Tell me the elo!
About the Clues
- Players A, B, and C are players from different skill levels.
- The rating gap between players is around 600.
- 5 clues (groups of graphs) will be given with player labels.
- 1 of the clues has players mislabeled (randomly switched around). If the clue contains multiple graphs, they are all mislabeled.
- All of them are from 1 + 0 bullet games with ±200 rating of the player.
- Guess the order in the comments and I will read through it later!
The games were exported from chess.com.
Clue 1: Game Termination
Clue 2: Time management
This is the average time left by move number by players.
This is the average time spent per move by players.
Clue 3: Engine correlation given position
This is the correlation between the top 7 engine moves given winning or losing positions. The numbers are proportions given each situation. Each row adds up to 1. For Player A, in winning positions, they kept an advantage, 56.5% of their moves were top 1 engine moves.
The following is the general percentage of top engine moves each player made.
Clue 4: Mate in X and advantages
Given they had a mate in X (completely winning position), the players made the following choices.
This is how well players kept their advantage or had a losing position in their games.
Clue 5: Time situation when missing mate
The following were the count of Time left when the players missed their mate in X. A Higher y count means they missed checkmate more. Most of them had less than 10 seconds left on their clock.
About the post
If you are curious if your predictions were right, you can check the answer here. I made this post to primarily understand if people could predict rating trends given some visualizations. This post is primarily for feedback and I would love to know which graphs were helpful in your thought process and which graphs weren't. If there are other visualizations you wish were included as a clue, leave them in the comment. The mislabeled clue was to prevent people from focusing on one clue and trying to easily draw conclusions. If you have ideas on how to make this more tricky, that would be lovely as well!
The motivation happened when I was analyzing 100 games between two very high-rated players and found out that these visualization techniques were not enough to distinguish them from each other. I decided to make a problem with 3 players with different ratings and wanted to know if the community could easily distinguish them with these specific standards. Thank you for reading the post and I will be reading the comments.