Hacker News Reader

Comment by conistonwater | Hacker News Reader

This time around there was an unusual amount of match strategy involved. In fact, what's interesting is that they always played the Spanish or the Italian, but in the one game where Karjakin wanted to (had to) win, he played the Najdorf.

Peter Svidler mentioned a very interesting suggestion that the players should play the tie-breaks before the match, which should produce more interesting games (in particular, it should avoid game-12-style games). But in any case, I think determining the structure of the championship match is a rather neat test-case of mechanism design.

View on HN

I'd like to see tiebreaks first in soccer.

Yasser Seirawan had an interesting suggestion. Make championship matches 13 games, not 12, with the challenger getting white 7 times and the defending champion getting white 6 times, and with the championing retaining the title in the case of a 6.5 - 6.5 tie.

One nice thing about Seirawan's suggestion is that it does not involve rapid or blitz. Many people don't like the idea that the world championship of classical chess might be decided by rapid or blitz games. (There are separate championships for rapid and blitz, held annually).

As computers get better and better at chess, I wonder if it might become reasonable someday to use computer evaluations to break ties? There has been work on using computer evaluations to measure how accurate and strong chess players are. Computer scientist and IM Ken Regan has done a lot in this area, both in the context of trying to determine how strong past players actually were, and in the context of trying to determine when someone is using a computer to cheat.

For instance, in this match Carlsen had winning positions in some of the drawn games, but was not able to convert. I don't think that in any of the drawn games Karjakin ever had a winning position. I suspect that if polled, most GMs would say that Carlsen outplayed Karjakin during the classical portion of the match. With computers, that could be put on an objective basis.

Svidler is an interesting commentator. He has a quick and sharp wit. During one of the earlier games, when Eric Hansen was co-commentator [1], they were talking about IM Lawrence Trent (Fabiano Caruana's manager). I didn't catch all of it as I was listening while working, but apparently Carlsen was playing an opening line or something that Trent doesn't think is good. Hansen said something like "I wouldn't want to be Lawrence Trent, betting against Magnus Carlsen". Svidler immediately shot back, "That sentence was too long".

[1] This was on chess24.com. Svidler and Hansen were commenting for them, not for the official broadcast. For the later part of the match, the chess24 commentary was Svidler and German GM Jan Gustafsson. Both the Svidler/Hansen and the Svidler/Gustafsson pairings were quite good.

Computer evaluation would cut off all the "human" factors involving in a match, especially championship one. It is part of any sport that being slightly better is not enough, you have to be able to turn that into concrete advantage over the scores. To put it another way, "match strategy" is part of the game.

We know with reasonable certainty that Carlsen is the stronger player, thanks to the rating system. Championship match has always be about more than being _just_ the stronger player.

The point is that Carlsen was ahead, so if enough games had been played, he eventually would have managed to convert one of those games to victory (by sheer probability).

Because the number of games played was too short, the game score wasn't allowed to reflect the true strength of the players.

Yes but that's a feature of almost every sport (to varying degrees). The aim of championships are not necessarily to determine who is stronger.

I think in chess the purpose of the match is to determine who is stronger.

If all you want to know is who is the strongest player overall, then we don't need these special championship matches, we'd just look at the overall rankings.

Being able to perform "on the day" is an integral part of every sport championship and the thing that makes them exciting.

Overall rankings don't tell you that. They also involve things like "who's been trying really hard lately"

I don't like the idea of computer evaluation at all. The measure of whether one has a winning position should be winning.

I agree but here's an anecdote. Chess tournaments have been decided by a third party. I recall the story of a match in Holland between two clubs, during which someone died of a heart attack. The clocks were stopped and never resumed; whether out of respect or distress (or both) the remaining participants did not want to continue the match, so the game positions were recorded, and a GM subsequently called in by the Dutch association to evaluate the positions and decide the result.

(I suspect no violation of the FIDE laws of chess, which concern themselves primarily with the conduct and outcome of individual games and have little to say on the deciding of matches).

Note also that computer evaluations of the strength of a position do not necessarily reflect the ability of a human being to play it. Not merely due to move depth and complexity (like when Stockfish comes back with a hilarious "mate in 24"), but also time: I recall moments in Carlsen vs Karjakin where the engines judged SK to hold a positional advantage despite mere seconds remaining on the clock.

> I don't think that in any of the drawn games Karjakin ever had a winning position.

Karjakin had a winning position in game 9, but did not find the key move Qc4. He analyzed this line, but did not see Bg8 and thought it's not winning.

As Greg Shahade pointed out[1], Kasparov had to beat Karpov to win the title, as did all the other world champions in the 20th century. He argues bringing back draw odds for the defending champ would lead to more interesting matches.

[1] https://gregshahade.wordpress.com/2016/11/28/how-to-make-the...

> Peter Svidler mentioned a very interesting suggestion that the players should play the tie-breaks before the match, which should produce more interesting games.

This is the first such suggestion I've heard that I actually like.

Since 2003, MLB determines home field advantage in the World Series by the winner of the All–Star Game. Not same but similar.

Interestingly, in the new CBA which was agreed to last night, they have removed this.

HFA in the WS might not be an advantage though. I'd argue it has actually been an advantage for the NL team if the AL team gets HFA in recent history because it helped their hitting. The last WS would be a good example.

The MLB system isn't similar at all.

Home field advantage in baseball is small and no certain road to victory and there's no specific series/game situation in which it would come into play in determining the result. To wit, the last World Series featured home team wins in just 2 of 7 games.

And the result of the All-Star Game is determined largely by the performance of players who will not feature in the World Series: at most 1/8th of the rosters are World Series participants, often much less. Further, the AllStar game is not managed with an eye towards maximizing the chances of victory, but instead to maximize the number of players to appear in the game; and pitchers that pitched two days before the game are not included, regardless of their skill level.

Both of the suggestions "Champion has draw odds" and "tiebreaks before the match" means that one player can just go on the defence for the duration of the match.

The only way to get both players to attack, that I've heard, is the classic "first to 6 wins" or some other number. In that version, no side ever wants to draw.

Unless you're 5-0 down, so you just aim to draw repeatedly until it's declared a stalemate.

No, that's the point: The game continues until a player reaches 6 wins. If you're 5-0 down, you can either resign or play for 6 wins. Draws don't help you.

But if you keep forcing draws, it doesn't let your opponent reach 6 wins.

Sure, but it can't be your strategy, since then you can never win either.

A draw is half a point. So with first to 6.5, you win the match if you have one game win and 11 draws.

I'm talking about first to 6 wins. Like the 1984 Karpov vs. Kasparov: http://www.chessgames.com/perl/chess.pl?tid=55015

Ah, of course. Sorry for the mistake.

What if you just have 13 draws? If you award any kind of points for draws, it seems like you're back to people aiming to draw.

I fear that the winner of that pre-match tie-break would play as defensive as he could during the match leading to even more defensive matches.

> the players should play the tie-breaks before the match

But then wouldn't the game before the 'tie break', essentially become the tie break in pressure/strategy etc?

Maybe I'm missing something, but I think not: the first point of it is that it will never happen that the match is even, someone will always be leading. So it affects even the very first game, unlike in the current set up, where the first game (Trompowsky?!) is not a real game. At the beginning of the first game, someone would already have to play like they are behind. It also means that a you (I think) don't get situations where a draw is "good enough", a draw always solidifies someone's lead. That said there might be other consequences that I don't see.

Reply on news.ycombinator.com