The Swiss System is a non-elimination style of tournament which can handle reasonably a fairly large number of players. The main rules of the Swiss System are:
 

1.No player is eliminated

2.No pairings should be repeated from earlier rounds

3.As much as possible, players are paired against opponents with the same score. In particular, in round two, round one winners typically play other winners, losers play losers and draws play draws.

4.Players generally end up with roughly the same number of games playing white and black.
 

Natural Pairings

In each round, the players are ranked first according to score and then according to their rating (or other assigned rank). If there are an even number of players in a score group, the natural pairings have the players divided in half, with the top player in the top half playing the top player in the bottom half, the second player in the top playing the second in the bottom, etc. For instance, with eight players, the natural pairings are

 

#1 vs #5

#2 vs #6

#3 vs #7

#4 vs #8

 

In pairings after round one, these natural pairings may have flaws. For instance, it's possible that the two players have previously played (which is especially common in score groups that include draws). It's possible for the two players to be restricted from playing by the pairing rules of a particular tournament, if, for instance, they are teammates. And it's very common for the two players to be due to play the same color. Usually, there is a minor change to the pairings that corrects this. For instance, if #2 and #6 have played earlier, either

 

#1 vs #6

#2 vs #5

#3 vs #7

#4 vs #8

 

or

 

#1 vs #5

#2 vs #7

#3 vs #6

#4 vs #8

will be similar pairings that avoid that problem. Both still have the top four playing the bottom four, both keep two of the four natural pairings and require just one swap of players in the bottom: in the first case switching #5 and #6, in the second #6 and #7.

 

Regarding color problems, there are two possible types of problem pairings. The more serious is an equalization problem, where both players already have more games with the same color (for instance, two players with BWB), so if they are paired, one will end up with now three Black to one White. The less serious problem is where the two players are due the same color on alternation, which means that they currently have the same number of Blacks and White, but have both just played the same color, for instance, BWWB vs WBWB. Note that it may be impossible to fix all colors in a score group. For instance, if in the eight player example, there are five players due White and three due Black, there is no way to avoid at least one pair with both due White. Usually, there will be a way to get as many colors fixed as possible (so that in this case, there is only one bad color pairing), though sometimes that may require a fairly substantial change from natural pairings.

 

There are two main systems of handling this that are available in WinTD, which differ in how aggressive they are in dealing with color problems and how they evaluate different pairing changes. You choose these using the section Pairing Rule Set.

 

US Chess methods

US Chess methods are designed to work in a broader range of situations, where the ratings of the players could be spread over a wide range and where there could be "pairing preferences" (such as restrictions against pairing teammates, or players from the same state) which could make it more difficult to find proper matchups. The natural pairings are compared against a proposed alternative set of pairings and might be rejected if the rating differences produced are greater than some preset limit. This is controlled by two values: an equalization or "high" limit (200 by default) and an alternation or "low" limit (80 by default). How these are evaluated is fairly complicated (see Pairing System: Color Correction), but the general idea is that larger rating changes are permitted to avoid the more serious equalization errors than to fix the less serious alternation errors. If the ratings in a score group are fairly widely spread, it may be correct under this method to leave some color problems rather than distort the pairings too much.

 

Note that under US Chess methods there can be more than one set of valid pairings. How WinTD chooses one is described in Evaluating Pairings.

 

FIDE Dutch method

The FIDE Dutch system differs in three major ways:

 

1.It is more aggressive in correcting colors—it will do anything it can within a score group to correct the maximum number of colors.

2.The ratings of players do not affect pairings other than for the original rankings.

3.There is (theoretically) only one set of pairings which is proper.

It is designed for tournaments with very strong players where color is considered to be a much more important than it is with lower-rated players. The algorithmic nature of the pairings ensures that there are no "judgment" calls by the tournament director that might be used for favor some players over others.

 

A Comparison of the Two Methods

To show ways in which the two approaches differ, let's assume what we have an eight player score group, and we're pairing the second round. In the first case, we have the following natural pairings, which show the players, their ratings and the color they had in the first round. We'll assume that any pairing is still OK.
 

 

Five players are due Black, and three are due White, so we will have to end up with one pair where both are due Black. Under FIDE pairings, transpositions are looked at systematically, where pairings which leave higher boards intact are chosen if possible. So the only correct pairings under the FIDE Dutch rules are 1 v 5, 2 v 7, 3 v 6, 4 v 8. Note that these pairings are the same, whether the players are spread across 500 points (as here), or whether they are much more tightly grouped. Under US Chess pairings, both possible switches are looked at: 1 v 6 and 2 v 5 or 2 v 7 and 3 v 6. The first is considered to be a 3 point switch (as it can be accomplished to swapping #1 and #2, so 2010 for 2007), and the second is 110 (#6 with #7, so 1810 for 1700) so the former is preferred.

Copyright © 2026 Thomas Doan