Author: Chris Whittington
Date: 06:43:28 11/05/97
Go up one level in this thread
On November 05, 1997 at 09:34:46, Andreas Mader wrote: >On November 05, 1997 at 07:45:44, Chris Whittington wrote: > >> >>On November 05, 1997 at 07:10:19, Bas Hamstra wrote: >> >>>[Chance best program wins tournament] >>> >>>>>1) With the number of rounds at the WMCC, what was the actual probablity >>>>>that the best chess program would win the tournament? >>>> >>>>Well, I suppose that depends on how much better the best chess program >>>>is compared to all the others... Let's say that the best program is >>>>72 points better (this converts to a 60% chance to win each game) than >>>>the other 33 entries. A quick back of the envelope calculation with >>>>lots of simplifying assumptions shows that the chance of the stronger >>>>program winning the tournament is about 10%, nothing like the 80% >>>>you'd like. >> >> >>Urrrgghhhhhh. What is this maths ? From Alice in Wonderland ? >> >>Methinks the general problem is that of chasing the holy grail of the >>'strongest' program. Maybe there is no such thing ? If you find it it >>runs away and becomes another version ? >> >>> >>>Thank you. I suspect in your calculation the winner must win all games? >>>That isn't necesarrily the case of course, but it gives an indication. >>>Anyway I'd appreciate it if you would e-mail me the calculation. Hm, 10% >>>would be no less than a lottery indeed. >> >>This is silly. Obviously Junior is a strong program. You can probably >>say that Ananse is a weak program. >> >>What do you want, everything pinned down to 50 decimal places ? This is >>KK mathematical logic, only allowed on rgcc :) >> >>Chris Whittington >> > >Hehehe! > >Maybe my password for CCC will be canceled by Chris for this It needs a committee to cancel passwords. You have input into this committee :) > but I wrote >a little program (with many simplifications, too) and 'played' 100.000 >times a 11 round tournament with 33 equally strong programs and one with >a 60 per cent win probability. Well there's the error. In the real event the other programs don't have an equal probability. Ananse appears to have a 0% win probability. > In the end the 'stronger' program has the >most points in 28.5 per cent of all the tournaments (many times together >with at least one other program). Not surprising, they are all equally strong programs. > >What I have to do now is to evaluate all 34 programs with certain >probabilities and write a swiss tournament manager to make this >simulation even better and to please Chris.... :) Thanks. you can entitle the paper "Andreas in Wonderland" and put it in as a dissertatiion for a doctorate award :) Should get one (60% probability) if the other papers by the usual culprits are anything to go by :) > >Andreas > >"Nobody will ask in a few years if a program played 'nice games' in >Paris, but maybe someone will ask for the winner." This is true. All that is left in the end is crass materialism :( Chris Whittington
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.