| |
BGonline.org Forums
Error rate calculated how means what, exactly?
Posted By: Phil Simborg In Response To: Error rate calculated how means what, exactly? (Daniel Murphy)
Date: Wednesday, 14 October 2009, at 9:21 p.m.
1. A good argument can be made that ELO is a better measure of skill than ER, as ER rewards for forced plays;
2. Howard Ring once told me that the only ER that matters is the difference between you and your opponent. For example, if you cube a non-double and he drops, the ER does not reward you.
3. Even a large sampling of matches does not necessarily provide a good comparison of skills--for many reasons. If all of A's matches were recorded in the finals of major tournaments, and B's were from computer or on-line games, there is a huge difference. If A's matches were mostly 11 points and higher, and B's were 3 or 5 point matches, again, a huge difference.
4. Even with a relatively large sampling, there could easily be differences based on one player having to face many tougher decisions.
5. A true comparison of skill, in my opinion, would have to be from a large test of decisions from identical positions, either by playing "duplicate backgammon" (I have a sample if you would like to see it), or by being tested. Even this does not directly tell us who might be tougher over the board.
It would be nice if we really could tell who is best, and who is better than who if the players are close, but I believe the luck factor in backgammon will always make this very difficult. I think we can determine fairly well if one player is better than another if the difference is significant, but otherwise, how could we be sure. Does anyone believe the best player in the world (if we knew who that was) would be more than a 2 percent favorite over the 10th best player in the world in a 13 point match? Would you lay 2 percent for any player ranked 11-32 against any other? The skills are too close and the luck is too great when the skill is that close.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.