FIDE 400 point rule

[MichaelBaron]

A Jeff Sonas said in an article:

800 or so point rule would make sense.

[Ian Rout]

800 or so point rule would make sense.

According to the FIDE rating adjustment table, at a rating difference of over 735 the higher rated player should score 100%, so an 800 point cutoff is equivalent to no cutoff.

The problem with applying the formula to large differences is that it assumes Elo probabilities have the same inevitability as probabilities for rolling dice or tossing coins. In practice scoring over 90%, even against inferior opposition, is difficult in a game that has a high margin of a draw, that can be lost with one mistake, and that gives the weaker player the advantage of the first move half the time.

[Bill Gletsos]

According to the FIDE rating adjustment table, at a rating difference of over 735 the higher rated player should score 100%, so an 800 point cutoff is equivalent to no cutoff.

I deliberately did not point that out in my previous post and wondered how many people would actually know it or both to look it up.

[Patrick Byrom]

... Around October 2005 FIDE changed from calculating the average rating of a player’s opponents in a tournament to doing game by game calculations for the tournament. It is at this time FIDE really should have gotten rid of the 350 point rule altogether as it really serves no purpose. ...

Thanks for this background.

Just to expand on your point, one player with a very low (or high) rating can significantly affect the average opposition rating, and therefore the expected score of a player in a tournament, so it makes sense to have a cut-off to prevent this. However, if ratings are calculated game by game, then the cut-off will only slightly change the expected score for each game, so the effect is minimal.

I completely agree that the useless cut-off should be removed.

[Patrick Byrom]

Just to give an idea of the effects of different systems, let's assume a nine player round-robin, with each player rated 2800 drawing every game, each scoring 4/8. Obviously there would be no change in their ratings.

Now add a single player rated 2000, creating a 10 player round-robin, with each player playing nine games. Assume the 2800 players again draw every game with each other, and all of them win against the 2000 player, each scoring 5/9, or 56%.

Without the cut-off, the average opposition rating for the 2800 players is 2711, and their expected score is 5.6, or 63%, so that without a cut-off, they would lose about 30 rating points, assuming a k-factor of 50 (for simplicity).

With a 400 point cut-off, the average opposition rating for the 2800 players is now 2756, and their expected score is 56%, so that with the cut-off their ratings would effectively not change - again assuming a k-factor of 50.

Finally, with game-by-game ratings, the only game that affects the ratings of the 2800 players is their win against the 2000 player. The expected score here is 99%, so their ratings would again effectively not change.