3

As people tend to like playing games when the outcome is not known before they start I would expect most games to be between players when the rating difference is not much more than 200. Is this the case for FIDE rated games?

Related questions:

Query for games where the difference of player ratings was big

2 Answers 2

5

I took the 2022 olympic games.

enter image description here

And this is its distribution.

enter image description here

A rating difference in the range 80 to 99 got a count of 213 games.


2018 Chess Olympiad

enter image description here


2016 Olympiad

enter image description here


2014 Olympiad

enter image description here


2023 Grand Swiss Open

enter image description here


2021 Grand Swiss Open

enter image description here


2022 World Juniors Open

enter image description here


The 4 olympiads combined.

enter image description here

Olympiad is fun and exciting, high mean and more spread, with some upsets expectations. More fun for a lower rated players. Are they going to change the first round pairing?

3
  • You have excluded most but the top players by your choose of events and chosen large events. Is this sampling topical of all FIDE gated games? Commented Jun 3 at 7:34
  • The issue here is to find the fide rated events or games. Can you supply the games?
    – ferdy
    Commented Jun 3 at 7:53
  • The olympic games are more representative. Strong and far weaker players often meet in the early rounds of the tour.
    – ferdy
    Commented Jun 3 at 9:02
4

Rating differences

For bigger sample size I took games from TWIC issues 1469-1520 (covering most of year 2023). After filtering for rated games played in year 2023, this left in sample 355,701 games.

Biggest rating difference was 1,561 Elo points, which is not surprising, since it occured in first round of Swiss system tournament. Swiss system pairs in first round top half of the players with bottom half, so large rating differences are expected.

Mean was 205 Elo points.

If we group rating differences on interval of 100 Elo points, we get:

Difference between Number of games Percentage of all games
0 - 100 103619 29.87
100 - 200 94731 27.31
200 - 300 68633 19.78
300 - 400 40737 11.74
400 - 500 19825 5.71
500 - 600 9424 2.71
600 - 700 4667 1.34
700 - 800 2261 0.65
800 - 900 1166 0.33
900 - 1000 582 0.16
1000 - 1100 319 0.09
1100 - 1200 183 0.05
1200 - 1300 69 0.01
1300 - 1400 54 0.01
1400 - 1500 20 0.005

So full 42.82% of games had rating difference bigger than 200.

Is outcome really known?

FIDE ratings aren't best at predicting results, since they change rather slowly compared to improvement rate of (at least some players). For most rating differences score achieved by player with smaller rating was larger than predicted just by rating difference:

Difference between Achieved score (%) Expected score (%) between
0 - 100 44.44 35.99 - 50.00
100 - 200 35.76 24.03 - 35.99
200 - 300 28.76 15.10 - 24.03
300 - 400 23.66 9.09 - 15.10
400 - 500 19.64 5.32 - 9.09
500 - 600 17.77 3.07 - 5.32
600 - 700 13.98 1.75 - 3.07
700 - 800 13.55 0.99 - 1.75
800 - 900 11.87 0 - 0.99
900 - 1000 8.50 0
1000 - 1100 8.15 0
1100 - 1200 4.91 0
1200 - 1300 7.97 0
1300 - 1400 7.40 0
1400 - 1500 0 0

Impact of tournament systems

Players in general don't choose their opponents - system of tournament in which they play determines the pairings.

Most open tournaments are run either through Swiss or Accelerated Swiss systems - difference between them being, that accelerated swiss system "skips" first two rounds of swiss system, so its first round of pairs is same as it would be in third round of swiss, if all results were as predicted by ratings.

In first round of swiss system top half of players by rating are matched against lower half of players - so if we have tournament with 100 participants, strongest player will play 50th strongest, second strongest will play 51th strongest, etc. This means that first round in swiss system will have very large rating differences.

In second round of swiss winners play winners (sorted by rating as in first round) and lossers play lossers, so rating differences still persist, albeit they are a lot smaller than in first round.

Additionally this way of pairing means that, if we have in bottom half of players a very underrated (or just lucky) player, he will keep playing against much stronger rated players as long as he keeps winning (or at least drawing).

3
  • Tournaments often have different sessions, with a max rating for each session (other then top), and a higher charge to "play up" in a session that is significantly above a players rating. Commented Jun 3 at 11:36
  • The swiss systems does tend to create pairing when one of the players is playing significantly above their rating. I expect the prediction value of players ratings would be much better with random pairings. Commented Jun 3 at 11:38
  • @IanRingrose multiple sections of course partition players a bit more - but sections are often pretty large. At least here in balkans it's common to have "only" two sections (commonly one above 2000, 1800 or 1600 and one under), which still leaves quite a large rating range inside section. Prediction value of ratings would be better, if players played more games - no matter the pairing system. However with OTB this is not really possible for majority of players - this is why online ratings are much better predictors, since they are based on larger number of games. Commented Jun 3 at 11:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.