Author: Ralf Elvsén
Date: 12:25:58 06/07/00
Go up one level in this thread
On June 07, 2000 at 12:03:30, blass uri wrote:
>On June 07, 2000 at 10:52:39, Jorge wrote:
>
>>On June 07, 2000 at 04:15:50, Ralf Elvsén wrote:
>>
>>>Yes, and Jorge, as I understand it you wanted to calculate the probability of
>>>getting exactly 37 non-losses and 4 losses out of 41 games. This in itself
>>>is a rather special result. I looked in a table and 33.5/41 indicates
>>>a rating difference of 260 points. The question should perhaps be put
>>>"What is the probability to get a performance rating difference of
>>>at least 260 points in 41 games, given that the "true" rating
>>>difference is 65 points?"
>>>
>>>Making assumptions about the rate of draws and neglecting the complicating
>>>factors Uri mentioned, this can be worked out, but I don't have the time
>>>right now.
>>>
>>>Ralf (not Ralph :)
>>
>>My apologies for changing your name Ralf :). Yes I agree, this is what we wanted
>>to hypothesize. True, my calculations is based on that special event (little
>>rusty on Statistics, but its coming back slowly), however if we add the rest of
>>the probabilities from 37 to 41 (w+ds), the amount practically negligible, as
>>you probably already know. Uri thx for the help, but the .592 was based
>>originally on Ralf's rate on BOTH Wins and Draws in the beginning. Dont know if
>>the drawing rate is 0.30 in this case. Quote "The number 0.592 is based on wins
>>and draws".
>
>"The number 0.592 is based on wins and draws" does not mean that 0.592 of the
>games are wins or draws.
>
>It means if we assume 30% draws that 0.442 of the games are wins for Junior
>and 0.3 of the games are draws so you get 0.442+0.3*0.5=0.592
>
>Uri
Aha, so that was the source of confusion :)
Ok, this is what I got but I give no guarantee that these
numbers are correct.
What is that the probability that you get at least 33.5 points
out of 41, given a rating difference of Rdiff and the
draw frequency df ?
df = 0.2 0.3 0.4
Rdiff = 65 0,000388 0,000151 4,38E-05
Rdiff = 100 0,003485 0,001772 0,000725
Rdiff = 150 0,036797 0,025179 0,014944
I think a 100-point rating difference between Crafty and Junior
is possible, given the huge error margin for SSDF-ratings (note:
I said possible, not probable. I have no opinion).
Anyway, these probabilities are pretty small, but all of them aren't
extremely small. I would look closer at this if I were in charge of
the testing. Luckily I am not :)
Ralf
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.