[Bug-gnubg] A player's average chance to win a game if GNU Backgammon's "Error rate mEMG" for both players is known

[Adi Kadmon]

Hello all,

I once posted here the question whether there were any data about each player's average chances to win a game, if GNU Backgammon's calculated  "Error rate mEMG" for both players are known (based on a large sample of games, and always available through GNU's accumulative "relational database").

As a reply I was referred to a table of chances based on "Elo rating", used in backgammon servers. Although this is nice by itself, I would much prefer the above data. That's because the "Elo rating" fluctuates much according to a player being new or old in the server, luck, "good/bad periods" etc; while, given many games as a basis, GNU's Error rate mEMG becomes much more stable and reliable.

So, for instance, where can I find data or formulas about the chances of  player "A" (average Error rate mEMG=13.33) to win a single money game against player "B" (average Error rate mEMG=18.50)?

Or, for that matter, how would GNU fare playing against itself over a very large number of games with settings to match these different Error rate mEMG's?

I guess I could run some such experiments with GNU playing against itself. But, of course, I'd prefer ready-made data if they exist. Indeed, I would be very surprised to learn that they don't, since they would be very valuable for players on real money through the web.

 

Thanks in advance for an answer based on reliable sources,

-- Adi