This is a part of a dataset that can be found here and it contains information about the most recent 20,000 games
(at its time) played by the top 100 teams on Lichess.
It includes information about each game like; rating of the players,
opening codes, game rating, number of turns, and who wins each game and how he/she wins it.
I found that both difference in rating between players and number of turns in a game play important role in deciding the winner and the final game result which yielded, in my exploration, to several findings:
- Most of the games are played between players of the same or close rating and lasted until about 40 to 70 turns.
- Players, as white or as black, are winning more when their ratings are higher than their opponents compared to when they have less rating than their opponents.
- Games that are played with high difference in rating in the winner's favor tend to end in less number of turns than those with lower difference in rating.
- There are more players as black winning the game with high number of turns specially when the rating difference is small.
As for the other variables interaction with the variables of interest; winner and victory status, I found the following:
- Player as white has more advantage over player as black in blitz rating games.
- Highest winning percentage for white is at opening group C while, surprisingly, black has slight advantage when using opening group E.
- Despite the fact that openings of group E are rarely used by the winner, they are popular among players of high ratings.
- For all opening groups, it takes higher number of turns to beat an opponent by checkmate than it takes for a win by opponent resigning.
The focus in my presentation will be on the obvious variations with rating difference and number of turns and I will leave out some of the subtle variations.
I will start with a scatterplot showing the variation of winners, as white and as black, with rating difference and number of turns.
Afterwards, I will show how variables like game rating and opening codes have a slight influence on who and how the game is won.
I will use bar plots for the distribution of winning percentages on game ratings and opening codes.
Then, I will show a scatterplot for the distribution of game ratings that are played with different rating difference between players and ended after different number of turns.
Finally, I will show a point plot of the openings that were successfully used by the winners with different victory results and their variation with number of turns.