Thema: Frage: Mm5 nur 1575 ELO ??
Einzelnen Beitrag anzeigen
  #208  
Alt 17.07.2017, 14:15
Benutzerbild von spacious_mind
spacious_mind spacious_mind ist offline
Schachcomputer Koryphäe
 
Registriert seit: 29.06.2006
Ort: Alabama, USA
Land:
Beiträge: 1.927
Abgegebene Danke: 47
Erhielt 327 Danke für 197 Beiträge
Aktivitäten Langlebigkeit
0/20 18/20
Heute Beiträge
0/3 sssss1927
Re: Mm5 nur 1575 ELO ??

Sorry ich Schreibe diesmal auf Englisch weil.

I am currently trying to create an alternative evaluation system where human players are evaluated by using Stockfish 8 at Lichess https://lichess.org/ and then compared with chess programs and 3 dedicated chess computers that are currently playing in my Division 3 Swiss Chess League which you can read about here at Hiarcs:

http://www.hiarcs.net/forums/viewtop...=8453&start=30

Basically my plan is to ultimately evaluate 100 human games per chess category = 200 human players. When I have finished the evaluation I will then remove 10% from the highest end and 10% from the lowest end per category to give me a good final amount of 160 players (80 games) per category for a good final performance analysis.

So far I have created a performance chart for each category with 30 to 40 players (15 to 20 games) therefore I am only about 20% complete.

The human players are selected by taking games that have 45 moves or more in order from top of Chessbase list. Therefore I do not jump any games unless they are 44 moves or less. This is done to simulate dedicated computers who we tend to play most of the time well beyond 45 moves and it also gives a sufficient amount of moves after the opening to evaluate properly all the phases of the game.

Here is what it looks like at 20% complete:



The above example is 2000 ELO Chessbase games.

I have done this for all the categories between 2400 ELO to 1400 ELO and here is a summary of what it looks like:



As you can see you can follow the mistakes and centipawn losses quite accurately between the categories where the average 1400 ELO players makes more mistakes than a 2400 ELO player.

I have now entered these results into my Division 3 Swiss League that I am playing and it looks as below:



The chess programs have only played 2 games each so far in Division 3, but in the end they will have all played 12 games each. So I will be curious to see where the chess programs will end up by the time I finish this League and hopefully I will be able to compare accurately their performance to the Chessbase officially rate FIDE humans who played their games under Tournament conditions.

I have excluded women tournament games and any rapid or blitz games as well.

Viele Gruesse

Geändert von spacious_mind (17.07.2017 um 14:25 Uhr)
Mit Zitat antworten