Lowest-vs-highest does indeed seem to be the most sure-fire way to kill two birds with one stone (or rather, one bird with no stones). I guess some feel it’s overkill, though.
I think Harkness (middle-vs-highest) may actually pre-date lowest-vs-highest. Some older rulebooks are not so clear. I can’t figure out what Harkness may have been thinking.
Bill Smythe
He may have been trying to reduce the benefit of the “Swiss Gambit”.
When I did it automatically (in the early '80s I was doing pairings before I bought the rulebook and initially thought they were so straightforward that I didn’t need to buy the rulebook, and then later discovered that I had been using a minor variation rather than the main rule) I did it to avoid the gross mismatches of bottom vs top. It seemed more consistent to have (in the example given and assuming 50-point differences between each adjacent pair of players) 5 games with 200-point differences and 4 with 150-point differences instead of 8 games with 150-point differences and 1 with a 50-point difference).
I didn’t think of it as dropping the middle, but rather as pulling a player up to become the bottom of the above score group.
Absolutely. The consistent rating spread is a secondary consideration, less significant than avoiding repeat play and less significant than staying within a scoregroup.
I think the key feature of a Swiss system is that it tries to maintain opponents at equal score levels. If you ahve enough rounds, the perfect score problem will solve itself. There’s no need to do that quickly, unless you have too few rounds to be certain they will be eliminated. Accelerated pairings are aimed at that problem.
The many variations of the Swiss system, not just the variations found in Chess, are all variations on trying to find what is best to do with the fact that there are odd number of players.
I’ve advocated that, when it is necessary to give a full point bye, it is better to give it to the player in the middle of the lowest score group rather than the bottom – especially in the early rounds. I’ve justified this by the thought experiment of introducing a player at the bottom of the wall chart who plays so very poorly that he is guaranteed to lose every single game and using that player to make the field an even number. In that case, you would be giving the “free point” to the player who is just above the halfway cut in the next score group up, rather than to the bottom player. Jeff’s comment makes me realize this is essentially giving the “float down” for the bye to the middle player of the group – in other words, the Harkness system.
When pairing a score group with an odd number of players, if you consider the players in the next score group down to be “weaker” than those in the current score group (not based on rating as much as on performance in the current tournament), then there is reasonable logic to the Harkness system. By assigning the float to the player in the middle of the score group, you’re basically pulling the top player in the next score group up, and the player just above the cut then gets a (slightly) weaker opponent than the next player up in the top half.
Every time somebody tries to justify the Harkness variation, the logic is always along lines similar to the above.
That logic works great for the player being downfloated from the middle of the upper group, but for some strange reason, the advocates of Harkness are conveniently silent about the consequences for the player upfloated from the lower group. That player, at the top of his score group and expecting a lower-rated opponent, is instead given an oppoinent who is not only (probably) higher-rated than himself, but higher-scoring as well. It’s totally contrary to the spirit of half-vs-half pairings.
The Harkness variation is an abomination from the distant past. It’s way past time to give it the boot.
Bill Smythe
I think it depends on the sort of tournament you are running. If you are running a class championship with a large prize, then all the players are reasonably close together in ability, and a bye could influence who gets a significant amount of prize money. In that case, it makes sense to give the unearned advantage to the player least likely to benefit from that advantage. The overriding concern should be trying to make sure that the prizes are allocated to the best performers of the day.
I, on the other hand, run Saturday afternoon, church basement, trophy tourneys. In that case, giving the bye to the lowest player deprives that player, usually a child, of a chance to play one of his games. Plus, when he goes into round two, he’s playing more difficult opponents than he really ought to be playing. For that reason, I’m pretty sure I’m going to use the method you have advocated in the future.
I did mention that it lessened the primary plus to a “Swiss Gambit”. Often that top player in the lower score group is significantly HIGHER rated than the bottom player in the upper score group, and also higher rated than the MIDDLE player in the upper score group.
As previously mentioned, Harkness is more applicable to an open tournament than a class tournament. I’ve seen multiple cases in open tournaments where an IM or GM that either drew or took a half-point round one bye ends up paired in round two against an E or below player that had a round one full point bye. Harkness avoids that and gives the IM/GM a bit more of a game.
I would not go anywhere near as far as calling Harkness an “abomination”, and I feel that it is worthwhile when used in the proper type of tournament.
It is certainly laudable not to give such a player a bye if possible. However, with lower-rated players, especially children, it is almost always possible to use cross-round pairings to avoid byes.
See 28M, Alternatives to byes, especially 28M1 and 28M2. You would be doing your players a great favor.
Bill Smythe
After some years of using cross-round pairings, I came to the conclusion that such games would often end up delaying the tournament (especially with an ASAP schedule, but often with a fixed schedule as well). That may have been mitigated by shortening the time control, but the K-3 section is already G/30 and going with G/20 is unpalatable to me for a dual-rated section (and in the HS section G/60 would sometimes need to be reduced to G/20 to avoid delays - I used to just tell the player after a while that there wasn’t time for another game and he was stuck with his bye).
I have a significant preference for cross-sectional pairings in an extra games section (with the players getting the scheduled byes in their section) and house games (generally in an extra games section) especially in a tournament like this one ( main.uschess.org/assets/msa_joom … 1-13008826 ). With a house game/extra game you have more flexibility than with a cross-sectional pairing and can thus often get an earlier start (especially in the later rounds when players in games on the bottom boards have a higher chance of being more social, inattentive or slow - it’s amazing how often it is one of the bottom boards that is the last game going in a section).
Of course the problem with this is, that in Mr. Lame’s case, his tournaments tend to be one section.
Alex Relyea
Even a one-section tournament can have an extra games section added.
Yes, and I’ve done this. I misspoke. What I meant to say was that cross-sectional pairings would be impossible in Mr. Lame’s tournaments.
Alex Relyea
But you could accomplish the same effect with a less extreme, more balanced approach. In looking for a downfloat from group A and an upfloat from group B, instead of jumping all the way up to the middle of group A while staying at the top of group B, why not go partway up group A and partway down group B? There is no reason to consider only the extremes (middle player or bottom player) while rejecting everything in between (anybody in the bottom half).
Let’s say, for example, that group A has nine players, A1-A9, and group B ditto, B1-B9. Instead of one extreme (A9 vs B1) or the other (A5 vs B1), why not something like A7 vs B3? The exact numbers could be determined case by case. The goals would be to (a) avoid blowout pairings, but still (b) give each player a pairing along the lines he is reasonably expecting. Harkness often fails miserably on point (b), as it can give player B1 an unexpectedly strong opponent.
This asymmetric thinking overlooks the viewpoint of one of the two players (B1 in the above example).
I prefer to think of a cross-group pairing as constituting a separate, tiny score group in between the other two groups. For example, when a 2.0 is paired against a 1.5, this pairing can be thought of as the 1.75 group.
Bill Smythe
For the most part, I deal with byes by playing when there are an odd number of players. It’s an option available with small, low stakes, tournaments.
I hope they grow into large, low stakes, tournaments, and then I might want to do something else, but for now, it works most of the time. I’ve actually only ever had one bye game, when there were an odd number in both the rated and the unrated sections,and that problem was solved when people showed up late and joined in round 2, evening out the sections. I will look into the rulebook suggestions, though.
It occurred to me that my system might make tournaments more susceptible to the “Swiss Gambit”. By pushing the climactic games toward the end, including for the players near the top, not just the top two, it might create a situation where everyone can look at the looming end of the tournament and realize that a draw just might do both players some good. It’s something I’ll have to be wary of.
Of course, for my tournaments, specifically, it won’t be a problem. No one is going to conspire to create a round 5 draw in order to get the coveted “Chess” bronze medal from Crown Awards.
I agree, that’s the best option in small tournaments.
The only question is, if you are jumping into the tournament in a later round after not playing in the early rounds, what score do you give yourself, for pairing purposes, for the missed games? The objective should be fairness to the player you pair yourself against. I think the best answer is to give yourself a score which will put you approximately in the middle (rating-wise) of the score group you put yourself in.
It’s really hard to be sure whether pushing the climactic games toward the end would, in any given case, increase or decrease the effect of the Swiss Gambit.
The real problem with delaying the climax is that the earlier games may not turn out the way you had hoped.
Suppose, for example, that in a 4-round tournament after 2 rounds, the standings are:
2.0 2100
2.0 2050
1.5 1950
1.5 1900
(plus other players with 1.0 or less).
If, in round 3, you postpone the 2100-2050 pairing, there is no guarantee that the 2100 would defeat the 1950, nor the 2050 the 1900. Statistically, a 150-point rating difference corresponds to about 69:31 odds for the higher-rated player. Throw in the possibility of a draw, and the odds could be more like 62:10:28 (win:draw:loss). With only a 62% chance in each game, the chances of both higher-rated players winning is only 62% of 62%, or about 39%. I don’t like the odds of 2100-vs-2050 even turning out to be an appropriate pairing in round 4. By not pairing them in round 3, you may have missed your chance to pair them at all.
Bill Smythe
Something you might want to consider is that a house player isn’t required to be a USCF member. That is you could, if you had an odd number in both sections, create an extra rated games section for the two byes, even if the unrated player isn’t a member of the USCF. Another possibility, of course, is to make that game unrated.
Alex Relyea
I wouldn’t. The desire to push the decisive games out is outweighed by the desire to pair equally scored players. One of the top two guys would be scored 3.0 at the end of round 3, and in round 4 he would play the highest scoring player he had not previously played.
With a round three board one draw you could have the 2100 playing the winner of the 1950-1900 game while the 2050 plays some 2-1 player (color issues could change that, such as if the 2100 went WBW, the 2050 WBB, the 2.5-0.5 A-player also WBW and the top 2-1 BWB).
OK, now I see your point, at least partially.
Since you want to maximize rating differences in pairings, you would, in a score group with an odd number of players, pull out the middle player, so as to drive a wedge between the top and bottom halves of that score group. This would increase (slightly) the rating differences in pairings within that group. This effect was especially noticeable in your 6-player example.
Presumably, you would also pull out the middle player in the next lower score group, for the same reason. Your cross-group pairing would then consist of pairing these two middle players against each other.
This gives rise to four possible ways of handling odd players:
I threw in the third one for symmetry. As far as I know, no TD has ever actually used it.
You invented the fourth one. (At least, I think you did.)
Features of each:
Although I continue to dislike Harkness (and inverted Harkness), I must confess some fondness for the double Harkness, at least for small tournaments where you want to decelerate the pairings as much as possible. Double Harkness seems to have every advantage over regular Harkness.
Bill Smythe