Quote:
Originally Posted by 409412
I think my comment was somewhat poorly phased so let get at what I am really trying to figure out. In the elo system for chess, if you play a computer that is 2800 (super gm strength) and you win 10% you are playing about 2400 level. Because that is the expectation for a 400 pt difference. That would make you better than 99% of the people who play. I just wonder if the same relationship holds for backgammon. So I just wonder how people at various levels do.
I also have some questions about the way these programs evaluate play. If you play a 30 roll game and make 2 dubious, or 1 bad, or perhaps even 1 very bad gnu will grade you as rather a bad player. But you might be making grandmaster plays about 95% of the time. If I could correctly solve 90% of the problems Bill Robertie posts I would feel that I was a pretty good player. So the relation of error rate to player evaluation is something else I find confusing.
For the record by the way, I am not anywhere near 50% more like 5 or 10 (I have not actually kept count) I just know it beats the snot out of me. But I feel like an upper intermediate level player
I believe that there is a similar FIBS relationship in BG (Very good chance that I am wrong about this)
I used to have a spreadsheet deriving the formula so that I could test friends on GNU and then inform them of how frequently I would expect to beat them based on the FIBS spread
FWIW I really dont have any interest in how often programs defeat me, all that matters is the equity loss per move and how that changes over time.