Computer pairings

[Alex Holowczak]

But if any of the arbiters want to explain why anyone still chooses to pair manually, rather than by computer program, I would be interested.

Arbiters like to do manual pairings for tournaments because they find it enjoyable. I don't find it enjoyable because I see it as an administrative task, and it's an administrative task that can be done far quicker by a machine than me. So why waste my time? (I'm usually the mug who has to write 100+ pairing cards out in advance, of course. It's quite possible that this has clouded my judgement!)

When checking pairings with computers, some arbiters tend to insist that "the pairings are wrong" or "the software is not fit for purpose". British arbiters are used to CAA pairing rules, and tend not to know FIDE pairing rules. Of course, the answer isn't that the pairings are wrong, or the software is not fit for purpose. The software gets the pairings right, it's just pairing to a different set of rules, and it's a set of rules that British arbiters tend not to use. So they think the pairings are wrong.

Surely the pairing algorithm is not so complicated it can't be programmed, is it?

It depends whose rules you pair to.

FIDE pairings are accurately implemented by various pieces of software. I've confidence in Swiss Master to get the pairings right. Tournament Director seems to be there too. Swiss Perfect gets it wrong, based on some pairings I've seen at a tournament I did last year.

CAA pairings are difficult to program. There are too many discretionary elements. Do you keep players apart from the same club in round 1? Round 2? Does it matter what scoregroup you're on? Do you keep players apart from the same family? Do you put players from the same family on the same side of the draw in round 1 to minimise their likelihood of playing later on? How do you tell the computer to look for these things?

Some arbiters argue that the CAA pairing system is better because it's fairer in certain situations. Fine. I argue that players don't really care which pairing system you use, so long as you make it clear on the entry form which pairing rules you're using, and which software you're using to do them. It's a necessary skill for arbiters to be able to explain the pairings that the computer spews out in case of a query, rather than just saying the computer did it. Someone queried two of their pairings at the Birmingham Rapidplay. One was easy to answer. The other needed a bit more investigation, and I contacted him after the event to explain why it was right. He seemed OK with it.

[harrylamb]

Surely the pairing algorithm is not so complicated it can't be programmed, is it?

I am neutral on this. I run English tournaments where I use manual pairings and play in several continental events which use computer pairings.

The answer to your question is that it depends!

Computer pairings demand a lot of rigidity and need rules that may be difficult and /or unwelcome in English playing conditions

Examples are
With computerised pairings players on the continent have to turn up 2 hours early to register for round one. Because once you do the draw it is difficult to change it. In fact under FIDE rules once you publish the draw you cannot change it.

Repairing for absent players is difficult

Avoiding playing certain players eg father and son together is difficult. In each of the last three years I have been paired abroad with my son. In forty years I have never been paired with one of my sons in England.

Similarly avoiding pairings between players from the same club is very difficult

Handling players who withdraw is difficult. As an arbiter in England I often have players who withdraw 5 minutes before the round starts. This is very rare on the continent. If it does happen it is handled by giving their opponents byes. Using English manual pairings you can repair.

I do not have a deep knowledge of the latest version of computerised programs. It may be that they have become more user friendly to English playing traditions. But because English pairings are traditionally manual there are a lot of flexibilities in manual parings that have become the expected norm. These flexibilities are difficult to replicate in computer programs

[John McKenna]

As usual Alex H and Harry L write a lot of sense in reply to Paul C. However, Alex writing out pairing cards by hand shows a predisposition for manual pairing due to the effort involved. Why not print off the pairing cards using the h/w and s/w available but keep them for emergencies? With 3-D printers just around the corner isn't it time to make full use of 2-D technology? I suspect that organisers and officials in England still favour a hands-on approach but the bots are coming, like it or not.

[Roger de Coverly]

Do you keep players apart from the same club in round 1? Round 2? Does it matter what scoregroup you're on? Do you keep players apart from the same family? Do you put players from the same family on the same side of the draw in round 1 to minimise their likelihood of playing later on? How do you tell the computer to look for these things?

You can minimise the likelihood of pairings you don't want by subtle changes to the ranking order, that's assuming all the methods work by ranking rather than grade. Beyond that, cannot certain pairings be prevented by use of a flag designed to avoid pairing players from Israel against opponents under instructions to default rather than play them?

You can get difficulties at the extremities, the sixth round of a tournament with only around 16 players. The logical sporting pairing of players with 4/5 playing players with 3/5 was one the program didn't want to deliver because of the difficulty it then gave pairing a player with 1/2 from 5.

I think e2e4 have demonstrated that you can handled re-pairings in the context of an automated pairing program. You don't need to abandon the whole idea of computer pairings just because of a few late arrivals or no shows. The British approach of getting the first round started on time is preferable to the Continental approach of sometimes starting an hour late while they try to figure out who has entered.

[Alex Holowczak]

Why not print off the pairing cards using the h/w and s/w available but keep them for emergencies?

Swiss Master does the pairings properly. Being a Dutch program, it doesn't have the option for printing Pairing Cards. In fact, what it calls Pairing Cards, we call Result Slips.

Tournament Director doesn't do CAA pairings properly, but it does allow you to print what British people call Pairing Cards.

It's cheaper, easier, and less resource-intensive to use Swiss Master and get it to do the pairings, in my opinion.

When I organise events, I make sure of a few things:
(1) Others have laptops with the software installed, in case mine fails (at the Warwickshire Championship, three of us had the software installed)
(2) I back-up my software onto either a memory stick, or a Cloud storage device (if I take laptop #2 as a backup, then the Cloud is easier than the memory stick)
(3) Pairings and wallcharts are printed. Obviously for a blitz this is difficult, but for a congress there's enough time. So I can then cut up the wallcharts to create mini pairing cards. This wouldn't be ideal, and would be a lot of manual effort. But of course, I only need to do this if at least one other laptop fails.

For the British Blitz, Dave Thomas, Matthew Carr and I will be there. I'll be the computer operator, Dave and Matt will be doing the floor arbiting, and feeding me with what Swiss Master calls Pairing Cards. Matthew has a copy of Swiss Master on his machine, which he's bringing with him. I'll e-mail the CSV to import entries the night before to both Matt and DT, just in case my laptop goes wrong the following morning. We have a printer installed on both mine and Matthew's machine already. (There's no backup printer, but the backup for that is that we can each write out the pairings for 1/3 of the field.)

In terms of backups for each other, Matthew has experience of using the software if I go down ill the night before.

Incidentally, if we were running a four-section Rapidplay - as we often do - then we'd have exactly the same organisational structure. This is different from the world of pairing cards, where the norm is to "control a section". You can see how quite easily, you need fewer staff on site if you're using computers to do the bulk of your administration.

[Alex Holowczak]

You can minimise the likelihood of pairings you don't want by subtle changes to the ranking order, that's assuming all the methods work by ranking rather than grade. Beyond that, cannot certain pairings be prevented by use of a flag designed to avoid pairing players from Israel against opponents under instructions to default rather than play them?

Yes, you can do that quite easily. In the case of FIDE pairings though, that's what it's for: To avoid that kind of issue where the opponent has to default or his government will do something nasty to him. I know of no relationships between brothers where playing one-another results in an international diplomatic incident, or the player being treated as a criminal. So I tend to avoid using it. I make it clear in advance though that that's what we're doing, so people know what will happen before they enter. If they still choose to enter, then they know what to expect.

I think e2e4 have demonstrated that you can handled re-pairings in the context of an automated pairing program. You don't need to abandon the whole idea of computer pairings just because of a few late arrivals or no shows. The British approach of getting the first round started on time is preferable to the Continental approach of sometimes starting an hour late while they try to figure out who has entered.

Absolutely, it's easy enough. We're not doing it for the Blitz, because the game only lasts 3 minutes per side, and there's not really enough float in the schedule. However, for the Rapidplay, we do re-pairing using the software, and it works fine.

[Sean Hewitt]

With computerised pairings players on the continent have to turn up 2 hours early to register for round one. Because once you do the draw it is difficult to change it.

It's very easy to change computerised pairings. There is no need to require players to turn up and register in advance for this reason.

In fact under FIDE rules once you publish the draw you cannot change it.

Unless the rules of the tournament state otherwise.

Repairing for absent players is difficult

Not in the slightest. It's dead easy and takes about 15 seconds.

Avoiding playing certain players eg father and son together is difficult. In each of the last three years I have been paired abroad with my son. In forty years I have never been paired with one of my sons in England.

It's not at all difficult. Computers have the concept of forbidden pairings and you can program it not to pair certain players together if you wish. Whether that is desirable is a different discussion, but it's easy to avoid if you wish.

Similarly avoiding pairings between players from the same club is very difficult

It's not. It's very easy, if that's what you want to do. See the point about forbidden pairings above.

Handling players who withdraw is difficult. As an arbiter in England I often have players who withdraw 5 minutes before the round starts. This is very rare on the continent. If it does happen it is handled by giving their opponents byes. Using English manual pairings you can repair.

Using a computer, it is dead easy to repair in these circumstances. However, I can't speak for any events other than my own but I've only ever one had a player withdraw 5 minutes before the round started. Such action, without good reaso, is unacceptable in my opionion. The player in question will never play an e2e4 event again!

...because English pairings are traditionally manual there are a lot of flexibilities in manual parings that have become the expected norm.

Surely all pairing systems were traditionally manual?! I think the difference is that most parts of the world expect arbiters to pair according to the rules. Some arbiters in this country seem to think it's ok to make it up as they go along.

[Stewart Reuben]

Alex Holowczak > Absolutely, it's easy enough. We're not doing it for the Blitz, because the game only lasts 3 minutes per side, and there's not really enough float in the schedule.<

If it is blitz, why does the game not last about 4 minutes each side? Why not each round being a two game match? That removes all colour problems. Then it is even easier to do by hand.
AND SOME OF THESE NEW-FANGLED COMPUTER PROGRAMS CANNOT COPE.

[Alex Holowczak]

If it is blitz, why does the game not last about 4 minutes each side?

It's 3 minutes plus 2 seconds per move. So if someone doesn't turn up, then the game will last 3 minutes plus 2 seconds per move.

Why not each round being a two game match? That removes all colour problems. Then it is even easier to do by hand.
AND SOME OF THESE NEW-FANGLED COMPUTER PROGRAMS CANNOT COPE.

The computer program we're using can't do double rounds. We're using that for web output of results as much as doing pairings and things. Swiss pairing cards are quite poor at producing web output. It's also quite tricky to produce FIDE-rating reports from pairing cards. The best way to create one is to put it in some pairing software...

[Stewart Reuben]

3 minutes + 2 seconds for a 45 move game at the maximum comes to 9 minutes. At a minimum it comes to about 6 minutes.
What has it got to do with somebody not turning up?

Personally I don't like 3 minutes + 2 seconds because my reaction time is no longer good enough to move within 2 seconds. And to think I used to play 1 minute chess. On the Countdown clock we occasionally played 40 moves in 1 miute + all the remaining moves in one minute. Najdorf said he didn't do gymnastics.

You agree with me, you are using a program that can't cope with other than 1, 0.5, 0. There are several on the market, of varying quality, that can. They may only accept 4 figure numbers, but that doesn't matter for blitz. Presumably the British Blitz Championship will be FIDE Rated, It can't be ECF.

I have absolutely no doubt, tournaments are better played as 2 game matches. That is easy for blitz where there is time.

[shaunpress]

The list of currently FIDE approved Swiss Pairing programs is at http://www.fide.com/fide/handbook.html? ... ew=article
However this list is slightly out of date as Vega is also now approved for the Dutch System, using the JaVaFo pairing engine.
It is also worth pointing out that each of these programs may produce slightly different pairings, as they were tested against the rules that were in force at the time, and that changes to the Swiss Pairing Rules have occurred since then. As a result the FIDE Swiss Pairings Program Commission (of which I am a member), will be re-testing all the approved programs against the current Swiss Pairing rules and either approving them again, or asking the developers to make the necessary changes.

Annoyingly, the one program that can easily do most of the things mentioned in the previous posts (manual repairing, avoiding pairings, multiple game matches etc) is probably Swiss Perfect, which of course does not do the actual pairings 100% correctly.

With my arbiters hat on, the program I mainly use for FIDE Rated events is Swiss Manager (mainly due to the desire to publish pairings and results at http://www.chess-results.com), although it does take a bit of learning and practice to discover all the tricks needed to use it well. The other program I like to use is Vega, as it produces quite nice web pages for the tournament website eg http://www.doeberlcup.com.au/DBL2013_re ... emier2013/

[Alex Holowczak]

What has it got to do with somebody not turning up?

I was talking about the concept of re-pairing in a blitz game. I'm unlikely to re-pair a game when both players turn up.

You agree with me, you are using a program that can't cope with other than 1, 0.5, 0. There are several on the market, of varying quality, that can. They may only accept 4 figure numbers, but that doesn't matter for blitz. Presumably the British Blitz Championship will be FIDE Rated, It can't be ECF.

It can handle it, but I have to treat the players as 2-player teams. This causes other problems, such as creating the FIDE-rating report. And it would give each team an erroneous colour history. It would be easier to use pairing cards. It is indeed being FIDE-rated.

[Alex Holowczak]

With my arbiters hat on, the program I mainly use for FIDE Rated events is Swiss Manager (mainly due to the desire to publish pairings and results at http://www.chess-results.com), although it does take a bit of learning and practice to discover all the tricks needed to use it well.

There are a few reasons I chose not to use it:
(1) It's the most expensive of all of the options.
(2) Questions about its ability to do the pairings. I know of one competition where the pairings were incorrect according to the software. All of the arbiters agreed it was wrong, but the regulations wouldn't allow a change. The results appeared online using Swiss Manager, but I don't know if that was doing the pairings.
(3) It didn't have a demo mode, that I could find. I would want to play around with it before committing to buying it.

[Stewart Reuben]

Alex >I was talking about the concept of re-pairing in a blitz game. I'm unlikely to re-pair a game when both players turn up<

Nor, I hope where neither turn up.

Colour is almost totally irrelevant for a two game blitz match. The players wouldn't even notice if they kept on having the same colour in the first game.

By the way, I bought the license to use Swiss Manager and Swiss Master, in order to introduce one in The Gambia. But, I abandoned that idea when I realised nobody had heard of the Swiss System, nor had a computer. I expect I could transfer the right. Nobody trusts me to do computerised pairings.

[Alex McFarlane]

With my arbiters hat on, the program I mainly use for FIDE Rated events is Swiss Manager (mainly due to the desire to publish pairings and results at http://www.chess-results.com), although it does take a bit of learning and practice to discover all the tricks needed to use it well.

There are a few reasons I chose not to use it:
(1) It's the most expensive of all of the options.
(2) Questions about its ability to do the pairings. I know of one competition where the pairings were incorrect according to the software. All of the arbiters agreed it was wrong, but the regulations wouldn't allow a change. The results appeared online using Swiss Manager, but I don't know if that was doing the pairings.
(3) It didn't have a demo mode, that I could find. I would want to play around with it before committing to buying it.

I'm not sure Swiss Manager is the most expensive. Swiss Master cost €60 per copy. Swiss Manager costs either €150 or €200 for a site licence. So for that money the ECF could buy it and allow all registered events to use it.

I find Swiss Manager awkward to use so wouldn't necessarily recommend going down the site licence route, thouh it is OK if you don't plan to make any minor alterations to the original draw.