I had forgotten (or hadn’t noticed) that FIDE also discourages repeat floats two rounds ago. It does indeed seem strange, if this happens, that an intervening float of the opposite type doesn’t neutralize one of the others.
Bill Smythe
That is the weakest restriction in the FIDE Dutch Swiss pairing rules (essentially the only version of Swiss pairing rules used). The very lowest “quality criterion” is to minimize the number of players who receive the same upfloat as two rounds prior. The next lowest is to minimize the number of players who receive the same downfloat as two rounds prior. You are correct that an upfloat does not cancel a downfloat.
I recall older TDs making a mark on their pairing cards noting if a player received a downfloat into another score group. It was the unwritten rule that a player was permitted to have only one such downfloat during the tournament. They also tried to avoid downfloats if a player received a bye or took a half point bye for any round. Upfloats were rare and used only when all of the players in the score group were similar in rating.
How do you have a downfloat without the other player having an upfloat?
A downfloat would be choosing a player in the higher point group to be removed from that group and paired with the lower group.
An upfloat would be choosing a player in the lower point group to be paired with the higher group.
An example where an upfloat makes sense:
2200 L 1600
1201 L 1604
1603 W 1200
1199 L 1602
1601 W 1198
You could downfloat the 1600, resulting in:
1604 1601
1602 1603
1600 1201
or you could upfloat the 2200 resulting in:
1604 1601
1600 1603
1602 2200
Aren’t you upfloating 1201 in your first example and downfloating 1602 in the second? Bear with me, I’m not a TD. I thought upfloating and downfloating apply to both players when they are in differing score groups.
I don’t understand Mr. Martinak’s pairings. When I have an odd number in a scoregroup, I don’t add a player from a lower or higher scoregroup and pair normally. Instead I take the lowest (or a very low) player in the top group and pair him with the highest (or a very high) players in the next group and create a heterogeneous group of just those two players, one getting an upfloat and the other getting a downfloat. Perhaps I am completely misunderstanding Mr. Martinak’s post.
I would have instead of his presumed two groups (1 and 0), three groups, one where the players total 2 points, one very small group where the players total 1 point, and the third while the players total 0 points. It is the same reason that I move the top rated player in the top scoregroup off top board if his opponent has a lower score than the opponent of another player in the group. I.e. if there are three players with 5 1/2 going into round 8 and the top rated has played the other two, they’ll play each other on board 1 while he’ll play the top (or a very high) 5 on board two.
Alex Relyea
I’m thinking the same way as you, hence my question as to how there can be a downfloat without an upfloat.
The old Harness odd-man rule (drop the middle) is no longer in the rulebook but would support your pairing labeling the 2200 as an upfloat but it was described as a downfloat of the 1602 (having the 2200 being a second 1602 that was paired against the 1600 because of team or family member issues would still pair the same players - it would not put one 1602 vs the 1604 and the other 1602 vs the 1600).
This is inconsistent at least with how FIDE defines the terms upfloat and downfloat. (US Chess rules don’t use these terms.) An upfloat occurs when a player is paired against an opponent from a higher scoregroup. A downfloat occurs either when a player is paired against an opponent from a lower scoregroup or when the player has an unplayed game for any reason (whether a requested bye, a “paired out” bye, or a no-show opponent).
I’m using pairing terminology (possibly local) from 1970-80s. Basically the question is whether it is better to pick an odd man in the higher group to drop down or an odd man in the lower group to bring up. When the lowest in the higher group had already played the highest in the lower group (usually after a big upset in the 1st round), we would compare the difference between the bottom two in the top group with the difference between the top two in the lower group. In the example, 1 vs 999. So, at that time, we would bring the 2200 up rather than take the 1600 down. The goal being to minimize the number of perfect scores.
But you’re using it wrong. Whenever one player downfloats, another (his opponent) upfloats. It makes no difference whether we use this FIDE terminology. The same concept exists in U.S. Chess, it just has different names (or no names).
The usual method (nowadays) is to downfloat a player at or near the bottom of the upper score group and upfloat a player at or near the top of the lower score group.
Or, if you use the Harkness option, you downfloat a player at or near the middle of the upper group and upfloat a player at or near the top of the lower group.
You could also use reverse Harkness – downfloat a player at or near the bottom of the upper group and upfloat a player at or near the middle of the lower group.
For that matter, you could even use double Harkness – downfloat a player at or near the middle of the upper group, and upfloat a player at or near the middle of the lower group.
While you’re at it, you might consider half-double-Harkness – downfloat a player about 1/4 of the way from the bottom of the upper group, and upfloat a player about 1/4 of the way from the top of the lower group. This might actually be a good option in a large, single-section tournament, such as the U.S. Open. It would, for example, avoid huge rating differences (such as 1000 points) between the floated players paired against each other in round 2.
No matter which of these options (or others) you choose, you should always think of it as a downfloat vs an upfloat. Do not think of the upfloated player as “joining” the upper group, because thinking of it that way suggests that you would simply pair this player “normally” (whatever that might mean) within the upper group.
It is better to think of the downfloat-vs-upfloat pairing as constituting a small, two-player group between the other two groups. For example, if one player has 1 point and the other 0.5, think of these two players as constituting the two-player 0.75-point group.
I agree that comparing rating differences near the bottom of the first group with rating differences near the top of the second group could reasonably be used to decide between (a) downfloating 999 from the upper group and upfloating 2 (instead of 1) from the lower group, versus (b) downfloating 998 (instead of 999) from the upper group and upfloating 1 from the lower group. But, either way, you are still downfloating one player and upfloating another.
Minimizing perfect scores is always one goal when you pair a player near the bottom of the first group vs a player near the top of the second group, regardless of details.
Bill Smythe