What’chu talkin’ ’bout, Willis?

[I seem to get a good number of readers, but not so many comments. This is one post where I really would appreciate some feedback. Apologies for the length … I bolded the most important parts!]

The Candidates Matches for the current Championship Cycle ended a couple days ago, and Boris Gelfand came out on top.

I somehow managed to guess Grischuk – Gelfand for the final match, but in the final, I had Grischuk advancing. Gelfand was a worthy winner, of course, but somehow, I think that Grischuk would have proved a more difficult opponent for Anand.

However, the matches have drawn a lot of criticism, some of it reasonable, some much less so in my view. There were two common threads amongst the unreasonable complaints in my view.

The first is that the winner was undeserving. It’s hard to imagine a chump coming out of these Candidates Matches, but because Gelfand was not a favorite (I certainly did not expect him to win it, even up to the final game), that doesn’t mean he was undeserving.

Aronian was generally considered the favorite before the event, but after failing to win an easily won endgame in Game 1 against Grischuk, he wasn’t able to break through later on, and then lost in the rapid tiebreaks, 2.5-1.5. Nothing fishy, he lost under the rules of the system.

The second is that these Matches were “boring” because of the high draw rate. With only 3 wins in 30 classical games, the 90% draw rate trumps even the 87% draw-rate in the “snooze-fest” that was Kasparov – Kramnik, London 2000. Nobody was holding me hostage in either event.

The chief offender here was apparently Grischuk because of some short draws with the White pieces. GM Moradiabadi, in commenting for ChessBase, apparently called some of his games disgusting (later revised to disappointing). Of course, Elshan’s handful of draws as white under 10 moves in the past decade are probably not as bad.

I’m not so surprised that 90% of the games ended in a draw. It’s very difficult to beat a strong chessplayer, and these guys are incredibly strong chessplayers. Grischuk saved some tough positions against Aronian and Kramnik, so should we knock him for being good enough to save those positions that most would have lost?

One suggestion was to extend the length of these matches. I don’t like that for a couple reasons. Match play has always been relatively cautious – it’s not because it’s 4 games that people are relatively cautious, it’s because it’s to get to the World Championship. If it was the length of the match, then did players think that rapid and blitz chess were somehow more controllable than a slow game?

Commentators like GM Sergey Shipov have also argued that longer matches (say 8 games matches at every round, already too much in my view with tiebreaks possibly to boot) might well have produced no more decisive games. As an example, despite Kasparov’s hefty plus score against Anand overall, they had many more draws than decisive games and had a number of stretches with more than 4 draws in a row. Meanwhile, in previous Candidate Match cycles, there were matches with double-digit games and only one decisive effort. Maybe we were actually saved from a lot more insipid chess.

So what can we do? Is there a system that might be more equitable, more interesting, or both? Giving Grischuk the mic:

“… [I]f you go back to the origins of competitions, the Olympics in Ancient Greece, and so on … [t]hey started off in order to identify the strongest person….”

So how do we do identify the strongest Candidate? Hard at work the other day, I had a little idea that I thought was interesting. I floated it by a few chessplayers, including a couple other GMs, and all of them seemed to think it made sense.

A minority of fans would like to see a return to the Candidates tournament, a la Zurich 1953. (As an aside, a lot of people like to talk about the good ol’ days with Candidates Matches, but that framework was not always the case!)  Charges of collusion have helped drive the shift away from double round-robins as the single qualifier, but I don’t think that’s as much of an issue anymore.

There are two issues I see with the Candidates event though:

(1)   It is quite possible for a the winner of a double round-robin to NOT have beaten his closest competitors (those in 2nd or 3rd place, for example). Won’t we just hear more about an “undeserving” winner in that case? For example, Carlsen won the London Chess Classic by losing to the next two finishers, and Nakamura won Corus with -1 against the rest of the top 5 finishers.

(2)   Another thorny issue is that of tiebreaks from a single event. Will a win/loss prove the difference ahead of two draws? Will an extra win with black outweight an extra win with White? But what if the win with Black was against the tail-ender?!

The match framework can solve that to some extent – to advance, you actually need to beat all your opponents. So I see some value in having the end stages of the Cycle built upon matches.

From an aesthetic point of view (and that of tradition), matches also differentiate the Candidates Cycle from – and possibly elevate them above – the standard round-robin that dominates the super-GM tournament circuit, all the while honoring the general tradition of having matches decide the World Championship.

Moreover, from what I’ve seen, while there are diehard match aficionados, the promoters of the Candidates tournament idea are not particularly averse to the idea of matches (if executed well). So a solution involving matches would seem to satisfy more people than any alternative.

So the first thing I’m going to work with is that matches will play a role at the end of the Candidates Cycle.

Still, while the match framework solves the direct competition factor, what about tiebreaks?

The most discussed example of a drawn match from the “golden age” of Candidates Matches was Huebner-Smyslov 1983, where the spin of a roulette wheel decided the outcome. (No, really – it did!) However, there was also Spassky-Portisch 1981, where the score was tied after 14 games (1 win apiece, 12 draws). Portisch advanced as he had won his one game as black …

In neither case am I satisfied with the advancing mechanism.

How about the current system with rapid and (if necessary) blitz tiebreaks? This makes no sense to me, as while the pieces move the same, those are rather different forms of chess. Why should this form decide the Challenger for the Classical title? There is a separate World Blitz Championship after all.

So, the second thing I’ll work with is that the Challenger for the World Championship (and the Championship itself) should be determined by classical play, not with quick-chess as the final tiebreak.

One idea (which has been tried before) is to play “First to X wins.” I even saw one post where this was claimed to be “the best that chess can offer in a practical world.”

I can’t see this being the best solution “in a practical world.” For starters, how will the logistics of consecutive matches ever be solved in advance? Who will sponsor Candidates matches that have no limit on the number of games? Couldn’t we still have a huge string of boring draws?

Instead, I think there’s an idea to satisfy all these conditions so far. Why not bring back draw odds?

Not with draw odds for the higher-rated player, but with draw odds for the player who does better in the tournament phase of the Candidates Cycle.

[View http://www.youtube.com/watch?v=Qw9oX-kZ_9k for some comic relief]

In my view, this seems to make more sense than a match system where all the previous results are thrown out. While the lower-seeded Candidate has shown the great ability to make it that far, with somebody having performed better, why should that person get the shot for the classical title simply because s/he is better at rapid or blitz?

The mechanics of the seeding could work something like this:

Seeds 1-4: Top 4 finishers from the combined Grand Prix and World Cup standings

Seeds 5-7: Top 3 by rating (not including the Grand Prix finalists and World Champ)

Seed 8: One host-country nominee

A portion of the Grand Prix would be like it was for this cycle – a series of strong round-robins amongst top 2700+ players, with points awarded based on finishes. I would incorporate the World Cup into the Grand Prix standings, but with a little extra value. So in the past cycle, a Grand Prix win was worth 180 points in the standings. Maybe the World Cup would be 240 points. It probably wouldn’t be enough to automatically qualify, but it would be worth quite a bit. The Grand Prix series would be able to accommodate pretty much all of the top 20 or 30 (depending on the number of events), while the World Cup would bring in the rest of the top 100.

In the match phase then, the top seed has draw-odds through all 3 rounds of Candidates Matches. The first two rounds of matches features 5 games, where the lower seed gets the extra White. The third and final round is 7 games, where the lower seed again gets the extra White.

My original idea was actually to have all the 3 rounds be 6 games, but on discussion with a couple GMs, draw-odds seemed to be too big an advantage then. With the extra white however, the lower-seed has an extra shot to prove that s/he should actually advance.

Essentially, the draw-odds factor puts the onus on the lower Candidate to show why s/he should leapfrog somebody who has had better performances over the duration of the Cycle. Then, if the match ends up in a tie, the lower Candidate hasn’t overcome the worse tournament performance over the Cycle and drops out. The match-phase is not on equal footing for all 8 players, as doing so completely throws out past results in the Cycle and then introduces the thorny question of how do you come up with a reasonable tiebreak when a draw is quite normal in the game itself.

In the end, I want to reward participation and classical performance – not playing and/or not playing as well as others should put you at a disadvantage for the later stages – and I think this setup does that to a better extent than the current system. There is a lengthy qualification process; matches with a real imperative to fight; and no rapid/blitz tiebreaks.

If the Challenger is one who had the draw-odds advantage, well – he earned it by winning in the earlier phases and then nobody showing superiority over him after that. If the Challenger overcomes the draw-odds factor, I can’t see anybody begrudging him/her the spot because he BEAT his competitors in classical play, not in rapid or blitz play.

There are a few side-benefits I can see as well. By rank-ordering the Candidates based on tournament results and not their rating, there is an incentive for the top guys to actually participate in the tournament phase of the Cycle. Thus, when the FIDE Grand Prix and World Cup roll around, guys like Kramnik and Topalov won’t necessarily rely on their high ratings to qualify for the Candidates Cycle. And with the highest-rated players likely to play, there might be more sponsorship and more varied locales (not that there’s anything wrong with towns in former Soviet republics …).

The importance of the tournament phase also could promote freer and more open play – when the difference between 2nd and 3rd place in the tournament phase is potentially draw odds in the quarterfinal match phase, there is more of an incentive for players to play to win, as now you don’t just get into the matches from the tournament play, but you are actually seeded based on your results.

Giving the lowest Candidate seeding to the “host country” also could avoid a situation that arose in the last FIDE Grand Prix, namely a game in which some (including at least one participant I think) suggested Mamedyarov had thrown his 11th round game to Radjabov. That was a huge boost to Radjabov’s qualifying chances, while Mamedyarov, who wasn’t in the Grand Prix chase anymore, was still safe with a host-country wildcard available. I also see no reason that a failure to qualify from the normal routes should be rewarded by starting on equal footing with the World Cup winner, for example. (He did lose in a normal fashion to Gelfand anyways.)

Finally, compared to “anti-draw” rules like the Sofia rules, the draw-odds factor keeps the players more honest. In the case of Sofia rules, there are easy ways to get around the spirit of the rule, and there is no penalty on the players doing so. With draw odds though, there is a real incentive to play everything out, honoring the spirit of the Sofia rules.

If you’ve made it this far, congrats – and what do you think? The basic idea came to me just before lunch, so I might easily have overlooked something obvious.


18 responses to “What’chu talkin’ ’bout, Willis?

  1. Hello,

    Very reasonable idea, definitely an improvement on the current state of affairs. Another interesting proposal floating around in the chess blogosphere these days is playing the rapid/blitz tiebreaks BEFORE the classical games, so that the winner gets predetermined draw odds. Thus, players will have an added incentive to push harder if they lose this pre-tiebreak.

    I think the two options are somewhat similar in spirit and the choice between them is a matter of taste. If you want to completely banish quickplay from Candidate Matches, your’s is the way to go.

    It will be interesting to see what FIDE thinks about this.

    • Yeah, I’d rather see quickplay removed from the equation as much as possible (it’ll still be there for the World Cup though). I saw the blitz/rapid first idea as well, and it is in the same vein to some extent, but the problem I see is that the better slow-play Candidate might well be different from the better fast-play Candidate.

    • GM Bhat, I really appreciate your blog.
      Its sad that being one of the best in the world at chess, isnt a guaranteed source of income, but appreciate your dedication and love of the game of chess.
      Keep up good work on the blog.

  2. Interesting idea. I like the fact that the top seeds come from tournament results (otherwise I was thinking you might get some weird results where a player doesn’t care about the tournaments anymore because his rating is already high enough, thus giving him the power to determine whether another player is going to come from the tournament pool or rating pool – but hopefully the higher seeding would be incentive enough to not “game the system” that way).

    “One player gets draw odds, the other player gets an extra White” seems dangerous to me, though. With numerical stability you run into problems if you try to produce a small number by subtracting one large number from another, and I feel that’s what you are doing here: taking the two large advantages of 1) draw odds and 2) an extra White in a short match and hoping that when you take the difference the resulting small advantage is of the right magnitude. For example, you could easily imagine that either one of those advantages could be markedly different based on individuals’ playing styles.

    Playing the tiebreaks before the classical games is a very interesting idea that’s new to me. Do you just play rapids and then blitz to break that tie, or do you do the whole thing in reverse (play blitz to determine your rapid tiebreak, then play rapid to determine your classical tiebreak, then play classical)?

    I’m not sure that you really need to give an advantage to the better seed at all, although then you run into the issue I mentioned in the first paragraph, where a player is already assured of entry and can try to manipulate the rest of the field.

    My current favorite tiebreak system is the US Championship time-odds vs draw-odds one where the players get to bid on which side they want. Maybe that feels too artificial for a world championship cycle, though.

    • Well, I’m not actually trying to get equal win-probabilities for the high/low seed. My original idea was just use the Cycle’s tournament phase to determine draw-odds, but that seemed to be too big an edge. The idea of an extra white for the lower-seed is only to mitigate that, not equalize.

      It’s different than the 5-4 or 6-5 blitz armageddon discussion, where you have an arbitrary coin-flip to see who gets draw odds. In that case, you are trying to equalize win probabilities, so it’s important to know which time control is most likely to do that.

      • Right, I realize that you are still trying to create an advantage for one side, but creating a small advantage by giving each side a different large advantage is very “noisy” and hard to tune. It’s like trying to move one step left by taking ten steps to your right and then eleven steps to your left.

      • That’s fair. I don’t really know what the win probabilities are for a match with draw-odds, and then a match with an extra white, so I am guessing that I “tuned” it to be about right.

        At least with draw-odds matches, I could probably go back to look and see how players did (may still not be a meaningful sample), but I’m not aware of any matches with an extra white game.

      • New post at https://vbhat.wordpress.com/2011/06/05/simulated-candidates-matches looks at some simulated results. It’s a couple-percentage points higher than I expected, but given some simplifying assumptions, it’s generally about 60-40 for the higher-seed in 5- and 7-game matches. Doesn’t seem to be such an insurmountable advantage to me.

    • Also, the blitz/rapid idea I think (as I saw “kxm” explain it elsewhere), was to play blitz first (pairs of games until a winner emerges), then rapid (if drawn, blitz is the final tiebreak), and then the slow chess.

      The US Champ system of “I cut, you choose” is theoretically good, but they use 60-minutes for the max time. Speeding things up again, as most of these games feature 60-20 in time odds. It’s an interesting system, but I feel that it’s more of a crapshoot when you could draw on tournament performances instead.

  3. Interesting thoughts. I’d like to see the removal of blitz, and think the 4-4-6 routine only encouraged that – a 6-QF, 6-SF, 8-game final would’ve made it harder for some matches to be determined by blitz. Grischuk’s comment about determining the strongest and then playing for blitz summed up everything that was wrong for me with the Candidates (I wrote about this in a bit of a rant here). I like the thought of a 6-6-8 format, with your ideas of determining progression without blitz playoffs. The qualification proceedure probably needs over-hauling too, and your ‘tweaks’ make sense to me.

    • I’m not sure 4-4-6 encouraged blitz too much – with Grischuk in the rapids, maybe so. He wasn’t getting anywhere against Kramnik, so why bother? A lot of other top guys, including Kasparov, have failed at that task.

      In addition, don’t forget that Kramnik had a number of whites to break through against him. He didn’t. It’s not like Grischuk just beat Aronian and Kramnik in quickplay, which is what I see a lot of people summarizing this as. He drew 8 games with them, and then beat them in quickplay.

      For example, I might have a plus score against some 2750+ GMs in blitz, but it doesn’t mean I’m going to survive 8 games against them to get there.

      Anyways, if you dislike the system, maybe you should send Grischuk some thanks for exposing its flaws! It was only after the debacle of Curacao 1962 that they scrapped the double RR and instituted matchplay to determine the Challenger.

  4. It depends what your goals are. As a very strong chess player, your focus is naturally on the quality and integrity of the game. My connection to chess is mostly as a spectator (despite our mutual lifetime score – I’m just sayin’), so I ask myself “what makes good entertainment? What can lift chess up into the ranks of basketball and poker as a popular spectator and eventually participatory event?”

    These changes are not it. If we want chess to be more than a novelty act in the cultural pantheon, games will have to be shorter and much more decisive. Yes, “quality” will suffer. But if done right, many, many more people would watch and play.

    I realize that many of today’s chess players may not care whether their game is popular with the general public. That’s fine. Then be prepared to give up hope of corporate sponsorships and resign yourself to Soviet-bloc tournaments forevermore.

    To succeed commercially, it’s not about the “product.” It’s never about the product. It’s all about the branding and marketing.

    • Haha, spoken like a true MBA!

      That’s true, I am focusing more on quality/integrity with this system above entertainment value. Draw-odds would inject a little more life into the games/matches, and the qualifying tourneys themselves could be more interesting. Participation of the top guys, as opposed to lesser 2700 GMs, does seem to get more interest.

      Branding and marketing are key for commercial success, and speeding up time controls with more decisive games is probably the only way to market it properly to a much wider audience. My changes are definitely incremental compared to a complete shift to quickplay.

    • That’s like saying a classical musician should give up her violin for pop music, just because there is more money to be made there right?

      Myself, and probably many other fans of the the game, want to see the world’s best players play the game to the highest standards of human ability.

      I don’t want to see a bunch of players hang their pieces all over the place because they only have 2 seconds per move. If I wanted to see that, I can just replay some of my games.

  5. Both the suggestion of playing tie breaks first and draw odds are very interesting. I also like Kramnik’s suggestion of raising number of games to 6 and 8 from 4 and 6.
    The candidates in a tournament format seems to have a history of collusion problems. In the old days there were too many instances of unfair play (Keres throwing to Botvinnik in 1946, Smyslov winning 1953 with Soviets ensuring that Reshevsky had tough games, Petrosian in 1963 over Keres, etc). If a group of players managed to get into a candidates with vested interests, this can happen again.
    A variation of draw odds could be to have the player with better results decide what he wants – the extra White or Draw odds. Someone like Topalov may go for the extra White and Kramnik might settle for the draw.

  6. Pingback: Simulated Candidates Matches | There and Back Again

  7. Pingback: Oh, the humanity! | There and Back Again

  8. Playing tie-breaks first seems like a very interesting proposition and guarantees a fighting match. I don’t think blitz is necessary though. I would go for a 6-games match at 25 10 time controls, and an armageddon rapid game where people would have to bet from 1 to 25 minutes with 10 sec. increments.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s