ChessBase.com Anti-draw Failure

.
At ChessBase.com on 2008-03-14, an essay by Kung-Ming “Victor” Tiong was published about the problem of unfought draws, known as “grandmaster draws”. Grandmaster draws occur most frequently in the last round of a tournament, when players can calculate that a draw will earn them a cash prize.
Tiong proposes the well-known Sofia and Bilbao off-the-board rules as the solution. The link:

http://www.chessbase.com/newsdetail.asp?newsid=4513

I see at least three major problems with Tiong’s essay:

[1] Tiong is wrong in claiming that the overall 60% draw rate in elite chess is no problem at all.

[2] Tiong fails to comprehend that gm draws are mostly an artifact of the high overall draw rate. Better to cure the underlying disease, instead of just bandaging one ugly symptom.

[3] Point-fiddling systems like Bilbao violate our common sense of justice. And the detailed justification they need has not yet been attempted by any of its proponents.

PROBLEM 1 speaks for itself. But because it is so easy to show, here is evidence that I am not the only person who disagrees with Tiong:

Stewart Reuben, Chairman of FIDE’s organizer’s committee:
{“The problems concerning the high percentage of chess games that conclude in draws has been exercising the minds of administrators of recent years. … [There is a] perception that a drawn chess game is somehow inferior to one with a positive result.”}

Reuben speaks of “draws” in general, not just of gm draws.
In the case of chess draws, perception is the only reality.

PROBLEM 2 has serious support from the fact that two chess-like games, shogi and Gothic Chess, are both free of any sizeable draw problems.
Why does chess have a gm draw problem when shogi (aka. Japanese chess) does not? Because the rules of shogi are engineered to make regular draws unlikely.
The same goes for Gothic Chess. G.C. games usually finish before the endgame is reached, which makes draws unlikely. Any gm draw problem in Gothic Chess could be eliminated by a simple rule requiring a minimum number of move-pairs before any draw offers; even tho that rule is ineffective in traditional chess.

PROBLEM 3 still waits for an answer as to why a player who achieves two draws has played less well than another who has one win and one loss. This is unjust. Perhaps the goal of Bilbao is to reward entertaining chess over better played chess: no thanks.
The ugly truth about Bilbao is its goal is to motivate players to make moves they judge to be semi-unsound. In other words, Bilbao wants players to make more weak refutable moves, more errors. The theory is there are unsound moves that (a) are so complicated that OTBoard refutation is unlikely, and (b) that lead to victory if unrefuted. That is a bold claim. Further, such moves are claimed to be plentiful enough that the overall high draw rate would be chopped (in half?) by adoption of Bilbao. However, Bilbao proponents are careful to never predict a specific number to describe by how much Bilbao would reduce the draw rate; because if they did we would better sense that Bilbao is nonsense.
Study the Bilbao theory yourself. Randomly pick some normal drawn games from any recent elite tournament. Then show us all the unchosen moves that Bilbao both wants the players to make instead, and that Bilbao would have the power to motivate. I doubt you will find enough such moves to justify the Bilbao hype. But until these specific moves are provided as evidence, Bilbao is no more than hype.

The inevitable appearance of cheating by collusion under Bilbao is yet another issue.


Floating in the ether is the bogus argument that a very high draw rate is logical because with perfect play Black can force a draw. But the priciple behind this argument fails when applied to other sports. Most sports have a very low or nonexistent draw rate, even though each player/team has all the resources needed to tie or not do worse than the opponent. Chess has a problem that other sports do not.

Blaming the players is always wrong. Any blame belongs to the current chess rules, and to those who control those rules yet fail to improve them.

The Sofia rule, which bans premature draw offers, is fine. But in Sofia 2005 the draw rate was still 60% (=36/60), though none were short gm draws.

BTWay, if all gm drawn games were resumed, more than 60% of them would still end in a draw.

Also, adoption of the Fischer Random Chess start positions rule would reduce the draw rate by only a small amount, from 60% down to perhaps 50%.

ANY SOLUTION?

So is any solution possible? The exorbitantly high draw rate in elite chess could reduced from 60% down to perhaps 30% only by adoption of an on-the-board rule change. But the change must retain the same “chess feeling”.
I suspect there is such a rule change option available. There is a rule that would both (a) directly add a little more piece power to the board, and (b) indirectly add piece power by reducing the rate of piece exchanges.
Shogi and Gothic Chess have tiny draw rates because they have more piece power.

But 99.8% of all tournament chess players automatically reject all on-the-board rule change proposals, sight unseen. Luckily minds were more open in 1475.
So the draw rate problem in chess will never ever be fixed.

GeneM
CastleLong.com …for FRC-chess960
.

Gothic Chess is hardly relevant, as the player/game base is too small as yet to yield meaningful stats. Do the rules of tournament-competition Shogi even permit agreed draws?

.
Here is Yasser Seirawan’s “S-Chess” idea for a chess-like game that would avoid the high draw rate problem (video posted in Jan 2008):

http://www.youtube.com/watch?v=Nht2TqabPr0

To me Seirawan’s idea is Capablanca Chess (or Gothic Chess) modified with a clever shogi twist, instead of adding more squares.

I feel Seirawan is probably taking this to an unnecessarily excessive extreme. He sacrifices too much of the “chess feeling”. He adds too much piece power to the board.
In any case, his shogi twist does…
(a) avoid distorting the 8x8 board, and
(b) correctly realizes that an increase in piece power is necessary to reduce the high draw rate in elite chess.

Seirawan’s shogi twist has the drawback of adding “invisible state” to the position. When you look at a chess game where a player has not yet castled, you cannot always be 100% certain whether the player has the right to castle: he might have moved Ke1f1, then Kf1e1, the position contains no proof either way.
The shogi twist has a much higher degree of “invisible state”. Tactical puzzle positions for S-Chess would not be sufficiently self-explanatory.


I believe a far less radical approach could chop the draw problem, while retaining the “chess feeling”.
.

“PROBLEM 1” is simply a matter of opinion; nobody can ever prove it or disprove it.

“PROBLEM 2” makes a true statement, that chess is inherently drawish, especially at the higher levels.

“PROBLEM 3” is again just a matter of opinion.

So the only way to attack the “problem” is to radically alter the rules of chess. Does anybody really want to do this?

Bill Smythe

.

Not to be argumentative, but I must disagree. I believe comparisons between chess and Gothic Chess are informative.

And by comparing computer vs. computer games in chess --with-- computer vs. computer games in Gothic Chess, the contrasting results are obvious. Gothic Chess games end in a middlegame checkmate far more often than chess games do. If a game does not reach the endgame phase, it ain’t very drawish.

Besides, just from fiddling with Gothic Chess the intuition senses the draw rate must be lower than in chess. It is hard to contain those two additional powerful pieces (that Capablanca evangelized). And the rate of exchange sacrifices (and the resulting anti-draw imbalances) is certainly higher in Gothic Chess.

Even though shogi is similar to chess, the draw rate in elite shogi is under 2%.
So why is it 60% in chess?

:question:
Would the shogi community be uphappy if their draw rate ballooned to 60% over the next decade?
Would the chess community be unhappy if their draw rate collapsed down to 2% over the next decade?
.

You didn’t answer the question. The vast majority of draws in chess are agreed draws. (How many games end in stalemate or bare King?) Do the rules of Shogi (in tournament competition, the only meaningful comparison) allow the players to agree to a draw? I agree that there are other factors in the structure of Shogi making it less drawish, but without more information it’s hard to tell which ones are most important.

In answer to your question, I’m sure many people would be happy if the percentage of draws went way down without changing the rules of the game. Would they be happy with a significant change in the rules? Well, that depends.

I read a letter to the editor in Chess Life within the past year or so that suggested changing the rules to make it a loss to repeat a position a third time. This is a minor change compared to adding new pieces. It adds new aspects to opposition and zugzwang that probably increase the chances of turning a slight advantage into a full point.

There’s a similar rule in Go; you simply are not allowed to repeat the position. While it would be nice to see more fighting games, all these suggestions strike me as using a nuke to kill mice.

This would be a huge change. For example, if a player is in check, and his only move out of check repeats the position, he would immediately lose because he has no “legal” moves. An opponent could force a win just by forcing a perpetual check.

Bill Smythe

I would think that, usually, it would be the person doing the checking who would first repeat the position three times.

In that case, a player who had a forced perpet in an otherwise losing position would lose.

Maybe that’s what you want.

Bill Smythe

That’s the minor problem with the rule. The major problems are that it destroys all endgame knowledge (for example, King and Pawn against King is now almost always a win). As a side-effect, no more speculative Pawn sacrifices–they’ll be far too risky: if you wind up a pawn down in an endgame, the game is essentially over–you will eventually lose when the no-repetition rule forces you to give ground.

Even worse, will lead to the dullest positions being routinely played out, probably for hundreds of moves. In opposite-colored Bishop engames (for example) the players would endlessly jockey for position, to see if they could get the opponent into a state where he has to repeat a position or give up ground.

In other words, imagine every game being Kramnik vs Kramnik. Except more tedious.

I really don’t understand the anti-draw people. We don’t need more decisive games, we need more exciting games. The way to do that is to reward interesting play: more brilliancy prizes and best-played game prizes; more invitations to the big-money round-robins to the players who play interesting games, and fewer to high-rated dullards. Throw in the Sofia draw rules, and problem solved. (For that matter, the Sofia rules seem to work quite well by themselves.)

They’re also free of any sizable world-wide fan-base. A vastly greater number of people all over the world play Chess. Why is that, do you suppose?

As best I can understand, the answer is – “No, the shogi players cannot agree to a draw”. They can ask the arbiter to declare the game as a hopeless draw, but such conditions rarely occur (meaning soon enuf one player achieves a decisive advantage, and he would not accept a draw offer).

So the Sofia rule in shogi keeps the draw rate near 2%. In contrast, in the Sofia chess tournament 2005, the draw rate was still 60%.

That is why I wrote that the chess unfought draw problem is merely an artifact of the deeper overall high draw rate problem.
Shogi has no unfought draw problem because shogi has no general draw rate problem. In elite chess matches and tournaments, the complete elimination of all unfought draws would still leave the same high draw rate unfixed.

I agree, “it depends”. But I think it depends on who is being asked.

PROFESSIONAL GRANDMASTERS:
I feel confident that the mass of financially precarious grandmasters rated 2500-2699 would flock to any well-funded chess tournament that played with a minimally invasive on-the-board rule change. They would certainly have the blessings of their families they support.

Evidence? With the conspicuous exceptions of Kramnik and Topalov, the majority of the world’s best chess players compete in the ChessTigers.de annual Fischer Random Chess (FRC, chess960) tournament. This example is a little unfair because FRC feeling exactly like chess after about 6-7 move-pairs; but still there is an initial shock at move 1. Also, the players report that they enjoy FRC.
FRC by itself would reduce the draw from 60% down to only 50%; not good enuf.

CLUB AMATEURS:
Any rule that substantially even slightly alters the “chess feeling” would automatically face nearly universal heavy opposition from the amateur chess community; even before they heard the description of the rule.

DILEMMA:
The dilemma is that while off-the-board rule changes leave the “chess feeling” untouched, it is no coincidence that they also fail to fix (or even dent) the high draw rate problem.
An innumerable variety of on-the-board rule changes would fix the high draw rate problem, but all with the drawback that they would alter the “chess feeling”; at least a little bit.

Would a huge reduction in the draw rate be worth tolerating a slight change to the chess feeling? Could a lopsided trade-off ever be worth considering?


RULE ENGINEERING CHALLENGE:

What is the least invasive on-the-board rule change that would truly cut the high draw rate in half?

This is a worthwhile academic question even if we never want to consider making any on-the-board rule change in live chess play. The investigation will teach us things.
.

Aren’t you loading the question? In order to get a a “huge” reduction in the draw rate you would need a quite significant change in the rules. A more realistic question would be whether a small reduction in the draw rate would be worth a small rules change. Since my answer to that is “almost certainly not,” there’s not much point in discussing specifics.

Worse still, what about something like K+R vs K+R, if triple occurrence and draw by agreement were not allowed? Both players would be forced to play on for hundreds of moves, until one of them got into a position where there were no legal moves, because every otherwise-legal move would create a third occurrence?

By the way, in that situation, would the player with no legal moves lose, or would it be a stalemate?

Bill Smythe

Well, that depends on whether the 50-move rule would still apply. Of course, if it is, that would give players a way to collude on draws. So we’d get all the disadvantages of destroying endgame theory and warping the game, without actually getting rid of fightless draws.

Anyway, various R+Ps vs R+Ps endgames, and other even endings with pawns, would be subject to the problem. I suspect the proportion of 100+ move games would go from less than 1% to 40% or even more. It would be very exciting.

I assume he’d have to lose, right? Otherise the no-repetition rule would be kind of pointless. But perhaps someone who advocates such radical rule changes should answer.

I don’t think there’s a good answer for this. Changes to the basic draw rules (stalemate, repetition, insufficient material) would do it, but are unacceptibly invasive.

I wonder how much the draw rate would go down if you eliminated castling?

I think that off-board changes would actually be more effective. The Sofia rules are a good start. (would you please stop citing only the year with the highest draw-rate at Sofia, by way?) In fact, they’re a sine qua non; if you don’t eliminate most agreed draws, nothing else can help.

An idea that hasn’t been explored anough is replaying drawn games at faster time controls until a decision is reached. There seems to be a tendency to want to use pairs of rapid games to eliminate the color advantage, and that’s a mistake: playoffs should be sudden-death. If you have White in the “regular” game, you will be very reluctant to draw if it means you have to defend as Black in the first sudden-death game. Changing the scoring rules so that either White gets less from a drawn than Black, or the higher-rated player less than the lower-rated, might work as well as playoffs. And, of course, prizes for decisive games, for brilliancies, for upset wins, and so on would do a bit.

One other thing to bear in mind is that the draw rate in games between the best players has always been high. As far as I can tell, if you look at tournament and match games from any decade, and consider only games between the elite players, they always draw over 50% of their games against each other. The difference is that, until the 1980s or so, most tournaments had a lot of “outsiders”–people like Suchting, or Chajes, or Donner–and the top players had to beat them to have a chance to win the tournament. Also, the outsiders would tend to play more decisive game among themselves.
I suspect that if organizers went back to holding larger round-robins not limited to the 2700+ club, the draw rate would go below 50% just from that.

What’s needed are sponsors and organizers who want to try some ideas out. As long as they run elite events that encourage draws, there will be a lot of draws.

At this point the discussion becomes an interesting intellectual exercise. So, let’s talk away, without seriously proposing any of this in practice. (We chess players are capable of intellectual discussions, aren’t we?)

How about the following rules:

  1. A draw can come about only through stalemate.

  2. Draws by agreement, triple occurrence, and the 50-move rule are not permitted.

  3. No player is allowed to play a move that creates a third occurrence of the same position (with the same player to move each time).

  4. If a player has no legal move, except for one (or more) that creates a third occurrence, the game is over. In that case, the result is a stalemate (draw), unless the player is in check, in which case the result is checkmate (loss).

  5. If a “dead position” arises (one in which no checkmate is ever possible, even using the new definition of checkmate as in 4. above, and even with the cooperation of both players), the game is drawn.

Sample dead positions in “normal” chess include K vs K, K+N vs K, and K+B vs K. Of these, however, only K vs K is a dead position under the proposed rules. For example, with K+B vs K, it would be theoretically possible for a player to be in check, and not have a way out without repeating the position.

This leads to our first question in this intellectual discussion. Is K+B vs K always a forced win (with best play by both sides), or is it always a forced draw (again with best play), or does it depend on the position? As a first step, come up with a K+B vs K position where white has a forced win beginning with a non-checking move.

Bill Smythe

Easy. White: Bd4, Kg6. Black: Kg8. White to play: 1. Bc5 Kh8 2. Bd4+ Kg8 (second occurance) 3. Bc5 Kh8 4. Bd4+. Now 4…Kg8 is the third occurance, so Black loses under the rules you defined.

I’m probably going to drop this now, since this exercise is not that much fun for me.