Situation 1:
Player A - 1999, 3 points
Player B - 1800, 3 points
Player C - 1799, 3 points
Player D - 1100, 3 points
Assuming pairing histories and colors work out, Player A must play a significantly tougher last round than Player B due to his higher rating, even though they compete for the same class prize.
Situation 2:
Player A - 2600, 3 points
Player B - 1999, 2.5 points
Player C - 1800, 2.5 points
Player D - 1100, 2.5 points
Assuming pairing histories and colors work out, Player B must play a significantly tougher last round than Player C due to his higher rating, even though they compete for the same class prize.
Wouldn’t it be fairer to have a general pairing principal like this: Within a given score group, the top-half plays against the lower-half, and given this restriction the pairings are random. ?
Also, similarly: If a score group has an odd number of players, a player at random from the (top half of the?) next lowest score group shall be moved up for pairing purposes.
If there are any biases to be handed out, I think it should be to the higher rated player, unlike as it is now. Although random seems better than both. In professional sports you have typical pairings of 1-8, 2-7, 3-6, 4-5 which contrasts to chess with 1-5, 2-6, 3-7, 4-8.
The original reasoning for this pairing method was “excitement” but philosophically, I would deem this principal inferior to “fair”. The NBA got a lot of criticism this season for forcing the Mavericks to play the Spurs, arguably the two best teams in the conference, in their tournament semi-finals. Chess doesn’t blink, but the NBA was forced to change their rules to make sure they would meet in the finals.
In my first situation above, maybe the 1999 player only drew the 1799 player and of course lost to the 2600 in situation two, and the 1800 goes on to win the class prize by beating esaily an 1100 that the 1999 player could have easily beat as well. So the probably better player did not win the prize, due to unfortunate pairings.
SADLY, had he beat the 2600, he still would have received only the same prize as the 1800 beating the 1100.
The principal of last-round class-prize pairings option offers some relief to the problem…but it is only an option…for the last round…sometimes it’s not a class prize at stake…sometimes the players are closely rated but in 2 classes… Anyways, something like this should have been the default for years (my humble opinion).
Class prizes are a bit arbitrary anyway, and you describe a threshold effect that occurs. But there’s another side to the story.
Maybe this 1999 player could have been rated 2000 and ineligible for the A prize. Or the 1800 should be 1799 and eligible for the presumably easier to win B prize. This whole situation, which you describe quite correctly, can be seen as their punishment / reward for “squeezing into” a lower class or “being stuck in” a higher class by virtue of a single rating point or just a few.
The 1999 player is almost certainly likely to win more money in this situation, taking everything into account, than if he were rated a single point higher at 2000. So the pairing effect you point out is generally trumped by the effect of competing for an easier prize.
It’s hard to design a proper system for rewarding mediocrity (i.e. rewarding people for having low ratings). This is just an example.
But it’s worth discussing the pairings that affect the top prizes. That’s analogous to the Mavericks vs. the Spurs. It’s not inherently arbitrary, and there’s no threshold effect. The pairings, however they work out during the competition, are supposed to be compensated by the tiebreak system. Various competitions use all kinds of tiebreaks and none at all, because nobody is sure of the best approach.
The NBA has a business interest in maximizing the probability that the two favorites meet in the finals (even moreso than maximizing the probability that the best team wins in the end, since the games rather than the title bring in the money), and some chess tournaments may feel the same interest.
For any specific set of pairing rules you can come up with cases that seem to favor one player or another. I can just as easily come up with pairing cases that tend to favor the higher rated player. Consider the following case:
Player A 1999 3 points
Player B 1930 3 points
Player C 1870 3 points
Player D 1800 3 points
Ignoring color allocations, players A and B seem to have easier pairings than players C or D, don’t they? Yet they are all going for the same prize.
The rules that we have tend to produce more balanced and MORE FAIR pairings than random selection. Yes, one time or another a certain player will SEEM to have an easier pairing and sometime DOES but it’s not necessarily so.
Take the case first case you brought up: an 1100 would seem to be an easier opponent than a 1799. But he had managed to score 3 points in the tournament – the same record the 1999 player had managed. And he managed to do this with a much tougher (unless you accelerated pairings) first round opponent! He might just be grossly underrated. For myself, I’d usually rather have the higher rated opponent. Playing for the prize isn’t all that important to me, I’d rather play a strong opponent AND have a better chance of improving my rating. Playing a grossly underrated 1100 player isn’t my idea of fun – you can’t win more than a single rating point and you can lose a bundle (enough to wipe out a whole tournament’s other results).
Hehe, I have to disagree/clarify a number of points you mention. In your situation Players C & D “have tougher pairings”. Compared to what? Than playing each other? I was arguing against a “set-in-stone” 1-5, 2-6, 3-7, 4-8, but I don’t think anyone wants to see 1-2, 3-4, 5-6, 7-8. Plus if you’re going to have a bias for any group in particular, it makes sense to reward the higher rated player, even if the rating is slight and/or statistically insignificant (you still have to do something).
How are random pairings less fair than a system which always and purposefully gives higher rated players within a top-half/bottom-half sub-section of a score group tougher pairings than their lower-rated counterparts?
Now, let’s revisit my earlier case you touched upon. 99.9% of the time the 1100 is going to be legitimately worse than the 1800 or whatever else was the other rating. But it doesn’t matter, assume I meant so: this lucky player got a Class H rated player in round 1, played down in round 2 to the highest .5 which was a 1200 and got a little lucky, then took game 3 when his far-higher rated opp dropped a queen…whatever, he still is legitimately an 1100 in my example… My take, I’d 100% rather have the easier pairing. Chess is expensive enough as it is, and I’d rather have a better chance at winning something than not. If I want a tougher opponent I can meet with a club member and schedule an extra rated game. Certainly I don’t want to be competing for a prize on an unlevel playing field. With regards to the risk/reward of playing the 1100 - my personal results show that I’m 46-0-0 against E/below players since I started keeping track of such results since 1997. I’ll take those odds. But even if I lost, it’s just rating points and probably puts me at a rating I more truly deserve.
I took the class pairings/prizes as an example. I was perfectly willing for it to be completely analogous to “Top” prizes. The same argument holds, as you point out, and maybe there are some distracting elements when you mention class prizes.
Fine, I’m a chess pro named Player A competing for a major prize.
Player A = 2600, 4 points
Player B = 2550, 4 points
Player C = 2500, 4 points
Player D = 2450, 4 points
Player E = 2449, 4 points
Player F = 2350, 4 points
Player G = 2300, 4 points
Player H = 2000, 4 points
If colors and player histories have no part to play, I would be mad that I have to play the last round against a senior master, while another senior master (player D) gets to play a mere 2000 rated player. He has about double the statistical likelihood of winning than I do. Why should the system reward him? Haven’t I already been playing tougher players than him in the first four rounds?? Haven’t I worked years at getting the rating I did? For what? To be punished with a more difficult game. Even if Player D and I both win, we still have only equal claims to all prizes.
This is a Swiss System flaw that can be eradicated. It is far more fundamental and important than color history which has had much more discussion. (When competing for prizes, I would much rather play Black against Player X, than White against Player Y who is 200 points above player X.)
I’m not speaking of complete ignorance in my proposal either. I’ve played in many (hundreds?) of tournaments in my other major hobby, Magic the Gathering Online. Pairings are totally random, so if I got a harder random opponent than someone else next to me I have nothing to else to say except that “the fates decreed it.”
Please note, if you wanted to peg me in how I rank these general pairing principals, I would say I like them in this order, (but the first two much more so):
Top-Half vs Bottom Half of Score Group, random pairings
Top-Half vs. Bottom Half, NBA Style (1-8, 2-7, etc.)
Whole Score Group, random pairings
Current USCF Pairing Method: (1-5, 2-6, etc.)
Also note, that random pairings are easy to do by hand. Flip pairing cards over, and have someone select at random.
If people worry about unethical Tournament Directors as the reason to favor the current system over my #1, please note my #2. Also, there are ways to ensure ethical random pairings (witnesses would be the easiest, maybe even one that the players agree to at the start of the tournament).
If you want to talk about pairing fairness as far as PRIZES are concerned, I have to definately disagree with your #2.
I’ve played in a number of big-money class tournaments. I’ve always gotten fairly easy pairings because I’m in the top half of my class. Nearly every game, I’ve been paired DOWN. Now you want to give those at the top of the class an even BIGGER advantage?
Really, prize money isn’t that big a deal unless you’re talking about one of the major tournaments. At those things, IMHO, those in the top half already have an advantage. Now, it’s not a HUGE advantage – after all, we’re all in the same class.
I doubt you’ll get very far with your ideas. Chess players tend to be a VERY reactionary bunch and don’t like change. If there was any real demand for such a change, a big organizer like the CCA would have already tried it. But good luck. It’s nice to talk about something besides politics. Thanks for that at least.
(After I wrote this post but before I submitted it, I saw Artichoke’s reply, and realized he hit all my major points. Oh well, I will post this anyway.)
Ben, Thanks for this interesting and thoughtful post.
I am not ready to completely agree that “excitement” was the overriding goal in the design of the chess pairing system, the system you have so clearly contrasted with the alternative common format of “best plays worst” used in the playoffs by the NBA and many others.
At the level of professional sports, the overriding goal needs to be giving the “best entrants the best chance” to reach the finals. This is because huge monetary amounts are at stake. Also, the fans would protest a system that rewarded weaker entrants with enhanced chances to progress toward the finals, thereby reducing the chances for more deserving entrants (it is a zero-sum situation).
“COMPETITIVE” GOAL
In contrast, in most local chess tournaments the amount of prize money at stake is low enuf that there is room for additional goals. I think one major goal of the chess pairing system was to generate “competitive” games at the club level.
Your post focused on the financial interests of the top rated class-A player, Mr. 1999. But the interests of Mr. 1100 should matter just as much. Mr. 1100 might well prefer to play someone rated say 400 points higher than he rather than someone a hopeless 800 points higher. Mr. 1100 did not pay his entrance fee mainly to feed a system skewed to shift his money to stronger players.
ARE CLASSES UNFAIR?
Indeed, if the arguments succeed in de-legitimizing the current “competitive” pairings system, perhaps they must also de-legitimize the entire concept of class sections (else an inconsistency of thought). Class sections are terribly unfair to Mr. 1801 compared to Mr. 1799 (when the class boundary is at 1800).
This is one reason by TD’s ought to alternately shift their boundaries for classes between tournaments, by 100 points.
RANDOM PAIRINGS
The proposal to make random pairings an option (between top and bottom halves) might have merit. Over several tournaments theoretically combined, the pairings would tend toward “fairness”. But in most individual tournaments the actual pairings would likely deviate from fairness, due to the short term variations inherent in any random process. I may doubt that the theoretical fairness would feel like adequate justification to any person feeling the sting of random unfairness in his particular tournament.
PROFESSIONAL CHESS TOURNAMENTS WARRENT FAIRNESS OF R-R
At the professional chess level the “competitive” goal is defined out of existence, leaving only the “fairness” goal.
The only truly fair system is the round-robin (R-R). Unfortunately there are usually more entrants than the R-R format can handle: 8 players requires 14 rounds, too long.
PIE RULE: As an aside, the number of R-R rounds necessary could be cut in half by adoption of the “pie” rule (so in each game neither color has any inherent advantage over the other). But the pie rule proposal will never be considered until chess960 (aka Fischer Random Chess) gains a place along side traditional chess1 in the brotherhood of chess (chess960 would not suffer from the constraints of the pie rule the way chess1 would).
BEST/WORST PLAYERS DEFINED?
So the “best plays worst” pairings format might have the most legitimacy at the professional chess level.
However, it was unclear to me from your post exactly how ‘best’ and ‘worst’ are defined. Did your comments treat Mr. 1100 as worst because he had the lowest rating -or- because you were imply Mr. 1100 had thus far lost the most games during the tournament?: I hope it would be the latter (ratings should not be over-emphasized).
The majority of players play for the experience, not the money. And the overwhelming majority of teachers, organizers and others who’ve addressed this issue in writing – here, in other online forums, and most of all, in print magazines and books – say that is the way it should be.
The Swiss System is inherently a crap-shoot anyway. By tweaking the rating numbers you gave while maintaing the same rank-ordering, one can just as easily create hypothetical examples that would look eminently fair, eminently unfair, in-between, or indeterminate. And every such hypothetical would have (so far as I can see) the same probability of actually occurring, as the one that you gave.
I don’t mean to attack the Swiss System, but the enormous influence played by “lucky” or “unlucky” pairings in determining players’ final scores is well known to everyone.
Just look at all the commentary surrounding the method of choosing a U.S. Women’s Champion this year. Although the strongest woman who competed in the Championship did end up winning the title, there was much hand-wringing beforehand – fully justified, in my opinion – that sticking a bunch of lower-rated women among higher-rated men in a Swiss System and giving the Women’s Championship title to the one who emerged with the highest score (as opposed to having the women play a separate Round Robin or Swiss among themselves), made the women’s title basically a roll of the dice due to pairings. That was pretty much a consensus of informed opinion before the Championship was held, as I recall.
How does that (real, not hypothetical) situation differ from the battle for a class prize within a larger Open, Swiss System tournament? Not at all, as far as I can tell.
Again, I’m not criticizing the Swiss System. I guess it’s like democracy: despite obvious flaws, it’s the best system we’ve got.
Even a Round Robin doesn’t create a perfectly level playing field. Think about it: If Topalov is playing, he’s the only one who won’t have to face Topalov!
So, in my opinion there’d be a poor cost-benefit payoff from reforming Swiss System pairings in a way that would introduce much more complexity – making TDs into virtual slaves of their software and turning the pairing choice into what most players would perceive as an impenetrable black-box – while quite likely introducing new pairing inequities no less material (or at best, only very slightly less material) than the current ones.
Gene did a much better job that I could at explaining these points.
I’ll just make one other minor point.
Once color allocations are taken into account, it’s somewhat harder to make “random-bottom-half” or completely random pairing work by hand. I guess a computer wouldn’t have that hard a time, but doing it by hand seems hard unless its “completely random” AND # due white and black balanced.
That’s a lot of criticism for mentioning just one idea I hope you liked the rest of the post. One thing I’d like to mention. Here you talk a lot about my giving a subset players a BIGGER advantage. The problem is that currently one subset of players is already given a BIGGER advantage (namely the lower rated players of a top-half of a score group). I’m just concerned that it seems unfair to reward that group in particular. Why are you against this transfer? I also disagree with you about prizes. I have to justify spending gas, lunch, and EF at a local Saturday Swiss to my wife. It’s hard on our budget. My being able to recuperate some costs is at least somewhat proportional to my being able to go - I don’t think I’m that unique in this situation.
I’m pretty sure my use of the word “excitement” is quoted directly from the Rulebook when talking about the pairing method, but as I don’t have the rulebook in front of me, I’ll just make a tentative claim of that.
With regards to your other goal of “competitiveness” I must respectfully disagree. When I want a competitive game, I can schedule a game with a fellow club member quite easily and play a rated game. We can meet at the club or some other location. When I have money riding on the line, even if what most people would consider smaller amounts, I like to think that I have equal shot at earning it with compared to my peers. How come local tennis tournaments or your kid’s end-of-season little league tournament don’t choose to employ USCF pairing methods? (I hope not too much of an apple/orange comparison as these are elimation events.) Here, there is no money riding on the line, but they would eschew the USCF pairing method if requested. And let’s not kid ourselves, the last round is basically an “elimination” round as far as prizes go.
With regards to the interest of the lower-rated player, the “1100” from my example: the system has to inherently give advantages to some party. The NBA for example (and basically all other sports) decided to heavily weight it for the #1 seed. In an eight-player score group for the last round however, the USCF has decided to weight it towards the #4 seed (then 3,2,1,8,6,7,5 seeds in that order). At least by USCF erring on the side of supporting the higher rated player, you give people something to aim for. Right now it’s “Darn, wish I was rated five points lower…” Eventually the 1100 will get better and reap the fruits of his labors. No reason to reward him for being 1100.
With regards to “feeling the sting of random unfairness”. Given the eight-player group, and given top-half vs. bottom-half pairings, Seed #1 is by definition currently given worse pairings than #s 2-4. Now randomly, #4 may still get the easiest pairing vs #8, but at least they had both equal chance of getting that easier pairing as opposed to it being systemic. I’d rather suffer the sting of random unfairness than the sting of institutional unfairness any day.
With regards to your RR paragraph, I can agree that RR is the most inherently fair system (by a wide margin), but that is off topic.
With regards to your last paragraph, I have no idea why when you not rank the #1 rated guy as the “Best” and the lowest-rated buy as the “worst”. It’s not like we’re talking about their eternal souls or anything, just chess prowess. Besides it again seems off-topic, but perhaps you’ll just have to reclarify your intent of that paragraph with regards to pairings.
Unfortunately, I feel you did not address a single point that I brought up in my situations - is this because you agreed with them?
Again, I disagree (somewhat) with the money issue. People play to win small-stakes events. Otherwise, make the events cost $2 to play for rating fees and the TD’s lunch money. Why the extra $20-$30-$40 tacked on (assuming a free site) if just playing for experience?
I realize that unfairness will take place, always and inevitably. However, my point basically reduces itself to this - when all things line up correctly (pairing histories, colors, etc.) - why favor a lower rated group of players than a higher rated group (especially when the higher-rated player is consistently given tougher pairings every round, and not compensated with extra prize should he/she win)?
The Swiss System is not the best system we have if there are suggestions out there which can be adopted, thus improving it. Your US Woman’s example is exactly why I’m concerned. This year, the best player got lucky you say (I have no clue) with the pairings (or at least not unlucky, which is perhaps the same thing). However, if it had gone wrong, this group would be advocating for something similar to what I am saying - please provide a system that while not overtly trampling over the lower-rated player is designed to ensure that the top-rated players are not punished for being top-rated players.
You mention other pairing inequities in a vague way, but fail to mention what they are. Please expand further on this, especially ones you think that should take precedence over punishing higher-rated players with tougher pairings.
Color-allocation is a good point which I had considered from the beginnning. It’s not a knock against my proposal as long as it doesn’t hurt what I’m advocating anymore than it disrupts the current system, which would be very hard to prove. But basically, to develop the system #1 from my previous post a little further taking into account for colors, we have:
Top-half players of a score group due White randomly paired against bottom-half players of a score group due Black. Top-half players of a score group due Black randomly paired against bottom-half players of a score group due White.
May I point out another advantage that “random” has over its predecessors, is that there is a much wider pool of legal pairings available to draw from in a given score group. It’s not constrained by 80-point rules and their ilk.
When one advocates for the current pairing method over what I’m proposing, I have to wonder what the bias is against higher-rated players in favor of their lower-rated counterparts. Random pairings make a lot of sense, and I feel even high-low pairings (with slight switches for color) are fairer.
It was amazing to me when I first started playing in big class tournaments. It seemed like EVERY game was against somebody around 100 points lower rated. Nearly every round. Guys from the bottom half of the class were paired against players rated 100 points higher most of the time.
With your system the top players will have opponents that are nearly 200 points higher rated. Players from the middle of the class will have opponents of about the same strength. I just don’t see this as an improvement. (and yes, I’m from the 2nd quarter of my class – but I don’t expect to be there long so I don’t think I’m biased because of my current rating).
BTW You mentioned my “criticism”. I trust you realize that wasn’t my intent. Just stating a different opinion.
Organize a tournament, pre-advertise your pairing experiment, and let us know what happens. I personally believe that every pairing system is going to have it’s ups and downs, and that the problem is magnified when a large group of players is in a 4 round event, etc, so you don’t have a chance to let people decide the results only on the boards.
However, swiss system tournaments are a good compromise with the needs to establish pairings, and it is well understood by most players. I think that you will run into significant resistance if you try this. (I especially worry about the “random” pairing options, and the accusations of cheating.)
I have two tangental experiences relevant to this.
On the MagicTheGatheringOnline tournaments which do random swiss pairings (not even top half vs. bottom half), I sometimes rue my bad luck when the ocmputer gives me a hard matchup. But what am I going to do about it? Just some bad luck - I move on. Within the context of top-half vs. bottom-half it is already clear to me that it is better. I have already given much situational evidence as to why at least some random is a good idea. It doesn’t systematically punish players for having a good rating being the obvious highlight.
The second experience is when I wanted to run a 24-round event (to be fair a 12-round double swiss) in one day at G/10. I announce in pre-tournament publicity and aloud in the player’s conference that pairings would ignore player histories over pairing players down a score group. I was expecting a tournout in each of the two sections around 20 players, and I wanted to avoid a situation in the last rounds where the top player had to play someone he was already ten points over in tournament standing simply because he had played the nine players above him. Some people ended up playing each other six to eight times, but it was great! A couple of the participants ran the exact same tournament format the next year when I could not. (Not related, but prize payout was neat because it was based on the amount of rounds each player won - so every one earned prizes - and this was also announced.) So, if I were in the mood to organize another tournament, I’d have no problem doing the new pairing method. No person could complain about preferential treatment once everything were random.
No pairing system will eliminate every real or perceived instance of unfairness. The important thing is to have an algorithm (more or less), rather than relying on any TD’s perception of what’s “fair”, a perception which could be skewed by the “squeaky wheel gets the grease” phenomenon, or by how well the TD knows or likes the players in question.
The top-end-vs-bottom-end suggestion (e.g. with eight players you’d have 1-8, 2-7, 3-6, 4-5) would, at least, eliminate the sudden discontinuities found at the mid-point of the half-by-half method (1-5, 2-6, 3-7, 4-8) because players near the middle would be playing each other, so I suppose on this basis it could be considered.
bbentrup thanks for clarifying your question. For determining top prizes, I like your ideas. As others have noted, top prizes don’t apply only to chess professionals. They can apply to all of us.
As I see it, the Swiss system has a democratic ideal. It is supposed to let everyone play a competitive game in every round (“excitement”). But we also want it to do a fair job at elimination. These goals are in some conflict, because the first goal dictates that we avoid mismatches (they are not competitive) whereas the second goal dictates that we seek them (they eliminate guys who don’t belong in a high score group).
The sentiment here is that any Swiss generates enough excitement and inclusion, so within the constraints of the Swiss we strive for the best elimination result and that will be my purpose in the remainder of this post. But it isn’t even obvious how to implement this tradeoff once we decide on it, because the choice really is between lots of smaller mismatches (current USCF pairing method) and a few bigger mismatches (NBA style pairings).
We could test the outcome of various pairing rules by computer simulation (Monte Carlo style). Simulate a bunch of tournaments with players winning at the probability their rating suggests. Have some draws too. Rating could be updated as the tournament goes along (i.e. a type of performance rating) or not. Run 1000 simulations with each of several pairing rules and evaluate the results. The best system will have final score groups (after the last round) where the opposition strength faced by each player is the most even. (This evenness could be measured by pre-tournament rating, performance rating, or another tiebreak system of choice.)
My tentative conclusions follow. I’ve managed to disagree with almost everyone about something or other , so I’ve highlighted points of disagreement.
If the system is fair all the way down the ranks, a situation like the US Women’s Championship should not be troubling. (jonnybear says that people were troubled.) You can say the champion was the best against general strong opposition not just other women.
I prefer to avoid randomness (unlike bbentrup) because (1) it’s harder on the TD, (s)he feels an exaggerated need to demonstrate scrupulousness and is open to after-the-fact questioning, and (2) many chessplayers are seeking an escape from randomness, otherwise they might be playing bridge, poker or Fischer random chess.
Among non-random pairings (disagreeing with tanstaafl), the fairest final outcomes are probably produced by:
Top-Half vs. Bottom Half, NBA Style (1-8, 2-7, etc.)
I think this is best because it’s important to play the big mismatches (1-8, 2-7) and get rid of low ranked guys. The problem is that the middle games (3-6, 4-5) pit players who are very close and should probably end up together in the same final score group according to pretournament ratings, performance ratings, or whatever we’re using (In contrast, Smythe Dakota thinks that this is an advantage.). Usually those games are decisive anyway; draw is an unusual result. Test results from Monte Carlo simulations would be really interesting. This issue is very amenable to testing (disagreeing with tanstaafl and Robgetty); you don’t have to try it first in a million live tournaments. I think that big organizations like CCA would be interested in the results.
Artichoke, you can’t disagree with me; I said to try it and let us know the results. I’m just skeptical on it replacing traditional swiss tournaments because there is a lot of inertia to overcome in a change like this. (Just think of how much argument was generated by people when they were told that they might have to break the habit of witing a move before playing it.)
I can be disagreeable to anybody! Your idea of a test is to run an actual tournament with real chessplayers who can and will complain about anything. My idea was to push around a bunch of electrons. Then armed with a 20 page research paper (and maybe you summarize your results in an article or letter to the editor in CL), you announce that your tournament will be a Swiss but it use a particular different pairing rule.
I didn’t understand the issue about writing the move down first either, when I read about it in CL. But I never saw CL mention that some people “write” their moves into computers. Once you see that, it’s obvious why FIDE (and now USCF) would prohibit it before making the move on the board.
The biggst objection I can see to pairing up a player at random is the randomness.
It will lead to complaints that the pairings are not fair.
I could see, however, a rule which says that no player can be paired up as the odd man more than once in a tournament as long as there is another player in the same point group who has not been paired up yet.
I can circumvent your case for a small tournament - just let the top half each pick an upside-down, unmarked, shuffled pairing card. Whomever they pick, subject to a final legality of pairings check, is their opponent. (You can even number the pairing cards and roll a die, if you didn’t want them handling the pairing cards.)
All of that doesn’t work for bigger events though, huh? I guess in Tennis, in FIFA, in sports draft lotteries you have random pairings available for all to witness to see who draws whom, but they just have to do it once, not N rounds. Still, it’s not impossible to set the computer up in a public place where anyone can see, announce that pairings are about to take place, have someone hit the pairing button and read what comes off the screen (or immediately hit the Print button). I certainly wouldn’t want to put anyone under the rigor of such gymnastics, but it is possible, open, and fair.
I agree that a repeatable alorithm is the best way to ensure anti-pairing manipulation, but it is at the expense of the higher rated players as I’ve pointed out here. We know what the default view of which is the lesser evil, but I’m still not 100% sure it’s the correct one (not forgetting the third option of a high-low pairing algorithm).
I also agree to close discussion on this. Hopefully I’ve opened some eyes and planted some seeds if the system does come under overhaul, but the system is fine enough (as proven over and over throughout the years) that I’m willing to let it be.
Pairings produced over time by a random system will be an average over its various outcomes, i.e. over various possible deterministic (repeatable) systems. I can’t prove it now (I would have to write code, hook in pairing software, run simulations, etc.) but I think that there would be a deterministic rule that’s close to the performance of the random system, according to any measure of quality you like. (And for some measures of quality, a deterministic system would beat the random system.) If that turns out to be true, there would be no good reason to take on the problems that random pairings would bring.
I’m not sure that deterministic systems are “at the expense” of higher rated players. If the current USCF system is unfair to them, the NBA system would be more than fair to them and shift the burden to the guy just above the middle of the section, as I understand the issue you’re getting at.
I didn’t notice that anyone besides bbentrup proposed to close this discussion, did I miss it?
I hope you liked the rest of the post. One thing I’d like to mention. Here you talk a lot about my giving a subset players a BIGGER advantage. The problem is that currently one subset of players is already given a BIGGER advantage (namely the lower rated players of a top-half of a score group). I’m just concerned that it seems unfair to reward that group in particular. Why are you against this transfer? I also disagree with you about prizes. I have to justify spending gas, lunch, and EF at a local Saturday Swiss to my wife. It’s hard on our budget. My being able to recuperate some costs is at least somewhat proportional to my being able to go - I don’t think I’m that unique in this situation.
, so I’ve highlighted points of disagreement.
I’m just skeptical on it replacing traditional swiss tournaments because there is a lot of inertia to overcome in a change like this. (Just think of how much argument was generated by people when they were told that they might have to break the habit of witing a move before playing it.)
Your idea of a test is to run an actual tournament with real chessplayers who can and will complain about anything. My idea was to push around a bunch of electrons. Then armed with a 20 page research paper (and maybe you summarize your results in an article or letter to the editor in CL), you announce that your tournament will be a Swiss but it use a particular different pairing rule.