3 ms·

If I get it right, the claim is that if A is better than B at fast time controls, they tend to be less dominant at slow time controls? If this were true statis

If I get it right, the claim is that if A is better than B at fast time controls, they tend to be less dominant at slow time controls? If this were true statistically, I think that should be reflected in the Elo spreads: Players winning more in blitz should mean they’re able to separate farther from the middle of the rating distribution. This doesn’t seem to happen at the very top end. Looking at the Elo range of ranks 1 to 100 to get a feeling for the distribution: Blitz: 2850 to 2610 Rapid: 2847 to 2620 Standard: 2864 to 2645 It seems if anything, the top players at slow time controls are even more dominant. Of course, top players are outliers by definition. Maybe it’s different in the middle of the rating range? (Source: https://ratings.fide.com/top_lists.phtml https://ratings.fide.com/top_lists.phtml)

usgroup 4y ago
I think this is an essential part of the analysis. The author makes a point of matching pairs of players who have played each other across multiple time controls so that he is comparing the same population of people across multiple time controls. Reason this is important is because the Blitz playing population is not the same as the Rapid or Standard playing population (they are literally a different mix of people and sizes of population), so you can't compare these populations directly against each other.
dmurray 4y ago
I think you would be better off looking at the spread among those top 100 players, rather than between those and some ill-defined "middle". There's no reason to expect the middle of the different Elo distributions to settle in the same place. Looking at the difference between the top-ranked player and the 100th, we get 240 points for blitz, 227 for rapid, 219 for standard. That suggests that among the elite, the best players are indeed more dominant at blitz than at longer time controls. However the effect size is small and you'd want to check for other effects causing this: eg pretty much all of the best players will be on the standard list, but some of them may not be on the blitz list as that is restricted to players who have played a rated blitz game over the board in the last two years.

usgroup 4y ago

I think this is an essential part of the analysis. The author makes a point of matching pairs of players who have played each other across multiple time controls so that he is comparing the same population of people across multiple time controls. Reason this is important is because the Blitz playing population is not the same as the Rapid or Standard playing population (they are literally a different mix of people and sizes of population), so you can't compare these populations directly against each other.

dmurray 4y ago

I think you would be better off looking at the spread among those top 100 players, rather than between those and some ill-defined "middle". There's no reason to expect the middle of the different Elo distributions to settle in the same place. Looking at the difference between the top-ranked player and the 100th, we get 240 points for blitz, 227 for rapid, 219 for standard. That suggests that among the elite, the best players are indeed more dominant at blitz than at longer time controls. However the effect size is small and you'd want to check for other effects causing this: eg pretty much all of the best players will be on the standard list, but some of them may not be on the blitz list as that is restricted to players who have played a rated blitz game over the board in the last two years.