Pairings for title-norm Swisses

[IM Jack Rudd]

Opening this one up to the floor: if we all put our heads together, we may come up with a good pairing system we hadn't yet thought of individually, or at least work out the merits and demerits of the systems currently in use. So what do we have?

(a) Standard Swiss systems
(b) Accelerated Swiss systems using the method we're using at Hastings
(c) Accelerated Swiss systems using the bonus point method
...
Any others we know of?

[John Hickman]

If Swiss splits by half, and accelerated Swiss splits by quarter, which is meant to improve things for norm-seekers, would a hyper accelerated swiss, splitting into eigths be better or worse?

Also Accelerated Swiss at some point turns to Swiss, so you could have three phases, Hyper, then Accelerated, then normal swiss.

[Ian Thompson]

Opening this one up to the floor: if we all put our heads together, we may come up with a good pairing system we hadn't yet thought of individually, or at least work out the merits and demerits of the systems currently in use.

Before you can work out the merits of different systems you need to decide what the objectives are. For instance, is trying to improve the chances of some players getting norms more or less important than the most deserving (which itself needs defining) player winning the tournament? If there are special prizes for which only some players are eligible (e.g. British Championship qualifying places, rating prizes, women's prizes, etc.) how important is it that the most deserving player wins them (compared to who wins overall prizes and other special prizes)?

[Ian Kingston]

One further consideration is implied in Jack's use of the phrase 'Standard Swiss systems'. The FIDE Handbook describes several different ways of carrying out Swiss Pairings, and it may be that one of these is better than others for achieving a particular goal. For example, the Dubov system:

attempt[s] to equalise the average rating of the opponents of all players of a score group. Therefore the pairing of a round will pair players who have played low rated players before with players having high ratings now

I don't know whether anyone has studied this. The effects on norm prospects, for instance, might be interesting.

[Sean Hewitt]

The swiss system was invented with one aim - to determine an outright winner in an event where there are more than 2^r players (where r is the number of rounds to be played). It was devised long before the concepts of norms were around.

It's entirely possible (perhaps even probable) that any attempt to improve opportunities for norm seekers, or more fairly split players within a score group (as required to determine British Championship Qualifying places for example) may result in compromising that main aim.

[Alex McFarlane]

Dubov has been tried on occasions. John Turnock I believe used it a number of times. It has generally proved unpopular with the players because of the variation in opponents in alternating rounds.

It is also recognised as Sean hints at, that the Swiss system is not good at differentiating people away from the top score groups. Grading prizes therefore tend to be a bit of a lottery, often depending on the strength of opposition of contenders in the last round ie a person half a point behind can have a much easier opponent than his rival.

[Ian Kingston]

Dubov has been tried on occasions. John Turnock I believe used it a number of times. It has generally proved unpopular with the players because of the variation in opponents in alternating rounds.

I'm not advocating Dubov - I have no experience with it - but I wonder whether the reason that players don't like it is simply a lack of familiarity. If (stress on the 'if') there are compensating advantages, then it might be worth persisting with. The accelerated Swiss leads to widely varying strengths of opposition as well. It might be that the best approach is not to have an Open event: a minimum rating might be more appropriate than trying to make the Swiss system cope with a 1000 point range between the strongest and weakest players.

In would be interesting to see computer simulations of the outcome of various systems with varying distributions of players' ratings. That's a far from trivial task, but it would at least remove some of the subjectivity from the discussion.

It is also recognised as Sean hints at, that the Swiss system is not good at differentiating people away from the top score groups. Grading prizes therefore tend to be a bit of a lottery, often depending on the strength of opposition of contenders in the last round ie a person half a point behind can have a much easier opponent than his rival.

I have benefited from this and suffered from it as well. I think most players have come to terms with this issue. It's the added complication of trying to provide norm opportunities that seems to be causing most of the debate.

[Alex McFarlane]

I agree that preventing low rated players from entering would make things easier from the point of view of the pairing system. This might be possible in a properly funded event with a major sponsor. Unfortunately most events rely on entries to break even and therefore cannot afford to reject any entry. Running a congress is a difficult business when it comes to achieving a balanced budget. To turn away potential revenue is just not a real possibility.

Without trying to wander too far from the point, it is significant that prize funds for congresses have not increased significantly in over 20+ years, indeed some have even reduced. In real terms this represents a huge cut in the prizefund. This acts as a major disincentive to younger players trying to make the breakthrough when they realise that a half decent living is not possible from the prizemoney on offer.

Computers should provide the potential to investigate alternative pairing systems but, as stated, conflicting objectives make the problem even more difficult. I agree players accept that the grading prizes are a bit of a lottery. With regard to titles, FIDE have made achieving norms more difficult due to changes in the regulations. This has the effect of meaning fewer titled players. But fewer titled players will have the effect of meaning even fewer titled players which could be a vicious circle. This would be a further disincentive to up and coming players. For this reason we must try to find ways of providing norm opportunities. Reaching an equitable balance is all important.

[IM Jack Rudd]

The best way to get round the issue with grading prizes is probably to use the "largest number in column F" method; i.e. award them for improvement over expected score, rather than for raw score. That way, variations in field at the end of tournaments become less critical.

[Stewart Reuben]

Jack is right there, except it also avoids the problem of variation in any round.
Jonathan Mestel wrote a long piece in the BCM (I think) where his conclusion was, 'There is nothng wrong with any Swiss Pairing System. It is how you rank the results that is all important.'

Various editions of my book have discusse various systems. Bustein has some advantages as has Dubov as have Accelerated Systems. The Bonus point system of Acceleration probably would work fine if they continued it to round 3.
I rather think the Fractional Accelerated System we use is not working as well as it used to because there has been a change in the demographics of the ratings of the players. The FAS System was used in the 1980s to cope with 250 players in the Evening Standard London Chess Championship. It coped very well by dividing he players into 10 groups. 1 played 2, 3v4 and so on.

If you are going to rank the players on their raw score, probably the standard top half v bottom half in rating order in every round is the most efficient at finding the 'correct' winner. It may also be the one that is most deeply flawed for determining rankings lower down.

Possibly the 'fairest system is a lottery one of players on the same score group. It will, with hindsight, have inequities, but these will vary from tournament to tournament.

Page 41-42 of 'The Chess Competiors Handbook' by Bozidar Kazic gives somed criteria that the London Chess Association decided that organisation wanted the system used to be met.

Stewart Reuben