Accelerated Pairings

[moderator]

This is from a discussion of the USCF forum regarding the use of "doubly accelerated pairings" or accelerating by 1/8's. (Not supported by WinTD, not in the rule book for reasons described below).

Apparently, this is due to some mistaken impression that if dividing into quarters is good, dividing into eighths is better. In a open section at Nationals, the ratings for the bulk of the players are fairly tightly compressed, since most players are above the limit for the highest Uxxx section. Accelerated pairings only work if the winners in the next lower group down fairly reliably lose to (or at least don't beat) the losers above them.

In standard accelerated pairings, roughly 1/2 of all first round winners play higher rated non-winners. The hope is that substantially fewer than 50% of them will win. If you accelerate by eighths, roughly 3/4 of the first round winners play higher rated non-winners and the gaps between those lower winners and their non-winning opponents will generally be much smaller. Since I had it handy, I looked at the K-8 Open section from the 2001 SuperNationals. That had 360 players, and did standard accelerated pairings. The rating gaps on the winner-non winner pairings were mainly in the low 200's. (Perhaps not surprisingly, the lower half winners outperformed expectations - 27% won their games).

If it had been accelerated by 1/8's, the rating gaps would have been as low as 90 (the 4th and 5th octiles are rather closely bunched) and most would be well under 150. The only way that that is going to be more successful at reducing the number of perfect scores is if there are a lot of draws, which isn't very likely in a population of K-8 players mostly rated below 1300.

I would be hard-pressed to think of a situation where accelerating by 1/8's would be useful, which may explain why it's not in the rulebook. Even the 1/6's accelerated isn't likely to be helpful very often.