gregshahade.wordpress.com/2017/ … d-explode/
Thoughts?
See Dubov system.
Random pairings are ridiculous. Every method has its problems. With random pairings you simply don’t know which problems you are going to have. Any potential problem thus becomes possible.
More promising is the link provided by wintdoan. Within each score group, the players due black are arranged in order by rating (highest at the top, lowest at the bottom) as usual. But the players due white are arranged in reverse order of the average rating of their opponents so far. Then, those due white are paired against those due black, one by one down the lists thus arranged. This means that, within each score group, players due white who have had stronger opponents are given easier pairings than those due white who have had weaker opponents. This tends to equalize the toughness of the field each player in the score group has played.
Since players are due white about as often as they are due black, each player receives the benefit of this “toughness adjustment” about half the time.
This system is still a “standard” Swiss, in that it pairs players with the same score whenever possible. So it will do nothing to reduce the likely number of eventual perfect scores.
If you want fewer perfect scores at the end, you need a system that pairs players with different scores against each other in such a way that the lower-scoring player is likely to win. That’s the idea behind accelerated pairings, where in round 2, top-half losers play bottom-half winners.
It’s also the idea behind the Verber system, which amounts to accelerated pairings on steroids.
Bill Smythe
I’m not sure I agree with you that accelerated pairings pair people with different scores against each other, to me it seems like it still pairs people with the same score, it just takes some of the players in a score group out of the pool of possible opponents for some other players in the same score group.
Accelerated pairings work best (if that’s their intent) for tournaments with large turnouts compared to the number of rounds.
Otherwise, it really comes down to when you want the strongest players in the tournament facing each other, early in the event when they’re likely to be playing to win, or late in the event when they may be more likely to play for a draw. 1-2 pairings might be more likely to achieve the former.
The version of accelerated pairings described in the USCF rulebook definitely does pair players with different scores, in round 2:
Round 1: Top quarter plays second quarter, third quarter plays fourth quarter.
Round 2: Top half winners play each other, top half losers play bottom half winners, bottom half losers play each other.
Round 3, etc: Revert to standard algorithm where players with same scores play each other, half vs half.
The idea in round 2 is that the top half losers should defeat the bottom half winners, so that all will be 1-1 after 2 rounds. Half of the top half winners will also be 1-1, as well as half of the bottom half losers. Thus, instead of having 1/4 of the field at 2-0, you will have only 1/8. Ditto 0-2. The remaining 3/4 (instead of 1/2) will be at 1-1.
No system that always pairs players with the same score will get rid of perfect scores faster than any other. It doesn’t matter if it’s half-vs-half, 1-vs-2, or random. (Well, there might be a slight reduction in perfect scores when high-rated players play each other, due to a slight increase in the number of draws.)
The only way to get rid of perfect scores quickly is to pair players with different scores, in a situation where the lower-scoring is likely to defeat the higher-scoring. This means pair higher-rated, lower-scoring players vs lower-rated, higher-scoring players.
Accelerated pairings accomplish that goal to some extent. Verber pairings (the one time it was ever tried) accomplished it on steroids.
Bill Smythe
In the only large tournament I ever played in with accelerated pairings, I went 2-0 and wound up playing the defending Nebraska champion in round 3 (about an 800 point ratings difference) and I was crushed like a grape. It was not a pleasant experience.
I played in several events with accelerated pairings back in the 70s when single section events were prevalent. As a C player I invariably beat a low rated player in round 1 instead of suffering my expected round one defeat at the hands of an A player. In round 2, I invariably faced an A player who had just been trashed by an expert or master. Said A player was usually tired, and I beat most of them. I doubt I’d have had the same results if I’d faced them in round 1 when they were fresh, but that’s hard to say. For me, then, the rating of my opponents pretty much was the same as I’d have had with normal pairings in reverse order, but my results weren’t the same. And the poor TD had one more perfect score than he should have had. Of course in round three I always got the squashed grape treatment.
How did I stay in class C so long when I was beating up on all those tired A players? Well, I lost to a lot of D players during that same time period. Evidently I played to the level of my competition!
I made it to Class B, briefly, but that was as high as I got.
I like d10 more than I like random pairings.
The suggestion seems to have been to use random pairings only at the start:
“Computers are pretty smart these days. It shouldn’t be difficult to set up a tournament so the first round or two is random, and then the computer could go out of it’s way to “equalize” the strength of opponents that people within a score group have faced throughout the tournament.”
The Dubov pairings that I linked to do standard Swiss in the first round and then start the process of trying to equalize the strengths of schedules, but there’s no strong reason to use standard Swiss rather than random pairings for round one. Dubov has a whole host of other problems, particularly as it would apply in typical US tournaments:
- Competition average isn’t a particularly good measure when the players are greatly spread out (and Dubov never even mentions how to handle unrated players). For a 1500 player, playing a 2100 vs playing a 2500 aren’t all that different in winning expectancy, but if you got paired with a 2500 you would get likely get much easier pairings for the rest of the tournament.
- For feasibility, the players are divided into which color they should receive and then only the players who are due White have their CA’s adjusted. (The CA’s for the players due Black are irrelevant). In a four round tournament, that means that players who have White in round one will only be looked at for equalizing strength of competition in round 3. (If you try to somehow push all the players’ CA’s towards each other at once, you will discover that there are too many moving parts to do it by hand).
- Dubov is likely just going to fail miserably if there are too many players who are blocked from playing each other by pairing preferences (team, state, etc.).
So you really need a long enough tournament that each player will see multiple adjustments to his/her CA and you need a field where the rating gaps are small enough that the winning expectancy is roughly linear and where players are generally free to play anyone that have not yet met. How many of those are there in the US in a typical year: maybe the major CCA opens?
I know that a local club runs one. I was considering running one, but now I see two many issues. What if the only legal pairing is a 3-0 vs 0-3 ??
It’s only ‘legal’ in the sense that in a Swiss players should not play each other more than once.
Otherwise, there are no restrictions against players having multiple games against the same opponent.
Oh, But if you call it a swiss won’t the system reject it ??
No, we don’t check for ‘valid’ pairings, because as far as the ratings system itself is concerned, any pairing is valid. (We do have limits on matches, but those are to protect the system from possible manipulation.)
About the only reason I ever came up with for our asking if the event used ‘swiss’ or ‘round robin’ pairings is the fact that RR pairings can be entered in classical RR form, where the columns represent the opponents rather than the rounds the games were played in.