Gashimov Memorial, Round 2 is LIVE and Open to ALL MEMBERS!Click here to watch!
Upgrade to Chess.com Premium!

Chess rating system


  • 2 years ago · Quote · #1181

    andreic

    @DavidMertz1

    Below are details about your observations. Except for the first aspect, I think the orginal  model is fine.

     

    1.  Promotions

     You are right about promotions , i did not think of how this will modify the score.  It will be correct to consider that pawns keep their value until end even if they promote to queen or someting else. The formulas are the same. 

    2. How to win

    You should not think if it's better to mate quickly or capture all oponents pieces and then mate. If you play at your best the score will be higher.  "it's better to just take the mate in 2 " as you said, the formula will ALWAYS give a higher score if you mate faster.

    I think you did not understand that the loser material does not count, i will try to explain better:

    "the score for the loser is: -material_winner/game_moves".The formula is the same as in case of a draw, except that the loser material is 0. Giving a mate to your oponent means the game is over ,and the loser material becomes useless(because it cannot play anymore) or  0.

    3. How to lose

    If a mate is predictable you should not resign , but avoid the mate as long as you can.

    4. Obvious draws.

    For the third situation (obvious draws because insuficient material) you can just stop the game, but consider the number of moves to be current_move_number + 50. The score will be accurate(close to zero for both sides) and you don't have to play 50 moves because you already know the material at the end, and the game length.

     

    As a general rule you should continue to play as long as the game length OR the final materials cannot be predicted. 

    The basic  idea is that the result should reflect the material gaining speed expressed as pawns/move.

    Examples:  A score of 1 means you gain on average 1 pawn per move.

    On the other side winning by promoting a single pawn to queen, after 50 moves  wuld give a score of just (1+4)/50=0.1 , if the king has 4 unit value.

    A  mate in for would have have much higher score , somewhere > 6.

    A draw with bishop and king versus king in 50 moves would have a score of 

    3/50=0.06(for player with the bishop) 

    and -3/50=-0.06 (for the other player).

  • 2 years ago · Quote · #1182

    GoldenGlory34

    Cool

  • 2 years ago · Quote · #1183

    Cenro

    Someday I expect to hit the double digits. 

  • 2 years ago · Quote · #1184

    Dodger111

    [COMMENT DELETED]
  • 2 years ago · Quote · #1185

    tommcgrath

    yes definetly d best !!! u r (wink wink he not the best)

  • 2 years ago · Quote · #1186

    Dodger111

    [COMMENT DELETED]
  • 2 years ago · Quote · #1187

    Zut1975

    I'm 99% sure Queen is worth 9, not 10.   Just an FYI.

  • 2 years ago · Quote · #1188

    Ziggyblitz

    I checked the rating adjustment for a game against an opponent rated 320 points below me, and was surprised to find that a loss costs me 20 points.  I've lost to players rated even lower and it hasn't cost me as much, usually 17~19 points.

    I can only guess that the reason is because of my opponent has a low Glicko RD which is 43, as he is a very active player.  Most opponents have an RD of around 60.  Shows the subtley of the Glicko at work. 

    I wonder what the lowest RD ever reached on chess.com is ?

  • 2 years ago · Quote · #1189

    tommcgrath

    yes me 2

  • 2 years ago · Quote · #1190

    tommcgrath

    Dodger111 wrote:
    oldbones wrote:

    yes definetly d best !!! u r (wink wink he not the best)


     Oh yes I yam....i got like 10 wins an one loss or something, that's like supar good

    wow well get yuor ratinjg higher and you weill be the best

  • 2 years ago · Quote · #1191

    chessmaster299o

    Wow, just that simple sentence or two recieved 15+ pages of comments.Smile

  • 2 years ago · Quote · #1192

    GordieSea

    Ii think the knight should be worth 3.5 relative to the bishop being worth 3.0.
  • 2 years ago · Quote · #1193

    arborvitian

    Hunadora wrote:
    But i think we have all have games where we played much weaker players and blew the game by not thinking as well as we should.

    I do!  I do!

    I went up against this one guy I never expected to beat in my wildest dreams, and I won three times in a row.  On the other side of the ledger, I completely screwed up and lost to someone who had a queen and a half dozen or so pawns against a king and three, and couldn't work out how to put me out of my misery.

    I should have tried harder to draw that last game.  Oh well.

  • 2 years ago · Quote · #1194

    Ziggyblitz

    ChristianSoldier007 wrote:

    actually someone did get zero, i don't remember who


    A zero rating would be achievable but you'd most likely need to cheat by using multiple accounts.  This would be the reverse of using multiple accounts to boost your rating.  I can't see anyone ever bothering to do such a useless thing.

  • 2 years ago · Quote · #1195

    Ziggyblitz

    waffllemaster wrote:

    It's more like a 1700 FIDE rating is a 2000 chess.com turn-based rating.

    Live chess is a bit closer I think, but I don't play here much either, and some of the ratings have changed.


    Everything I've read on chess.com says that comparisons between different rating systems/pools are very difficult to make.  However I also read that FIDE ratings are on average, 200~300 points LOWER than chess.com turn-based.  Same as Chess 960 ratings are on average 300~400 points lower than standard turned-based ratings, (different pools of players + less games played).

  • 2 years ago · Quote · #1196

    fasttime

    A rating can give you sort of weakness . Do not look at the opponent's rating when

    you play, find a way to cover it up so it   does not matter anymore.

  • 2 years ago · Quote · #1197

    Ziggyblitz

    fasttime wrote:

    A rating can give you sort of weakness . Do not look at the opponent's rating when

    you play, find a way to cover it up so it   does not matter anymore.


    I agree.  However I know that I can lose to players rated way below me and win against players rated above me, so I always try to play carefully...and you never know if and when your opponent decides to use some engine assistance.

  • 2 years ago · Quote · #1198

    MuskieMan33

    How about the different game types and the points associated with those?
    I was wondering if anyone could break this down for me, I'm a little confused as to why a 10 min game I only get 15-30 points, and then I play a 15|10 game and will get 100+ points for a win... Is it that because I'm not an extremely high rated player? Or is it a combination of who I play, their rank, my rank, and how the game plays out (ie. if I lose with a majority of my pieces remaining on the board or lose with just my king on the board).

  • 2 years ago · Quote · #1199

    AndyClifton

    It's doubtless because you've played far less games in Standard.

  • 2 years ago · Quote · #1200

    DavidMertz1

    MuskieMan33 wrote:

    How about the different game types and the points associated with those?
    I was wondering if anyone could break this down for me, I'm a little confused as to why a 10 min game I only get 15-30 points, and then I play a 15|10 game and will get 100+ points for a win... Is it that because I'm not an extremely high rated player? Or is it a combination of who I play, their rank, my rank, and how the game plays out (ie. if I lose with a majority of my pieces remaining on the board or lose with just my king on the board).

    The pieces on the board do not matter.  What matters is your rating vs the rating of your opponent, and how accurate the algorithm thinks the ratings of you and your opponent are.  For example, if your opponent has played very few games recently, your rating will not move as much because their rating may not be as accurate.  But if YOU have not played many games recently, then your rating will move MORE because YOUR rating is deemed to be less accurate.


Back to Top

Post your reply: