Is there a tournament system that pairs players based on ratings being similar?

Question

Clearly games are more fun when the result can't be predicted in advance hence pairing players who have not recently played each other and have similar ratings has advantages. After a few games it should be possible to pair based on each player's tournament rating. (It would also be possible to have rules to maximise how often masters are paired with non-masters with similar strengths.)

Maybe alternate between a rating difference that is large enough that a draw is a good result for the lower rated player and a difference that is small enough that a draw is expected.

Such a tournament would not need any limits on the rating of players, but somehow prizes will need to be allocated so people at all ratings have an incentive to play as well as possible.

Answer 1

Swiss tournaments somewhat attempt to do what you want in their later rounds. Also, tournament organizers often have different ratings-based sections for larger Swiss events to reduce the number of games played with large ratings differences. However, I expect you are aware of this and are looking for other systems.

Perhaps the closest is the "Quad", if this fits your idea. It is somewhat common in the United States (at least the mid-Atlantic with which I am familiar), although I don't know how often it's played elsewhere. Each entrant plays three games in what are (seemingly) always one day events.

In a quad the players are split into groups of four, based on ratings: the top four rated players are in one quad, the next four in a second quad and so forth. Each quad then plays a round robin (i.e. everyone plays one game against each of the others in the quad). Typically only the winner of a quad wins prize money; in the case of ties the prize money is split.

When the number of entrants isn't a multiple of four the lowest rated group will have a number of entrants other than four - I've seen groups of three, five, or six used when that happens. A Swiss format is used for a "quad" with more than four players.

Of course, there can still be large rating disparities if there are insufficient numbers of entrants with ratings in close ranges. Unfortunately, that is unavoidable when there aren't many players and there is a large difference in ratings.

Answer 2

I want to give an unequivocal "No" as the answer, but that's too simple. General Swiss pairing rules (for where you want to find a winner from lots of players / teams and few rounds / matches) require you to not pair like that but there can be exceptions.

Let's start by looking at FIDE's Basic rules for Swiss Systems. It's mostly about the principles for a way to do fair pairings. The closest it comes to matching players with similar ratings is:

e In general, players are paired to others with the same score.

So, if players with similar ratings play at about the same level then after a few rounds they are likely to end up on similar scores and end up paired against each other but in general that doesn't happen most of the time kind of by design as we see next.

FIDE's General handling rules for Swiss Tournaments goes into more detail and makes it clear how pairing similar rated players is avoided:

2 Before the first round the players are ranked in order of, respectively

Strength (rating)
FIDE-title (GM-IM- WGM-FM-WIM-CM-WFM-WCM-no title)
alphabetically (unless it has been previously stated that this criterion has been replaced by another one)

3 This ranking is used to determine the pairing numbers; the highest one gets #1 etc.

This is key. When the pairing is done the first step is to split the players into score groups (you play against someone with the same or similar score) then each score group is split in two according to number of blacks and whites to try and ensure fair numbers for the players. Then in each score group and colour the groups are split in two according to their initial ranking and the high white group is paired against the low black group and vice versa. This pretty much guarantees mismatches in the early rounds in big fields.

This is done to ensure fairness in general, to give the best chance of fairly finding the best players according to how they play and to try and make the key pairings (between the players playing best) happen in the later rounds when they have a clearer idea of what they are playing for.

However there is one case in which these rules can be bent towards pairing players of more similar ratings against each other.

Acceleration

If a Swiss tournament has enough strong (IMs, GMs, WGMs, WIMs, FMs, WFMs etc) players from enough different federations then players who do well enough can achieve a title norm (IM, GM, WGM, WIM, etc.) If such a player has to play a weak player in round 1 and maybe also in round 2 then this lowers their average performance rating over the tournament and can decrease the chances of achieving a norm if there are too many weak players.

The solution to this is to use accelerated pairings. The details are spelled out in FIDE's FIDE-approved Accelerated Systems.

The basic method is to split the initial ranking of the players into two groups, the top half rated players and bottom half rated players, and to give all the players in the top group one or two "fictitious" points for the first two or three rounds, pair as normal and then remove the fictitious points after the two or three rounds of "acceleration".

This means that the two groups will end up playing within their group against more similarly rated opponents. Within the two groups they are still ranked and paired according to rating (after score) so this reduces the rating disparity but doesn't eliminate it.

Answer 3

I believe you are looking for the Monrad system. This is a variant of the Swiss system in which, in each round, players are grouped by score and then listed in order of descending rating (or seeding), as in the more familiar type of Swiss tournament (otherwise known as the "Dutch system"). But, instead of pairing the top half against the bottom half, it instead pairs the first player against the second player, third player against the fourth player, and so forth. It does make allowances for modifying pairings to ensure, for instance, that no player faces the same opponent more than once in an event. If you don't care about avoiding repeat matchups, then a further variant of that called the Danish system does exactly that - it pairs players as described above, without regard to whether they've been paired earlier in the event. Either way, this method of pairing makes it more likely that, in each round of an event, players are more likely to face an opponent of similar rating (or similar seeding in the tournament).

One of the big disadvantages to doing it this way (in my opinion) is that, particularly in a larger event, it can put substantially lower rated players in contention for a share of the same prizes being contended for by much higher rated players, without ever having any of those lower rated players face an opponent rated substantially higher than themselves. So (just as an example) you might find that a player rated 1200 could end up in a position where they might be able to split first prize with a player rated 2200, simply because they have the same number of points at the end of the event - even though the 2200 rated player had to work a lot harder than the 1200 rated player in order to get those points. Whether you think this is fair or not kind of depends on your goal - do you want the prizes to be awarded to the players with the strongest performances in the event (taking into account the strength of their opposition), or to you want the prizes awarded to the players who just accumulate the most points in the event (without regard to the strength of the opponents they faced)? Either goal could potentially be valid.

Answer 4

Go tournaments commonly use the McMahon system, which is basically like Swiss, but the players are given different amounts of points (wins) beforehand for free, based on their rank/rating. In a way, it can be considered a Swiss tournament where the first few rounds are only simulated, except that the initial scoring is usually not actually done like that, but directly on ratings.

With winning the tournament being based on the score (wins) gathered during the actual games, and the initial score, this means that it's impossible for players outside the top group to win the tournament. Hence, it should only be used in cases where there skill levels of contestants vary enough that it's possible to identify beforehand some skill level below which a contestant would already be very unlikely to win the whole tournament anyway. Or in other words, the top group should be set so that everyone who just might be capable of winning should be in it.

Like in Swiss, N rounds are needed to find winner from a top group of 2^N players. (And more rounds can bring confusing results, e.g. with a top group of 16, a five-round tournament may end up with two players in the top group finishing with 4 wins each.)