![[Bodog/Bovada/Ignition] Unofficial Thread](https://d30akgybcov179.cloudfront.net/icons/hamburger.svg){kind=icon}
![[Bodog/Bovada/Ignition] Unofficial Thread](https://d30akgybcov179.cloudfront.net/icons/close-sidebar.svg){kind=icon}
[Bodog/Bovada/Ignition] Unofficial Thread
![[Bodog/Bovada/Ignition] Unofficial Thread](https://d30akgybcov179.cloudfront.net/icons/quote.svg){kind=icon}
05-18-2014
, 03:06 AM
05-18-2014
, 03:50 AM
05-18-2014
, 09:40 AM
05-18-2014
, 09:47 AM
Quote:
Phatty is correct. iPlayPLOhigh is wrong.
We are not talking about estimating winrates. Obviously we need larger sample sizes for that.
We are talking about estimating the proportion of winning players. We do not need nearly as large of a sample size for that.
Phatty also acknowledged that it wasn't online. But he offered something. When making inference on humans, data on monkeys is better than no data at all. And in this case, for this particular question at hand, I think the leap is much much smaller. The leap may be bigger for the question of estimating winrates.
We are not talking about estimating winrates. Obviously we need larger sample sizes for that.
We are talking about estimating the proportion of winning players. We do not need nearly as large of a sample size for that.
Phatty also acknowledged that it wasn't online. But he offered something. When making inference on humans, data on monkeys is better than no data at all. And in this case, for this particular question at hand, I think the leap is much much smaller. The leap may be bigger for the question of estimating winrates.
05-18-2014
, 12:18 PM
Sigh, statisticians hate graphs like this. It looks like they did a grouped histogram, but presented it as a line graph, and without telling the group size or group centers. Just from looking at it, I'm assuming they used a group size of 1BB/100 and centered each group on the whole numbers. Even once you figure all that out (or make an educated guess at it), it's hard to use that graph for any actual data without gridlines (or using a drawing program to add your own), since all that really matters is the exact value of the heights of the groups -- the lines connecting them are just fluff. Even a simple scatter plot of each group would have been 10x better.
05-18-2014
, 01:49 PM
05-19-2014
, 11:39 AM
05-19-2014
, 11:55 AM
Last edited by iPlayPLOhigh; 05-19-2014 at 11:56 AM.
Reason: Just kidding, would immediately move and never log on again
