Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: are you serious ?

Author: Dave Gomboc

Date: 23:54:31 12/16/99

Go up one level in this thread


On December 17, 1999 at 01:47:28, Dann Corbit wrote:

>On December 17, 1999 at 01:36:19, Dave Gomboc wrote:
>>On December 16, 1999 at 20:08:25, Dann Corbit wrote:
>>
>>>On December 16, 1999 at 19:09:09, pete wrote:
>>>>On December 16, 1999 at 18:50:55, robert michelena wrote:
>>>[snip]
>>>>>Seriously, my highest rating was around 1620.
>>>[snip]
>>>>then if you are really serious which I tend to believe to some extent look at
>>>>the ELO system ; now assume for one second the progs play at about 2500 USCF .
>>>>Ok ?
>>>>
>>>>Ok , you have your own experiences ( them progs are simply unbeatable , which is
>>>>predictable as the rating difference should be about 900 points to you  ) , but
>>>>now think about a player rated about 2000-2100 USCF which is _far_ away from
>>>>master strength ; see the number of points he can expect from the top programs ?
>>>>
>>>>Do you think you really are competent to make a fair judgement here ?
>>>
>>>Using the above as a 'frinstance to model with,
>>>The oft repeated table:
>>>
>>>Win expectency for a difference of 0 points is 0.5
>>>Win expectency for a difference of 100 points is 0.359935
>>>Win expectency for a difference of 200 points is 0.240253
>>>Win expectency for a difference of 300 points is 0.15098
>>>Win expectency for a difference of 400 points is 0.0909091
>>>
>>>2500 - 2050 = 450.
>>>Between 9 % 5% of points will be won by that difference.
>>>An occasional win should not be at all surprising.  With 100 gmaes played, if
>>>your rating were 2100, you should get 9 points (on average).  Anything from 18
>>>draws to 9 wins.
>>>
>>>Win expectency for a difference of 500 points is 0.0532402
>>>Win expectency for a difference of 600 points is 0.0306534
>>>Win expectency for a difference of 700 points is 0.0174721
>>>Win expectency for a difference of 800 points is 0.00990099
>>>2500 - 1620 = 880.
>>>Between 1% and 1/2 of 1% of the points will be one (much closer to 1/2 of 1%)
>>>So play 100 games under tournament conditions to get one draw.
>>>
>>>Win expectency for a difference of 900 points is 0.00559197
>>>Win expectency for a difference of 1000 points is 0.00315231
>>>
>>>I don't think (however) that an argument from math will prove effective either.
>>>
>>>I'll bet that the really good players like Vincent score remarkably well against
>>>programs (unless their Achille's heel is tactics).
>>
>>The normal distribution does not accurately predict the occurance of large
>>upsets.  Elo himself discussed this in his book, which you can reference for
>>further details.
>Of course, for a single contest, it does not accurately predict anything.  So
>what does he say about large upsets?  More frequent that predicted?  Less
>frequent?  I don't have the book (does anyone know where to buy it from?)

Upsets are more frequent than predicted when using a normal distribution.

What I don't understand is why people don't take real tournament results and
figure out the correct distribution.  Maybe they actually do... <shrug>

Dave



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.