ECF Swiss Pairing Rules

[Stewart Reuben]

My book addresses what I regard as a considerable defect of the Seeded Swiss Pairing System, which system is however excellent for determining the winner of an event.
Page 62 of the current edition explains the problems. Consider a 64 player tournament and the opponents of players 32 and 33 assuming all games are won and all go according to seeding. After 4 rounds, they will each have 2/4. 32 meets 64, 15, 48 and 19, the sum of which is 146. 33 meets 1, 50, 17 and 46. Now the sum is 114. 33 has met easier opposition. The effect is even more marked if all the games are drawn by all the players in the first three rounds. Then 32 plays 64, 62, 60 and 58. 33 plays 1, 3, 5 and 7. I have twice observed this effect in the British championship.
Players dislike the 'bouncing' effect. In 2001 in Bermuda I played a 5 round Swiss. I won rounds 1, 3, 5 with white against players I was expected to score virtually 100%. Rounds 2 and 4 I lost to a GM and an IM with black.

Most people nowadays think that quality Swiss Pairings consist of using a good program competently. They do not consider the underlying statistical effects of a system. Indeed they are not even aware of the options.

Of course I favour Accelerated Franctional Pairings as I devised the system - accidentally. The bonus point system has the disadvantage that, in the third round, players with 2/2 commonly meet weaker opponents than those with 1.5/2. This problem will be observed in the Olympiad in Dresden.

I am simply ignorant of the French Accelerated System. Both David Welch and I would welcome information.

The Burstein and Dubov Systems also have considerable merit.

Stewart Reuben

[Edward Tandi]

I like the suggestion of recording manual deviations. This could be applied to the paper system too. Could we mandate it in the rules?

To what end?

For better visibility. Once it becomes a requirement to record and publish this information, I think the people performing pairings will be more careful how they apply some of the looser options. It might be an idea to have this displayed at the tournament too, so that the players can scrutinise it.

To whom are you going to send the full draw, and the deviations (properly recorded)? What are they going to do with them when they get them?

Well, results are generally sent to the ECF and FIDE. I would expect them to take a cursory view over the deviations to check that the deviations are reasonable. There shouldn't be may cases to examine.

What problem is going to be fixed by this additional work? And who's going to pay for it?

I think having to justify pairing alterations will reduce the amount of arbitrary tweaking performed. Sure, it's a little more effort for those doing the pairing and those accepting results for grading.

If a deviation is found to be questionable (unlikely), the matter can be taken up with the arbiter. I would not expect players to be penalised for pairing mistakes.

[Edward Tandi]

I am simply ignorant of the French Accelerated System. Both David Welch and I would welcome information.

The Burstein and Dubov Systems also have considerable merit.

Having read your post, I'm not sure I can call it French, it's just the first place it's been observed in real use. Details:

SWISS SYSTEM WITH PROGESSIVE ACCELERATION

OBJECTIVE - This variant of the Swiss system aims to better match opponents of approximately equal strength during an open tournament. The goal is to allow international norms during an open tournament where less than of the players have FIDE rating.

PRINCIPLES – When pairing at the start of the round, each player has a global score (GS), which is equal to the real score gained by play (RS) plus a fictitious score (FS) between 0 and 2:
GS = RS + FS
Rather than relying solely on fixed full-point fictitious offsets to separate the groups, this system awards additional half-points based on achievement during the tournament. This allows the groups to merge together more gradually.

CALCULATION OF THE FICTITIOUS SCORE
Players are divided in 3 groups A, B, C on the basis of their Elo rating. Each group includes a minimum of 25% and a maximum of 50% of all players. Elo rating is used to limit the groups, which are decided by the arbiter before the start of first round. The groups may be set according to the prizes determined by the organizers. In general, group A includes players having an Elo rating greater or equal to 2000 points and group C includes players with an Elo rating of less than 1600 points.

Like the standard accelerated system, at the start of the tournament, players in group A have two fictitious points (FS = 2), group B one fictitious point (FS = 1) and group C no fictitious points (FS = 0).

When a player of either group B or C gains at least 1.5 real points (RS ≥ 1.5), their fictitious score is increased a further half-point. When a player gains their third real point (RS ≥ 3), their fictitious score is increased another half-point. Thus group B players can eventually attain the maximum of number of fictitious points (FS = 2).

When a player of group C gains at least 4.5 real points (SR ≥ 4.5), their fictitious score is increased by a half-point for the third time.

When a player achieves N/2 real points (where N is the number of rounds in the tournament), their fictitious score is brought to 2.

Before the penultimate round, all fictitious points are cancelled and the system becomes a usual Swiss system.

I'm not sure the terms "fictitious" and "progresive" are accurate, since they arise from translation. But I think you get the idea.

[E Michael White]

Stewart Reuben

I’m afraid I’m not convinced by your example. Surely player 32 has met easier opposition not 33 as you state, as lower opponent numbers are stronger. The average opponent number is only 8 places different and will get closer as the event proceeds. This does not look wrong to me unless you are thinking of a 5 round tournament in which case 64 players is a bit high.

The examples you give are also extreme cases because there is a lowish probability of all games going as seeding or of being drawn.

[Stewart Reuben]

Michael
Of course there was a typo. I meant 32 has met easier opposition.
I can assure the effect is very real in my experience. One way to test the robustness of a system is to choose extreme examples. Also it is easy to explain to an outsider.
5 round events are common. I have had no problem running 6 round Swisses for 250 players in the 1980s. You might be lucky and get one person with 6/6 with a normal seeded Swiss. One Evening Standard London Open nobody succeeded in scoring 4/4 with fractional accelerated pairings.
How can progress be made unless there is experimentation? How can there be experimentation if new people simply put the information on computer and then print out the pairings without understanding the principles? If FIDE had a rigid position, then there could be no progress.
Stewart

[E Michael White]

Stewart

But why do you regard it as a problem if 32 has met slightly easier opposition on average after 4 rounds ? He hasn’t had easier opposition in each round.

The drawback of large Swisses with a small number of rounds is not that there may be joint perfect scores but that the imperfections in the method do not have time to iron out. After 4 rounds your example shows that player 32 has met opposition an average of 8 players lower per round than player 33. In later rounds this difference will reduce. 8 players does not seem significant in a tournament of 64.

If FIDE moved to default draw types which eventually were computerised arbiters could still override to improve the draw as you suggest; this ability would not be lost. Reporting deviations from standards would improve matters long run I suggest. As Edward Tandi agrees, posting these at tournaments would deter frivolous tinkering and allow quality assessment later. I don’t think sending all these reports to Fide is necessary unless a player wishes to challenge the draw or the arbiter wants to draw attention to his improvement, so maybe an arbiter should keep the reports for say 3 months. Of course once computerised it might be easier to e:send the whole lot each time to FIDE.

I didn’t used to think the draw mattered much until I saw a heated argument abroad between a GM, an IM and an arbiter when the draw was not as expected. The arbiter simply said he could perform the draw as he wished and would not budge. Afterwards I was told the IM would just miss his chance to achieve a norm or grade level even if he won which would affect his invitations to future events. When someone’s livelihood is affected in this way I think you should consider improvements to the methods.

[Stewart Reuben]

I regard it as a problem if there is a built in bias against certain players. You don't have to agree. You are also entitled to dismiss a difference of 8 places per round as unimportant.
There was one British Championship where two players had very similar ratings. They drew their first 3 games with the same colour. After 4 rounds the higher rated player had met no rated opponents. The lower rated had played all 4 above 2350. The higher rated player was advantaged because he met weaker opposition, but disastrously disadvantaged if he had entered the tournament in the hope of getting a norm (highly probable). I don't dismiss this as unimportant and it is not just a one-off statistical quirk. There was no error in the pairings made. Thus I think the Seeded system is lacking.

Crosstables are sent to FIDE electonically. Checks could already be made if variations looked suspicious or if there were complaints.
A pairings controller who responds 'I'll do what I want' needs retraining. It is more likely that what happened is that the last round rating of the opponent of the player was too low. Thus he could no longer gain a title norm even if he won. Another possibility was that he was paired with an untitled 2400. That meant he had met an inadequate number of title holders. Instead he could just as well have met a 2400 IM. Today I probably would not make such changes.
But the rules are much more user-friendly than they were in 1977 when I started out on this road.
Stewart Reuben