Author: Christophe Theron
Date: 14:59:29 11/17/99
Go up one level in this thread
On November 17, 1999 at 03:06:29, Stephen A. Boak wrote:
>On November 16, 1999 at 07:32:57, Christophe Theron wrote:
>
>>On November 16, 1999 at 06:49:28, Stephen A. Boak wrote:
>>
>>>On November 16, 1999 at 02:16:10, Christophe Theron wrote:
>>>
>>>>On November 15, 1999 at 23:01:17, Stephen A. Boak wrote:
>>>>
>>>>>Based on the following reported results for Chess Tiger 12.0 at 40/2hr time
>>>>>controls, and assumed ratings shown for the opponents, I calculate the
>>>>>Performance Rating for Chess Tiger for 122 games at 2726.
>>>>>
>>>>>SSDF Chesstiger12.0 (A450)-Fritz5.32 (P200MMX), 2575, 30,5-13,5
>>>>>SSDF ChessTiger 12 K6-2 - Shredder 2 200MMX, 2503, 28.5-11.5
>>>>>ChessTiger12-Fritz6, 2681(?): +5 =1 -0 Thorsten Czub
>>>>>SSDF Chesstiger12.0 (equal)-Hiarcs 7.32 128MB K6-2 450 MHz, 2646, 19.5-12.5
>>>>>
>>>>>--Steve Boak
>>>>
>>>>What happens to the rating if you take out the games played by Thorsten? Because
>>>>there is no evidence that Fritz6 is as strong as Fritz5.32...
>>>>
>>>>
>>>> Christophe
>>>
>>>Without the Thorsten games, ChessTiger 12.0 is 2711 for 116 games. Yes, I
>>>assumed Fritz6 was equal (absent evidence) to Fritz5.32. Dropping those 6 games
>>>is not very significant since so many games have elsewhere been reported against
>>>opponents with established ratings.
>>>
>>>--Steve
>>
>>OK, thanks.
>>
>>Do you have a program or a spreadsheet to compute this?
>>
>>
>> Christophe
>
>I use a very simple spreadsheet with a very simple formula. Can you read an
>Excel file? or another style spreadsheet? Let me know which one you can read,
>and I'll send it to you by eMail.
Yes I can read Excel97 files.
>The simple formula (it can be very easily calculated by hand or programmed in
>whatever language you want to use) is:
>
> A. Add up all opponent ratings for *all* the games,
>then
> B. Calculate (Total ChessTiger 12.0 score - Total Opponent score)*400,
>then
> C. Divide the grand sum (A+B) by the total number of rated games.
>
>It's that easy.
>
>NOTE--for the Step B calculation, you can simply use
> (CT12WinQty - CT12LossQty)*400
>if it is easier to use the Win and Loss counts for CT12, instead of Total Scores
>for CT12 and its Opponents. Here is another way of looking at the same
>calculation and how to 'program' it:
>
>[RtgOpp1*QtyGamesVsOpp1 +
> RtgOpp2*QtyGamesVsOpp2 +
> RtgOpp3*QtyGamesVsOpp3 + ...
> RtgOppN*QtyGamesVsOppN +
> (CT12TotScore - OppTotScore)*400] / TotRatedGames
>
>I call this the +/- 400 (plus minus 400) rule for Performance Rating
>calculation. It is the same as averaging your individual game performance
>rating for each game, based on:
>A. If you win, you get your opponent's rating plus 400 pts.
>B. If you lose, you get your opponent's rating minus 400 pts.
>C. If you draw, you get your opponent's rating.
>
>This, by the way, is the basic formula used to establish a provisional rating
>for a new USCF player in the United States. After approx 20 games or so, the
>provisional rating becomes a 'permanent' rating, the +/- 400 rule is no longer
>used, and the USCF then uses a version of ELO rating formulas to alter the
>'permanent' rating thereafter.
>
>The +/- 400 rule is used, however, any time it is desired to calculate the
>Tournament Performance Rating (TPR) for a player in a particular tournament.
>The tournament can be USCF or FIDE type, or any similarly rated, ELO-based
>system. If I received a TPR of 2850 in a FIDE tournament, that would mean my
>performance level (score obtained versus my particular opponents and their
>particular ratings) was about the same as what Kasparov would be expected to
>obtain, if he played those same opponents instead of me. :)
>
>It might work also, over a large number of games, on another rating system, such
>as the English or British rating system. But the magnitude 400 is probably
>scaled for mathematical ELO-systems like USCF and FIDE. You often see the TPR
>ratings shown, for example, in the London Chess Centre's This Week in Chess
>(TWIC) reports of major tournaments (as a column in the results crosstable). In
>this case it is used as a measure of how well a player performed in a single
>tournament, against a certain field of opponents.
>
>In this instance, since CT12 opponents are rated at the SSDF ELO-based system,
>the Performance Rating I calculated for CT12 may be compared to other comp-comp
>ratings issued by SSDF. Thus is approximates the SSDF rating that CT12 has
>earned so far due to its score versus those particular SSDF-rated opponents.
>
>I keep my personal human results, including score and opponent rating, on a
>spreadsheet. I use the TPR +/- 400 rule to calculate my TPR, tournament by
>tournament, as well as my 'running TPR' (which changes game by game) over the
>last 50, 25 and 15 games. These TPR figures are easily graphed by my
>spreadsheet program (Excel)to track my performance trend over time. I even
>separately calculate my TPR with White pieces, then my TPR with Black pieces, so
>I can see my trends there, analyze my strengths and weaknesses, and improve my
>worst areas. As you would expect, the TPR with White pieces is normally always
>higher than the TPR with Black pieces. It is interesting to calculate how much
>different those TPR ratings are--if you like statistics like I do! :)
>
>--Steve
Thanks a lot for all these explanations. I have printed you message and will
keep it somewhere for future references.
What you explain sheds some light on things I have heard several times but was
not able to understand.
I assume the SSDF is using these rules too?
Christophe
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.