1968
WPP: 96
Location: 6max limit tables
|
|
| What’s the source code in? I'm just wondering if you coded it or it was just out there. If you could send me that source code, that would kick ass. |
|
|
|
|
|
|
|
|
|
Posted: Wed, 29 Jun 2005, 6:31pm Post subject: |
|
|
4-of-a-Kind
Joined: 14 Oct 2004
Posts: 1791
WPP: 58
Location: PEANUT BUTTER JELLY TIME
|
|
| It's in PHP. It's not pretty, I just knocked it together in 10 mins. Send me your email address if you want me to send it to you tho. |
|
|
|
|
|
|
|
|
|
Posted: Wed, 29 Jun 2005, 6:36pm Post subject: |
|
|
4-of-a-Kind
Joined: 03 Mar 2005
Posts: 3537
WPP: 96
Location: Your place or my place
|
|
M +- (1.96X)
Where M is the mean, and X is the standard error (which is Standard Deviation / sqrt(n)?
So you take the standard deviation (input by user), devide it by the squareroot of the number of hands (input by user), and multiply it by the z score (which is 1.96, and you don't need to care why), then add it to (and subtract it from) the average winrate
Ugh I'm leaving out a step... *thinks thinks thinks* |
|
|
|
|
|
|
|
|
|
Posted: Wed, 29 Jun 2005, 6:37pm Post subject: |
|
|
4-of-a-Kind
Joined: 03 Mar 2005
Posts: 3537
WPP: 96
Location: Your place or my place
|
|
| And im a dipshit because its all explained on the 2+2 link. |
|
|
|
|
|
|
|
|
|
Posted: Wed, 29 Jun 2005, 6:38pm Post subject: |
|
|
4-of-a-Kind
Joined: 03 Mar 2005
Posts: 3537
WPP: 96
Location: Your place or my place
|
|
| Oh, and note that your hand-size should be above 10k for it to be reasonably accurate. Small sample sizes don't fit the central limit theorem |
|
|
|
|
|
|
|
|
|
Posted: Wed, 29 Jun 2005, 7:16pm Post subject: |
|
|
4-of-a-Kind
Joined: 14 Oct 2004
Posts: 1791
WPP: 58
Location: PEANUT BUTTER JELLY TIME
|
|
| euphoricism wrote: | M +- (1.96X)
Where M is the mean, and X is the standard error (which is Standard Deviation / sqrt(n)?
So you take the standard deviation (input by user), devide it by the squareroot of the number of hands (input by user), and multiply it by the z score (which is 1.96, and you don't need to care why), then add it to (and subtract it from) the average winrate
Ugh I'm leaving out a step... *thinks thinks thinks* |
I'm multiplying by 2*STDERR instead of 1.96 but I'm bad at math and I don't know why. Is this ok y/n |
|
|
|
|
|
|
|
|
|
Posted: Wed, 29 Jun 2005, 7:41pm Post subject: |
|
|
4-of-a-Kind
Joined: 03 Mar 2005
Posts: 3537
WPP: 96
Location: Your place or my place
|
|
| Definitely change it to 1.96. Thats the z score for .05 (corrosponding to 95% conf) |
|
|
|
|
|
|
|
|
|
Posted: Wed, 29 Jun 2005, 7:56pm Post subject: |
|
|
4-of-a-Kind
Joined: 14 Oct 2004
Posts: 1791
WPP: 58
Location: PEANUT BUTTER JELLY TIME
|
|
| euphoricism wrote: | | Definitely change it to 1.96. Thats the z score for .05 (corrosponding to 95% conf) |
Though I won't likely understand it, can you tell me why? Or give me some kind of link? |
|
|
|
|
|
|
|
|
|
Posted: Wed, 29 Jun 2005, 8:05pm Post subject: |
|
|
Full House
Joined: 15 Oct 2004
Posts: 666
WPP: 95
Location: Decatur, IL
|
|
| TylerK wrote: | | euphoricism wrote: | | Definitely change it to 1.96. Thats the z score for .05 (corrosponding to 95% conf) |
Though I won't likely understand it, can you tell me why? Or give me some kind of link? |
It just means that 95% of the normal distribution is within 1.96 Standard Deviations of the mean. Unless you are doing work for a project where accuracy is extremely important the difference between 2 and 1.96 will be hardly noticable and would just give us like the 95.4% confidence interval(So in our case the 2 is probably fine). |
|
|
|
|
|
|
|
|
|
Posted: Thu, 30 Jun 2005, 8:12am Post subject: |
|
|
4-of-a-Kind
Joined: 03 Mar 2005
Posts: 3537
WPP: 96
Location: Your place or my place
|
|
Yeah, thats pretty much right. Its not vital, but theres no reason not to change it and increase accuracy
Basically, to get a z-score you take your percentage of confidence wanted (as a decimal) and subtract it from 1. so 1-.95 = .05, look that up on a ztable and it corrosponds to 1.96 (95% is basically used all the time, so everyones memorized it). You could do 99% the same way, just google a z table and do 1-.99 = .01, and look up .01 |
|
|
|
|
|
|
|
|
|
Posted: Thu, 30 Jun 2005, 4:02pm Post subject: |
|
|
Where's the AI button!?
Joined: 12 Jul 2004
Posts: 4533
WPP: 87
Location: TagFish
|
|
| With a 95% confidence interval, your true winrate is between -0.9 BB/100 and 5.56 BB/100. |
|
|
|
|
|
|
|
|
|
Posted: Thu, 30 Jun 2005, 4:03pm Post subject: |
|
|
Where's the AI button!?
Joined: 12 Jul 2004
Posts: 4533
WPP: 87
Location: TagFish
|
|
| You know, that is probably the most useless thing I've done today. |
|
|
|
|
|
|
|
|
