Ratings are unstable and are statistically flawed (subject to pool expansion inflation). To evaluate my progress over time it would be nice to compare quality of my moves to a reference. A graph that allows a person to select time period for averaging (day, week, month), variant (rapid, blitz, etc.) and plots average centipawn loss of his games over time would be perfect. Basically, like a rating plot but with centipawn loss on the y axis.
Ratings are FAR more reliable than average centipawn loss. Centipawn loss is often HIGHEST in games against people at your own level or games where both sides are progressing a plan with a narrow path to success. They are lowest in extremely one-sided matches, and this is compounded by the fact they are no longer tracked once one side as a +5 or +6 advantage.
Your percentile is a much better yardstick if you insist on using something other than rating. Centipawn loss is practically useless outside of the context of a single game.
Centipawn loss is inaccurate. If you win a Queen or Rook in the first ten moves then maintain your advantage to the end your score will be 10 or under. If you play a wild game your score will be higher. I have won games where my Centipawn loss was over 60.
oooh a cheater in the forums! When your cheating at a game no better thing to do then call attention to yourself am I right?
Players who cheat do not play the Kings Gambit as white usually lose if black plays accurately and all AI give Black clear advantage after 2. f4. Please try it
This topic has been archived and can no longer be replied to.