SO WHO'S GOING TO SHEFFIELD? I AM!

[Simon Ansell]

I feel very fortunate that my first ever game in the British Championship at the age of 18 was black against Viswanathan Anand and the year after I started with white against Michael Adams. Had accelerated pairings been used I wouldn't have had the chance to play either of them.

Mine was White against Pete Wells, my coach at the time. Non-accelerated pairings. I was 14 and rated ~2240, probably in the 10 lowest rated in the tournament. How times have changed.

[Alex Holowczak]

Well of course if we're just going to award the titles to the "likely" winners without the need for a 'fair' competition...

Well, Stewart's argument is that it's unfair on the third quarter players. The probable winners aren't in the third quarter though. In fact, at-a-glance there are no female players in the third quarter at all; Susan Lalic is in the second quarter, and the rest of the women are in the fourth quarter. So I don't see where the third quarter argument comes in to this. Similarly, at-a-glance the Under 18s apart from Zhou seem to be equally spread throughout the third and fourth quarter, with perhaps a slight bias to the fourth quarter. One may sneak into the second quarter. I don't know which of the players are the next best Under 21s.

So based on the presented logic - accelerated pairings are unfair to the third quarter - then that doesn't seem to be a problem at all, because that's where a minority of the candidates for those titles are situated.

[Richard Bates]

I feel very fortunate that my first ever game in the British Championship at the age of 18 was black against Viswanathan Anand and the year after I started with white against Michael Adams. Had accelerated pairings been used I wouldn't have had the chance to play either of them.

To be fair, again, if you take the view that the method of pairing shouldn't be used to improve the chances of norms at the expense on what might be best for determining the tournament winners, equally one probably shouldn't select the method of pairing with the thought that players low down in the draw should get a "once in a lifetime" shot against GMs.

Both are clearly potential attractions of one system or the other, but neither should be central factors.

There is IMO an argument for acceleration in finding the British Champion on the grounds that it increases slightly the chances of all the leading players playing one another and getting the toughest overall fields. Because it probably (assuming results go reasonably as expected and someone doesn't unexpectedly sneak through to 2.5 or 3 out of 3)means one less "meaningless" game against players at the lower reaches of the field who are unlikely to be a factor in the tournament as a whole. Although of course taken to the extreme you might as well have an 12 player all-play-all with the "best" players.

On the flip side you can weigh up that the leading players will meet earlier in the tournament and you might get larger floats in later rounds.

[Alex Holowczak]

To be fair, again, if you take the view that the method of pairing shouldn't be used to improve the chances of norms at the expense on what might be best for determining the tournament winners, equally one probably shouldn't select the method of pairing with the thought that players low down in the draw should get a "once in a lifetime" shot against GMs.

I think Loz was expressing the view that by using the system that is perceived to be the best for determining the tournament winners, the "once in a lifetime" concept is another benefit of it.

Both are clearly potential attractions of one system or the other, but neither should be central factors.

There are benefits to acceleration in certain circumstances, but I really don't think they exist this time:
(1) Very few of the candidates to win are in the third quarter, the people whom the system allegedly benefits
(2) An abundance of titled players in the field for norms

I can see the reasons for using acceleration last year; it made a lot of sense. Particularly if you buy the "norm-seeker" argument. This year, I think it's a waste of time and effort on the part of the people who have to put them together.

Although of course taken to the extreme you might as well have an 12 player all-play-all with the "best" players.

Last year many argued in favour of this, but no one has mentioned this year...

On the flip side you can weigh up that the leading players will meet earlier in the tournament and you might get larger floats in later rounds.

Based on memory, this seems to be the bigger problem.

[Stewart Reuben]

1) Virtually every player in the event who seems to care about the matter seems in favour of using the "normal" seeded Swiss system.
That is a verty cogent argument. I have no idea whether the players have been consulted and, if not, why not. I have no idea how many of them undestand the issues.
Alex's 3rd point is simply not correct, although it was the original purpose of Acceleration. Why does he think Seeded is 'normal'. It isn't used in the World Cup, or the Olympiad and wasn't in England prior to 1965 or the British before 1984.
Nigel is in a different position from players at the top of the second half. He does not suffer from the bouncing effect.

>(1) Very few of the candidates to win are in the third quarter, the people whom the system allegedly benefits.<
Thus presumably it is OK to systematically disadvantage them, possibly year after year.


Richard asks why it is easier to get a norm in 9 rounds than in 10 or 11.
The system was built around 9 rounds, the most common number. There are rounding errors when dividing and approximatig to the nearest whole number. I have extracted the table below from the full ones.
GM norm Ra of opponents
9 rounds 7/9 Rating range 2380-2434. 6.5/9 2435-2475. 6/9 2476-2520.
10 rounds 8/10 Rating range 2380-2407. 7.5/10 2408-2451. 7/10 2452-2490.
11 rounds 9/11 Rating range 2380-2389. 8.5/11 2390-2425. 8/11 2426-2467.

It is true that there are occasions when 10 or 11 rounds requires a lower percentage score than 9. 7.5/10 in the range 2408-2434. Particularly 11 rounds against a field 2390-2434. Only for 2426-2434 is it a pronounced lower percentage. Whether 8.5/11 is easier to get than 7/9 depends on the player. Many people find it easier to get their norm after 9 rounds, safe in the knowledge that, if they fail, there will be a second chance.
Some people have said, OK but you can drop a low rated opponent from your field to improve the Ra. That is true, but only if you beat him.
Stewart Reuben

[Richard Bates]

On the flip side you can weigh up that the leading players will meet earlier in the tournament and you might get larger floats in later rounds.

Based on memory, this seems to be the bigger problem.

Depends on your view of what is "a problem". Large floats in later rounds are caused by the leaders having played all of their leading competitors. It's not necessarily good for an exciting climax to the tournament (although if there are nevertheless a few players in contention, having played each other, it may reduce the chances of quick last round draws), but it arguably means that a worthy winner has been found (ie. perhaps less likely that someone wins by avoiding many of the other best players). But then maybe later games between the leading players would be more bloodthirsty when the tournament depends on the result (whereas meeting in earlier rounds could increase the chances of playing safe with a draw).

[Richard Bates]

Richard asks why it is easier to get a norm in 9 rounds than in 10 or 11.
The system was built around 9 rounds, the most common number. There are rounding errors when dividing and approximatig to the nearest whole number. I have extracted the table below from the full ones.
GM norm Ra of opponents
9 rounds 7/9 Rating range 2380-2434. 6.5/9 2435-2475. 6/9 2476-2520.
10 rounds 8/10 Rating range 2380-2407. 7.5/10 2408-2451. 7/10 2452-2490.
11 rounds 9/11 Rating range 2380-2389. 8.5/11 2390-2425. 8/11 2426-2467.

It is true that there are occasions when 10 or 11 rounds requires a lower percentage score than 9. 7.5/10 in the range 2408-2434. Particularly 11 rounds against a field 2390-2434. Only for 2426-2434 is it a pronounced lower percentage. Whether 8.5/11 is easier to get than 7/9 depends on the player. Many people find it easier to get their norm after 9 rounds, safe in the knowledge that, if they fail, there will be a second chance.
Some people have said, OK but you can drop a low rated opponent from your field to improve the Ra. That is true, but only if you beat him.
Stewart Reuben

That's not quite correct - I asked why a 10 or 11 round norm required a higher performance. Surely the rating bands are simply determined by the requirement of a 2450 performance, 2600 performance or whatever? All that determines the level of performance is the specific level of opposition - which is a matter of chance. Maybe we are differing in what we mean by "performance"?

[Alex Holowczak]

1) Virtually every player in the event who seems to care about the matter seems in favour of using the "normal" seeded Swiss system.
That is a verty cogent argument. I have no idea whether the players have been consulted and, if not, why not. I have no idea how many of them undestand the issues.

Well, Thomas Rendle expressed this opinion. He's in a far better position to know than I am. Certainly the only person on this Forum who seems in favour of its universal adoption seem to be yourself. I'm in favour of it in certain situations, which don't appear to exist here.

Alex's 3rd point is simply not correct, although it was the original purpose of Acceleration. Why does he think Seeded is 'normal'. It isn't used in the World Cup, or the Olympiad and wasn't in England prior to 1965 or the British before 1984.
Nigel is in a different position from players at the top of the second half. He does not suffer from the bouncing effect.

The original purpose of acceleration is what I was driving at. I think the seeded Swiss is "normal" because:
(1) It's used almost universally in 2011, at virtually every tournament in the world. The World Cup is a knockout, so it's a different kettle of fish. The Olympiad only introduced acceleration when it was shortened to 11 rounds, and reverted back to unaccelerated in 2010. The British used the lottery system pre-1984, I understand.
(2) Seeded and lottery are the only systems the vast majority of arbiters ever mention in terms of ways to run a tournament. If accelerated was so good, I'd have thought it'd get at least a passing mention.
(3) It's the pairing system that players are most familiar with. Most players couldn't explain how the second round of an accelerated Swiss works. The point of the original seeded Swiss was to provide something that took away the lottery element; i.e. there were defined rules as to who goes where. It became more transparent. Accelerated pairings are not transparent, because players are unfamiliar with how they work.

I guess the main reason I consider it to be normal is because I'm 21 and it's the only pairing system I've ever been exposed to! Based on what people seem to be saying, yourself included, it seems to me to be that you use different Swiss pairing systems in different situations:
Lottery - where a very small percentage of players have grades or ratings. For example, junior tournaments at lower age groups.
Seeded - where the majority of players have grades, and thus can be assigned seeds.
Accelerated - where you can see future problems with people potentially being tied on 100%.

I had considered decelerated pairings, when someone mentioned them on here. My thinking was that you effectively divide players into seeded groups of 4 ideally, but maybe 6 if your number of players isn't divisible by 4. You basically have an APA, or a mini-Swiss with 3 rounds if you have 6 players. After 3 rounds, the groups merge and you pair as normal. With this system, if you have 32 players, you'd have a maximum of 8 on 3/3. With a normal seeded Swiss, you'd have a maximum of 4 on 3/3. So you've effectively slowed the tournament down by 1 round. I'm yet to think of a situation where this would be useful!

>(1) Very few of the candidates to win are in the third quarter, the people whom the system allegedly benefits.<
Thus presumably it is OK to systematically disadvantage them, possibly year after year.

I don't understand why third quarter players are disadvantaged relative to other quarters of the draw? I understand they may bounce, but surely no more than, say, a fourth quarter player would?

[Sean Hewitt]

I had considered decelerated pairings, when someone mentioned them on here. My thinking was that you effectively divide players into seeded groups of 4 ideally, but maybe 6 if your number of players isn't divisible by 4. You basically have an APA, or a mini-Swiss with 3 rounds if you have 6 players. After 3 rounds, the groups merge and you pair as normal. With this system, if you have 32 players, you'd have a maximum of 8 on 3/3. With a normal seeded Swiss, you'd have a maximum of 4 on 3/3. So you've effectively slowed the tournament down by 1 round. I'm yet to think of a situation where this would be useful!

The Hewitt system

It would be useful where you have more rounds than you need to determine the winner of the event. It would be used to get your 'decider' game played in the last or penultimate round though I'm not sure that it would be worth the trouble!

[IM Jack Rudd]

The football World Cup and other similar tournaments are decelerated knockouts. The format seems to be popular and well-understood.

[Simon Ansell]

1)I have no idea how many of them undestand the issues.

I don't claim to be an expert, but it is my limited understanding that the main (only?) reason for using accelerated pairings is in cases where there are more than the optimal number of players for the number of rounds. I don't know the maths, but 2^n has been quoted in various threads. It seems to me that 11 rounds for 82 players should find a proper winner under a normal Swiss-system.

I realise there are advantages and disadvantages to every system. I'm perfectly capable of understanding the maths, so would appreciate a proper argument as to why acceleration, which seems to me to be the system the sponsors and public are least likely to understand, is being used.

[Stewart Reuben]

Being simplistic. If you play against an average field of 2380, for a GM norm you need 7/9 TPR 2600, 8/10 TPR 2640 or 9/11 TPR 2662. Patently the first is easiest. As I showed above, higher Ra is more complex.

I didn't say I was in favour of universal adoption of Acceleration, although one could understand why I might think that, having accidentally invented the system. Certainly while I ran the British completely, it was not used. It is the Seeded System I am now against.

Alex, have you not read my book? Take 64 players and everybody draws all their games for the first 3 rounds.
33 v 1,3,5,7. 31 v 63, 61, 59, 57 with the same colours. Ludicrous. Unlikely, but it has happened similarly to players in the British. Testing a system to its limits often shows whether it works.

Bouncing. In Bermuda I played White v 1400 and won. Black against 2500 and lost. White v 1600 and won. Black against 2400 and lost. White agaist 1800 and won. I was 2200. I think that was unsatisfactory, 3/5 and playing nobody in my reasonabled range. Surely it can't happen to the bottom strength and rated players?

The Olympiad does not use Acceleration anymore; my advice was ignored and it was a disaster. The Burstein system is used, relying on Bucholz. That is, I think, more complex than Acceleration. So is Dubov.

It is absolutely correct to say both Accelerated systems are more difficult to understand than a Seeded system.

When I was last totally the organiser of the British in 1997 there were 83 players, 15 GMs 4 of them 2600+, 17 IMs, 1WGM, 10 FMs, 2 WIM, 1 WFM. Jovanka Houska had no title. The median was about 2295. I can't remember Acceleration being discussed. Was it stronger than 2011? You must define your terms.

I have played in a Swiss using the newly coined term, 'Decelerated System'. It worked OK for me in the Dutch Open Championship. I think the playoff problem for the Football World Cup is even worse than the Armageddon System is for chess.

Stewart Reuben

[David Gilbert]

You have to stay up late to catch Steve Conner’s updated entry list. The new figures for 19 July 2011 – available around midnight – shows 787 entries. Comparable numbers leading up to Canterbury last year were 722 on 18 July 2010 and 751 for 20 July 2010. Looking good!

At the moment Sheffield stands in 8th place in the league table of championships held in the 2000s, with potential to rise another couple of places at least.

Edinburgh (2003) 1009
Torquay (2002) 973
Scarborough (2001) 960
Scarborough (2004) 940
Torquay (2009) 935
Canterbury (2010) 891
Great Yarmouth (2007) 824
SHEFFIELD (2011) 787
Swansea (2006) 768
Liverpool (2008) 742
Millfield (2000) 711
Isle of Man (2005) 503

Poor Richard Bates looks like sitting opposite Nigel Short in the first round. I personally will be cheering him from the analysis room. From a purely selfish spectator point of view I believe the accelerated draw gives the watching public and the media a far more competitive round of matches to start with. And there is usually at least one shock on the four demonstration boards – Andrew Martin’s 'predict the result competition' is often not won until late in the first week. So fear not Richard - as a sequel to Scott Pilgrim and Bobby Fischer, we might have Richard Bates Against the World!

PS: I did enjoy Scott Pilgrim Against the World - I doubt Bobby Fischer Against the World will be half as funny!

[Richard Bates]

Being simplistic. If you play against an average field of 2380, for a GM norm you need 7/9 TPR 2600, 8/10 TPR 2640 or 9/11 TPR 2662. Patently the first is easiest. As I showed above, higher Ra is more complex.

Stewart Reuben

You're right, it is simplistic. The absolute advantage for 9 round norms over 10 and 11 round norms comes from where the base level of 2380 is set (so 10 and 11 round 2600+ performances don't get the advantage of being easier at levels below 2380). However when one considers that these "easier" 9 round norms are basically at the level of 7/9 it is easy to see that it isn't relevant for this tournament. Nobody who has 7/9, 8/10 etc is going to have played a field as weak as 2380 or even close.

[Richard Bates]

Take 64 players and everybody draws all their games for the first 3 rounds.
33 v 1,3,5,7. 31 v 63, 61, 59, 57 with the same colours. Ludicrous. Unlikely, but it has happened similarly to players in the British. Testing a system to its limits often shows whether it works.

Stewart Reuben

I'm not so sure this is a good example. Any system can be shown to have "fundamental flaws" if you program it with unexpected outcomes. Seeded systems make assumptions about rating equating to strength. You don't necessarily show up flaws in the system by putting in results which suggest these assumptions are rubbish. If everyone draws and is arguably therefore of equal strength (whatever the ratings say), how can the lower seeds be said to have tougher pairings?

The longer the tournament the more the "biasses" tend to disappear anyway.