Quote:
Originally Posted by erdnase17
I think people should question how these numbers were produced.
Take as an example 60s 4800 games = -1.28% EV
It appears that whoever chose these numbers made it so that the 80% confidence interval is [0.498, 0.510]. However it is much more usual to use a 95% confidence interval which would require 4800 games = 0.18% EV.
In other words, the rules were made by lowering the confidence level, while using the standard 95% confidence level would require much bigger samples or demand higher EV...
Imo allingirl777's system would be much better than this: a self-regulating system preserving individual freedom, where each member of the group is required to play x% of other members. Also less hassle managing the group and no need for these arbitrary EV requirements.
Of course everyone should question everything, not only the old system, but why so many were hanging on to the old (flawed) system. And also the new system. Questioning everything is always good.
That said, two main points:
1) EV ROI is going to result in a higher confidence than regular ROI. So if you use binomial to come up with 80% in your above example, it's actually far higher when you're talking about EV ROI, and it's not really 80%, more like 90%+.
2) When you're moving up to a new level and beating an opponent, is it really fair if you can't take his place or get into the "club" if you are 85-90% likely to be beating him? What is this "standard 95%' that you speak of? I've never heard of that. Imagine if pros waited until they were 95% certain that they beat a new level to move up. Nobody does that, it's absurd, and it's over conservative to insist that 95% is the minimum % required for someone to prove they are better than you.
All that said, I love Hester's system. It's pretty obvious that there isn't enough support to run with it, so improving the current system drastically is the very best solution that is out there, and I believe that has happened in the $60s and $100s levels.