Final rankings extra details.

[Janzert]

For those interested the final "ratings.txt" output from bayeselo can be found at .

Also for the final rankings I had bayeselo run it's likelihood-of-superiority calculation, the LOS matrix result from that can be found at (warning it's a 81MB text file with lines over 18,000 characters long).

One possibly interesting thing to pull out of the LOS is how far up and down the rankings you have to go from a particular bot to find the point where bayeselo believes there is a less than 10% or greater than 90% chance of superiority for the rest of the ranks. That information can be found in . I had planned to integrate this into the ranking display but I changed my mind since I don't want to change the database schema and the ranking display code just for this one time (it takes too long to calculate the los during an ongoing contest).

Janzert

[iouri_]

Thanks for the calcs!

95.6% probability that I beat Slin, I guess that means that I pretty definitely got 2nd place. That's reassuring. And bocsimacko pretty definitely got 1st, but we already knew that. Pretty close finish for Astek and jimrogerz though, they practically share the 4th place.

[SUA.khb]

626 9309,SUA.khb,196443 698 33 33 525 68% 544 0%
The 4th win-% result =)

[barabanus]

This los matrix isn't totally correct, there're some players that would win much more than 0.1% vs 99.9% against much less ranked players, though we see 0 vs 999 in the appropriate matrix cells.
May be that's because elo rating system is based upon normal distribution of performance which isn't a precise suggestion when applied to bots.