FIDE vs US Chess Pairings

The FIDE Dutch System is designed for stronger players, generally smaller strength ranges. While the US Chess pairing system is a set of rules and guidelines that allow for feasibly pairing even a fairly large tournament by hand, the FIDE Dutch is a description of a computer algorithm which can be very complicated in late rounds.

The two main characteristics of FIDE Dutch are:

  1. Color trumps basically everything other than repeating pairings. Absolute differences in ratings are ignored; order is all that matters.

  2. Predictability (once pairings are posted, they can’t be changed) and non-randomness are crucial. If many players have the same ratings (common in U.S. scholastic tournaments with many unrateds), players are assigned pairing numbers arbitrarily (by default, in alphabetical order) which are maintained throughout the tournament. This is largely irrelevant in “professional” tournaments where relatively few players have identical ratings, and those that do generally aren’t in the same score group in most cases. It’s very relevant when many players have the same rating.

FIDE also has its share of goofy rules

Consider C.8 and C.9 (they are C10 and C11 in the 2025 Rules)

C.8 minimize the number of topscorers or topscorers’ opponents who get a colour difference higher than +2 or lower than -2.

C.9 minimize the number of topscorers or topscorers’ opponents who get the same colour three times in a row.

As background, FIDE goes a bit farther than US Chess rules in dealing with “three in a row”. In US Chess rules, you don’t break a score group to avoid three colors in a row. In FIDE Dutch rules, you do if neither player is a “topscorer” (above 50% going into the last round), that is, if you have two and only two 0.0’s and both have an absolute preference for the same color, you break the 0.5 score group to provide two other opponents so as not to pair two players with the same absolute preference. FIDE Dutch also treats players who are already +2 or worse in color identically to players who face three in a row in the same color—except in the case of the topscorers, where the C.8 vs C.9 come into play. Note that this only matters if: this is the last round AND you are pairing players above 50% AND you have at least three players who are way out of whack in color in the same direction, AND there is no way to avoid pairing two of them AND there is a choice between pairing a player worse than +2 with both a player facing three in a row (but not worse than +2) or another facing worse than +2 AND the more “natural” pairing has the worse than +2 paired with each other. In a first go, I figured I could get away with basically combining C.8 and C.9. It took until I was at 50,000 simulated tournaments with around 500,000 rounds before I hit a case where the difference between those mattered.’

The deterministic nature of FIDE pairing rules is greatly overstated. I defy any one to take the listed rules and figure out THE ONE AND ONLY CORRECT PAIRING for dealing with the last sentence in A.9:
“The pairing process resumes with the re-pairing of the PPB [Penultimate Pairing Bracket]. Its downfloaters, together with the players of the collapsed scoregroup, constitute the Collapsed Last Bracket (CLB), the pairing of which will complete the round-pairing.”

Unless the intention was that “the pairing of which will complete the round-pairing” for the CLB was to be pretty much “anything that works” (possible, as this is generally just the bottom few score groups), there are at least three reasonable ways to handle that, and the method chosen by SwissPerfect wasn’t on my list of three.

(Note, the 2025 rule change basically confirmed that “anything that works” was actually what was intended; in fact, they repealed the entire description for backtracking through the pairings to fix the bottom score groups to be replaced by

1.9.3 If it is impossible to complete a round-pairing, the arbiter shall decide what to do.

That being said, what can we learn:

  1. US Chess doesn’t really deal with repeated “downfloats” and “upfloats”. While I think a two-round lookback as is done by FIDE Dutch is overkill (particularly because you could have an upfloat and a downfloat in the previous rounds, which instead of washing means you shouldn’t repeat either), a one round lookback seems like a good idea (at least as an option).
  2. I think US Chess deals better with color assignments when there are missed rounds. FIDE treats WBxWB the same as WBWBx—if two players have the same W-B difference, you work back through the rounds IGNORING UNPLAYED ROUNDS until you find a difference. US Chess works back until there is a difference, period. Aside from being simpler, the US Chess rule is based upon the idea that color matters more as you get later in the tournament: WBxWB has a greater claim to White because she had B in round five, while WBWBx had B in round four.
  3. US Chess rules for evaluating transpositions and interchanges are based upon those being used to correct “color” problems. Of course, there are many reasons that A shouldn’t play B (probably more in US chess-land than FIDE—we are more likely to have pairing preferences involving teams, clubs and states). The FIDE Dutch rules don’t care WHY A can’t play B; it just has a deterministic set of rules for choosing which alternative set of pairings is to be used. (Any set of transpositions is better than any set of interchanges; a single interchange is better than any double interchange, transpositions are examined in lexicographic order aiming to keep high boards as intact as possible, …). The 80 and 200 point limits are probably reasonable for tournaments run on Elo ratings with the standard chess scales for a “low” (<80 point difference is statistically insignificant for class players) and “high” (>200 point difference really isn’t insignificant at almost any level except very, very low ratings). The US Chess rules are basically that you don’t do a >80 point swap to fix alternation errors, >200 point swap to fix equalization errors but there is no limit for pretty much anything else. A problem with employing that in practice is that in late rounds, there can be pairings which can’t be done because the two have already played; some pairings where equalization is a problem and some where alternation is, and you probably can’t fix everything with a few simple switches. So how much of each switch is to fix what? The Rulebook is silent on that, and it’s not clear that there is any good answer to it. Just hope that you never have to stray more than 80 points.

For more on this topic, see FIDE vs. US Chess Swiss Pairing Rules