|
BGonline.org Forums
PR or Luck? Here is the real answer if you are looking for (Long but informative)
Posted By: Matt Cohn-Geier In Response To: PR or Luck? Here is the real answer if you are looking for (Long but informative) (Matt Cohn-Geier)
Date: Wednesday, 10 February 2016, at 10:09 p.m.
1 - is already long established and in fact considerably weaker than the conclusion that others have come to or attempted to prove - that the luckier player ALWAYS wins. This is due to how "luck" and "luck factor" is defined. In any case, if a player is less than a 100% favorite before a match starts and then wins the match (hence performing better than expectation), then of course he got lucky to win.
2-4 - this was not a study, I just did the calculations. It is in fact possible to calculate this because PR is by definition a measure of how many errors you made, in other words how much % chance of winning you give up, or in other words, how big of a favorite you are, and hence what the mean is. Basically very similar to Rick's post about ELO. If you know what the odds are that you will win a match then you can also calculate the variance, standard deviation, confidence intervals, etc.
Of course there are some problems with trying to apply this practically. For instance, maybe the 6.3 player can make adjustments and not lose as often as expected against the 3.7 player. On the other hand, maybe the 3.7 player can make adjustments and win MORE often than expected against the 6.3 player. Anyway, it is definitely a more accurate measure of skill to say "player A who wins 58% against player B" than to start involving imperfect third party judges like XG, start doing data collection, etc. So calculating the mean and std dev is much less prone to error and bias than trying to accumulate data and run simulations on it. For example, you are not hindered by sample size.
You could do similar calculations involving a 10+ player. I am pretty sure a 3.7 player would be a huge favorite there.
|
BGonline.org Forums is maintained by Stick with WebBBS 5.12.