Thoughts on replacing Swiss system tournaments with McMahon

I would like to start a discussion about the drawbacks of popular Swiss tournaments. I am talking about big tournaments where players of various strength enter. I am not talking about elite tournaments like Grand Swiss where the games are competitive from game one, even for top players.

The problem with the Swiss system is that in a big field there are many games between opponents so mismatched (when it comes to strength) that those games aren't very interesting. Typical scenario for me for example (2300 ELO a bit lower in rapid/blitz) is:

-first two games I play against beginners/intermediate players and I win without a fight

-third game I am paired against a GM which is fun for me, probably not so much for him. I usually lose that game

-fourth game I am paired against much weaker opponent (often below 2000ELO) and I usually easily win

-fifth game I play against a much stronger player, I fight but usually lose

-sixth game I play against much weaker opponent, often around 2000ELO and I usually win

This continues and I get maybe 2 games against more or less equally rated opponents. Of course I don't always win against weaker players (especially in rapid and blitz as I am an amateur who doesn't play much) and I don't always lose against stronger players but the point is that most of my games are mismatches with one side being 75%+ favorite rating wise.

This is a problem for popular tournaments because people who travel to those don't have unlimited resources to play all the time. They usually pick a few tournaments a year, schedule the vacation, pay for the hotel and want to have fun playing. If you play 4 tournaments a year and 2/3 of your games are mismatches it's really hard to find motivation to travel. It's bad enough in rapid/blitz but it's absolutely ridiculous in classical when you can easily spend 4-5 out of 9 days playing games which aren't fun for anyone involved.

The way some chess tournaments work around that is introducing sections like u1800, u2300 and 2300+. The problem with that is that if you are at the top of rating range you will not get to play against stronger opponents which is what a lot of people like. It's also more difficult for the organizers to split the tournament (as it's expected every section gets separate room or at least part of the room, needs a separate arbiter etc.).

I would like you to consider the solution to this problem called McMahon system (https://en.wikipedia.org/wiki/McMahon_system_tournament, https://senseis.xmp.net/?McMahonPairing). This is standard in go tournaments. The way it works is that the tournament starts not with everyone having 0 points but with stronger players starting with more. Consider 300 players 7 round tournament. We could have groups of around 50 players, first 50 starting with 5.0 points, next 50 with 4.0 points etc. Usually McMahon is done by rating and not number of players.  So for example 2400+ players are in the top group and then it's 1.0 point less for every 150 ELO with the bottom being 1800 or 1600 (so all players weaker than that start from 0). You can also split brackets by 0.5 point, not 1.0. The details don't matter much for the purpose of this discussion.

The result is that you play a lot more games against equally matched opponents which makes it more fun and allows your rating to grow faster if you are underrated. It has a disadvantage that if you are not in the top group your chances to win the tournament are greatly diminished. This I argue is not much of a problem because players who are so severely underrated to stand a chance to win the tournament and not getting to the top group at the start are very rare and it will be corrected after 1-2 tournaments anyway so long term it doesn't really hurt them. It does make it much better experience for everyone else though.

The experience go players have with the system is that even if they travel to a weekend tournament of 5 games they play 5 interesting games. Again, this is very important for amateurs who don't play chess full time. It's also important that the point handicap you get at the start of the tournament doesn't influence your ELO rating. If you are under or overrated it will be corrected quicker than in normal Swiss because you will get more games against people closer to your rating.

I would like to invite you to give the idea a serious thought even the reaction of many people when they first hear about it is "unfair for lower rated players". I argue it's beneficial for everyone and would make the chess playing experience much more fun and competitive. What prompted me to write this post is my experience in European Rapid Championships - 13 round long Swiss tournament for more than 400 people.  I played 3 games with opponents 400+ ELO weaker, 4 games against opponent around 200 ELO weaker and 4 games against opponent about around 200 ELO stronger while the remaining two were against players 100-150ELO weaker. I lost all the games against stronger players (got black in all of them and didn't play well) and didn't lose a game against anyone lower rated although I did draw some. While the tournament only took two days and was very fun for me due to flawless organization, nice atmosphere and some fun games it was a bit spoiled by so many mismatches. I didn't have the snowball chance in hell to win it all (I am an almost 40 years old recreational player). I wouldn't mind starting 1 or 2 points behind the top guys as long as it means games against players on my level from round 1. If I did exceptionally well I would get to play some GMs anyway and in the next tournament my rating would be much higher. If I did poorly I would finally get to play against 1800 players but if I did reasonably within my rating I would play most of the games against people my strength and it would be much more competitive.

Summarizing this already long post, the McMahon system has the following advantages:

-games are competitive from game 1
-there are way fewer mismatches as the tournament goes, especially if you set the brackets in a smart way
-players from the top bracket play many more games against each other making the final result less dependent on luck in pairings/fringe result. As it is even the eventual winner of the tournament played 4 games against people around 400 (or more) ELO weaker, negative effects of Swiss pairings influence the top guys the least but even they can easily get 30%-40% of non competitive games.
-as you play more games against equally ranked opponents you will correct your rating in 2-3 tournaments even if you are hugely underrated

-it's better for norms (both classes and titles) as you will get people around your strength (and stronger if you are doing well)

-it's better than sections (u2300, u 2000 etc.) because if you do exceptionally well you will start playing against stronger players in a few rounds while you don't have any chance for that in a sectioned tournament. You can still give prices to people who start in the same point bracket. No need to section the tournament.

My experience makes me very reluctant to travel to classical tournaments as I can't afford to waste 3-4 days out of 9 days vacation. I imagine most people with families and jobs are in similar situation. I hope that maybe the idea catches on and if enough people support it might well be implemented one day.

Some additional resource:

-https://en.wikipedia.org/wiki/McMahon

-https://www.britgo.org/organisers/handbook/tournament4 (in chess the bars width should be adjusted because of draws which don't occur in go. For example the recommend size of top group in 9 round tournament in go is 10-26 players it could easily be much higher in chess)

-https://handbook.fide.com/chapter/C0405 it describes one of such systems (Baku method). McMahon is more well established and widely used method with more clarity than accelerated Swisses. If it caught up it would hopefully be used more often than accelerated pairings are now. The introduction to the section in the FIDE handbook though gives good overview of flaws of the Swiss system. It also states McMahon is OK to run in FIDE rated tournaments.

>>I would prefer the Swiss tournament with the option to compete with ONLY fellow GMs if I were a GM. Same if I were an IM, etc...

You would get almost that with McMahon. If you are say 2500+ player you would be in a top bracket and play a first game against other 2500 players. Now depending on details even if you lose that first game you would either play against a player from a lower bracket (for example 2400-2500) who won the game or again against someone from 2500+ bracket. Only if you lose more games you start playing against people from lower brackets who won their games.

It's also better for people who has a chance to make a GM norm (or any other norm for lower titles/classes). If you are a promising FM/IM with say 2400 rating you score for example +2 in your first 3 games and you start playing against GMs from the top bracket. Instead as it is now your first two games are against much weaker opponents which don't really contribute to your chances of making a norm. Then you get a GM and if you lose that game you are back to playing -300ELO opponent which ruins your chances of scoring a GM norm.

>> Sounds like the McMahon system gives an unfair scoring.

This is the gut reaction of many people who first hear about it (and it was mine as well many years ago) but it's better than the solution of sectioning the tournament. Everyone in the top group starts from the same score anyway. If you were in u2300 section you wouldn't get any chance to win the main tournament anyway (as you are not in it). McMahon doesn't really hurt players who don't have a chance to win. It just provides more competitive pairing.

Your student analogy doesn't work because McMahon starting handicaps don't influence ratings. If you do well you will get to higher bracket in the next tournament and even in that one you start playing against stronger opponents after a few wins. It's similar system to having a lot of sections in the tournament (say u1600, u2000, u2300 and 2300+ but more fluid in a sense that you can get a game with someone much stronger if you do very well.

I can't overstate how much improvement that would be for everybody attending a Swiss tournaments. Top players, promising young players, weaker players, beginners and organizers as well as it's much simpler and less hassle than introducing sections.

Sounds like a wonderful idea. Unfortunately I've never heard about McMahon system used in OTB chess tourneys. Maybe chess.com and other sites could start proposing it in the tourneys in order to make it more familiar to chess players.

>> So, you want to play people higher than 2300 right? No problem. You agree to play 2000 rated players in Poland, and if you go to Germany or France you would then be paired with 2500 rated players. You wouldn't automatically get to play higher rated players unless you agreed to also play lower rated players too. Those that have to travel would play higher rated players, those that don't travel as much would be stuck playing lower level players. You want the norm? TRAVEL!!!

That doesn't make sense really.  Even in the same country it means traveling to play. You can't work  normal hours if you are in a tournament. It doesn't matter if it's another city in Poland or in Spain for that matter. In fact it's easier for me to get to for example London than to a town at the other part of Poland. In both cases I am not staying at home and will be taking vacation from my job. It doesn't matter what country it is. In McMahon I play first few games against similarly rated players and then if you do well I play against stronger ones and if I play badly I play against weaker ones.

>>If you lose to a lower rated player you would lose points, no?

Yes but in rating changes are more reliable if you play against people on more or less your level and not in mismatches.

>>So, by playing higher rated players initially you would guarantee that you didn't lose as many points for a loss. I disagree, it looks like McMahon handicaps do influence rating changes.

For every game there are two players. One with higher rating and one with lower rating. The problem are games which have huge rating difference. McMahon minimizes the number of those games. You can only start playing much higher rated players if you do well in the tournament and much lower rated players if you do poorly. In normal Swiss you just get a mismatch after mismatch.

>>If you want to get norms and play against higher rated players, you should have to travel then. It seems like the fairest way to spread the "wealth". You play lower rated players at home, and in exchange you get to play higher rated players abroad.

While in McMahon everyone gets chances for a norm if they do well. Even 2300 player could score a GM norm if they score a lot of points in initial rounds as they would start facing IMs and GMs quickly. Chances for norms actually increase in this system and are less dependent on luck in pairings.

Why introduce such artificial divisions? It shouldn't matter what country you are from if you are playing in a tournament. McMahon "spreads the wealth" quite well. People who do well get to play higher rated guys, people who do poorly will get lower rated opponents. You start playing against equally matched opponents and you have to win your chance to play against a stronger guy or face lower rated opponents if you do poorly. It's the same in Swiss by the way but the differences are bigger. In Swiss you bounce around between extremes all the time: if you do better than expected you get top guys, if you do worse than expected you get easy wins without fight. You keep bouncing around between those two for most rounds and it really not as fun as it should be.

I have a Brilliant idea if you can't or don't want to play for any reason don't join.

>>Let's say you had 12 players, 6 were 2000 and 6 were 2500. The McMahon system seems to say you start with a "segregated" first round. 3 games would be at the 2500 level and 3 games at the 2000 level. What do you do after?

This is a bit artificial example because there aren't any players in between. While McMahon still works in such case let's consider a more realistic example: 12 players at 2500+, 6 players at 2350, 6 guys at 2200 and 6 players at 2050

The tournament starts (assuming bard set to 150 ELO, again this is done depending on number of players and rounds in the tournament. Usually the top group is wide enough to include everyone with a chance to win there).

What happens is that people with 2500+ rating start the tournament with 3 points, people with 2350-2500 rating start it with 2 points, 2200-2350 start with 1 point and the remaining start at 0 points. From now we basically play a normal swiss tournament. After the first round (assuming no draws for simplicity) it will look like this:

6 guys at 4.0 points

9 guys at 3.0 points (6 of them losers from top group and 3 of them winners from the 2350-2500 group)

6 guys at 2.0 point (3 of them losers from the 2350-2500 group and 3 of them winners from the 2200-2350 group)

etc.

Now 2nd round is paired according to the swiss rules. Some of the 2500 players now get winners from 2350-2500 group and some of the winners from 2350-2500 group now get losers from 2500 group. Notice that not only was the first round more equally matched but the second is as well. In the third round someone who win 2 games from and started from 2200-2350 bracket will most likely play either against 2500+ guy who lost two games or against 2350-2500 people who scored 1/2.

This means that the more you win the stronger players you play (like in Swiss) but the gradation is smoother.  Let's not get into technicals how to set the initial brackets. You do that to include everyone in the running in the top group and spread the rest to have a competitive tournament.

>>Only if worst 2500 player loses enough? And at which point?

Yes, the same as in Swiss. The difference now is that 2500 player "wastes" two first rounds playing against 2000 guys while in McMahon they would play the first one against other 2500 player and then the winner would play 2500+ again and the loser someone from 2350-2500 bracket. There are no mismatches. You can "move up" in McMahon if you win enough.

>>If you are going for a norm, shouldn't there be a standard? What you are saying with the McMahon system is that you can get an "easier" norm.

I am saying that you will get more equally matched opponents and chance for a norm. As it is in Swiss if you don't luck out and get enough GMs/IMs from enough federations you won't get the norm. In McMahon if you are already say 2300-2400 player you will start playing against other FMs/IMs and if you do well you will quickly start getting GMs. McMahon in general means less mismatches.

@OneTrickPony_82: What's the advantage of this system over just having different groups playing different tournaments?

If there was a demand for it, that type of tournament would be run. I would have to look at FIDE regulations, but is that pairing system a valid one for norms?

It also sounds mostly useful for large events, where groups would have a sufficient number of players to keep similar ratings closer together. Another option would be using Swiss accelerated pairings, which works in a similar manner, though it generally has different goals.

>>What's the advantage of this system over just having different groups playing different tournaments?

It's more smooth. Groups are set somewhat arbitrary. For example if you have u1800, u2200 and 2200+ then some of those group can still be wide. Also if you are near the top of the group you won't get games against stronger opponents even if you do very well. It's also more difficult to handle with accomodation, arbiters etc. McMahon is basically more finely grained sectioning and it's more fluid as you can get games against stronger people if you do very well. 2199 ELO player will only play against 1800-2199 people while 2201 player will play only against higher rated ones regardless how they do even though the rating difference is 2 points. In McMahon the difference when it comes to average opponent is much smaller. Even if you are in a lower bracket to start with you get similarly rated opponent (say 100 ELO lower) but then if you start winning you can easily get opponents stronger than higher rated person who lost some games.

It achieves similar goals as sectioning (having different groups) but in more fluid and easier to organize way. There is a reason for example European Rapid Championship wasn't divided into groups: it's more difficult to handle from organizational point of view. Still it would be so much more fun for everyone if it was played using McMahon and the top players would get to play more against each other as well to make the final results more reliable.

>>If there was a demand for it, that type of tournament would be run.

Well, with that attitude we would never get any progress. The system is huge success in go world and frankly I think it's so much better than Swiss that everyone would have much better experience that it would gain wide support. It's just not that easy to think of it if haven't seen how it works before. There is certainly inertia and tradition. Every change must start somewhere so I am trying to get the spark going. I honestly think it's such an important change for chess being more fun game it trumps all the other experiments like time controls, number of rounds etc.

>>I would have to look at FIDE regulations, but is that pairing system a valid one for norms?

I don't know. Maybe if enough people support the idea then the regulations could be changed. It's much better than Swiss in a sense as you get more opponents from similar rating range so the norms from McMahon would be more reliable than those from Swiss.

>>It also sounds mostly useful for large events, where groups would have a sufficient number of players to keep similar ratings closer together

Yes, the bigger the tournament, the lesser number of rounds and the bigger the rating difference between highest/lowest rated players the better it is but it would make vast majority of tournaments much better. You just don't do that many brackets in smaller events. You don't really  need to keep the groups together though. If someone is rated 2200 and wins 3 first games it's ok for them to play a 2400 IM who won one and and lost two for example. You still get games with people of different rating it's just not as extreme as in Swiss. You have to "earn" a game against a better opponent.

Sectioning allows lower rated players to still win a prize if they perform well (and often their rating then climbs into the next section so they're playing up in the next tournament they go to). How does McMahon handle your average tournament player wanting to compete for a prize?

You still have final tables for the different sections - a little messed up, of course, since e.g. players performing extremely well might have to compete against players from a higher rated section, but you have this kind of distorsion with the Swiss system anyway.

While I found the approach quite strange when reading the OP, I find it quite appealing now. Works.

Large tournaments generally divide players into sections grouped by rating, or in youth events sometimes age. When the sections themselves are large (we face this with the Washington Elementary Chess Championship that may have 300 in third grade), there is an option to use accelerated pairings.

>>Sectioning allows lower rated players to still win a prize if they perform well (and often their rating then climbs into the next section so they're playing up in the next tournament they go to). How does McMahon handle your average tournament player wanting to compete for a prize?

You can still have prizes for say u1800 or 1800-2200. Just set McMahon bars accordingly (so all u1800 starts with the same number of points, all 1800-2200 with the same number of points etc.). You may want to give a price for u2200, u2000, u1800, u1600 instead of 4 prices for u2200 (for example) to make it work better with McMahon.

>>It sounds like the McMahon system is very similar to accelerated pairings used in swisses here in the U.S

I've never seen this used here but yes, the idea is similar. I think McMahon is clearer as you can determine top group and lower bars in clearer terms (for example in big strong Swisses all 2500+ players in top group, then lower groups every 150ELO down to say 1600). It's better in McMahon as it lets you define terms in clearer way and doesn't force you to have groups of equal size but other than that it's similar. I would say it's a better/clearer way to achieve the goal of accelerated pairing.

>>Large tournaments generally divide players into sections grouped by rating, or in youth events sometimes age.

Yeah, that achieves similar goal to McMahon but with many quirks I mentioned in the original post. McMahon is a better way to achieve what accelerated pairings or sectioning does. For example European rapid/blitz championship wasn't sectioned (probably because it's more difficult to organize and cut-offs always make some people unhappy as some of them get 0 chance to play against stronger opponents which is part of the fun) while it could very well be run using McMahon making it more fun, competitive and fairer for everyone (I say fairer as McMahon produces more reliable results than Swiss for the top players as they get to play more games against each other making it in effect a tournament with more rounds and less dependent on pairing luck for tie-breakers like Bucholz). I feel sectioning isn't done enough or done at all because of difficulties it introduces while McMahon would be much easier to carry on.

Sounds like a very interesting suggestion.

Is there any demand among GMs to this pairing system?

>>Is there any demand among GMs to this pairing system?

I think GMs in general and top GMs especially are the least affected by Swiss bouncing. Sure, in the big open a GM plays maybe 2 or 3 hugely mismatched games at the start but then every game is competitive as they fight for the top spots. Still, it would save them a few days of playing vs much weaker opponents so they would play more against other contenders for top spots which makes eventual winner less dependent on pairing luck. Another thing is that you can now organize 7 round tournaments instead of 9 round ones and still have more reliable results. I think for professionals for whom travelling is a major cost there are benefits as well.

I think players who wouldn't be in a top group benefit the most though. It's just much more fun and competitive experience for everyone involved and much easier for organizers as well (no need for sectioning, capping number of entrants, minimum ELO requirements).

Take Grand Swiss for example: it was very difficult to get in as the idea was to get around 150 players as you can only play 11 rounds and you want top players playing each other. You wouldn't get that with more entrants. If you use McMahon and have big enough venue you can easily have 500 or 1000 players. Just set the top group cut-off at 2600 or 2650 and accept all comers. With McMahon it doesn't hurt Carlsen, Nakamura or Akobian that there are many 1800's in the tournament. They start with 8 points less (assuming 100ELO = 1point) and will never be paired with the top guys.  They might get to play against a 2000 or 2100 if they do well though.

Another example is World Rapid: right now you need to be 2550+ to get in. The main reason is again that you get to play 15 rounds and you want all top guys to play each other. If you make it 500 person tournament there will be too many pointless games. Use McMahon and your only limitation is the venue. Organizers make more money as there will be many players willing to travel to World Championship.

I think you could also reconsider round-robin tournaments. Right now the main reason to organize those small events is that you want GMs to play each other and not battling it with a random 2100 Joe in big Swiss. So what organizers do is to make "GM group" for like 10 guys, then "A group", "B group", "C group" or what not. It sucks you missed the higher group by a small margin as now you will only player weaker guys. Make it McMahon and everyone plays in the same tournament. Everyone gets the chance to play higher rated guys and the results are still very reliable.

McMahon is just big win for everyone: players and organizers. It just needs a few tries so people realize the benefits and the chess world will be so much better for it.

This is why tournaments here are grouped by rating...

>>Become a TD and organize your own tournaments. 

I don't have resources for that, at least not yet. Still, just because I don't organize my own tournaments doesn't mean I can't point out an idea to improve them. It's also well tested and working idea. It's likely many TDs didn't even know about it though.

>>This is why tournaments here are grouped by rating...

Yeah, that's another solution but as pointed above that has many disadvantages in comparison to McMahon and those disadvantages cause many organizers to forgo sectioning to make it more manageable (I pointed out European Rapid but that's the case with most Swiss tournaments).

Is it the same as Accelerated swiss?  I remember that it was applied in Cappelle la grande chess tournament in France as the number of players exceeds 700...

I think its a very good option to play tournaments with these systems, I agree the mismatches really annoying. 

Unfortunately, McMahon isn't currently FIDE-approved.

I'm a TD and I'm thinking of trying Baku Acceleration, which is similar.