The king's gambit solved by computer analysis

Fredrik

I just wanted to recommend this article, which I find really cool. The king's gambit is the chess opening defined by the moves 1.e4 e5. 2.f4 exf4. It used to be a very popular opening, but fell out of favor as the top chess players got better at playing it with the black pieces.

According to a computer analysis that was recently carried out, the most popular move 3. Nf3 loses by force to 3. ...d6. Another popular move 3. Bc4 loses by force to 3. ...Nf6. The only way that white can avoid losing the game against perfect play is to play the strange-looking 3. Be2.

Fredrik

Great. Now I feel like a complete idiot. When I read the article, my first thought was that it was an April fool's joke, but I checked the date on the article and found that it was posted on April 2. So I figured it had to be true, and didn't look too closely at the details. Of course it was a joke article. Their explanation for the date on the article was that it was still April 1 (for five more minutes) in American Samoa.

johnqwertyful

Sigh. You got my hopes up. I am an avid king's gambit player. I play it now almost without fail.

As white, I almost always play E4, if they respond E5 I play F4. If they go Sicilian, my main goal is rapid castling and development. I prefer Sicilian to E5, but it's too pawn-y.

As black, I defend against E4 with Sicilian almost without fail. I defend against D4 with D5 or Queen's Indian. Either way works, it just depends on my mood. I never accept Queen's Gambit.

Dickfore

I would've asked what does the term "loses by force" mean?

arildno

That for every move White can do, there exists at least one winning strategy for Black.

I_am_learning

Now, what should I feel like then? I read your OP, went to the link, read all the article, believed everything, was fascinated* by it, and returned to PF to read further discussions on the topic and found your second post.

*I am an intermediate chess player and I was once fascinated to find out that we have now table-bases of endgames for up-to 7 pieces on board.
Although, rigorous calculation is practically possible for endgames, I should have realized that it must have been impossible for openings.
But given the link on PF, and given the site of "chessbase", and provided I didn't have enough deep knowledge, I believed the whole thing.

I seldom check dates on articles. If checked, I mostly see the year <to get a sense of when did it happen>, much less the month, and almost never the exact date.

Shouldn't they have dis-claimer somewhere, at-least after having been so long. If I hadn't read your second post, perhaps I would carry false knowledge for eternity. There isn't enough hints on the article for those who aren't literate on the field.

Fredrik

I'm glad I'm not the only one who fell for it. I think I should have seen it right away, because I'm a physics nerd who should understand the numbers involved. But it was an incredibly well-written article, I have to give them that. Both the questions and the answers were well thought out, and the pictures added credibility to it. It seemed like there was too much effort behind the article for it to be a joke.

In the follow-up article, Vas Rajlich said that the joke is completely given away by the claim that IBM let him use their super computer for four months. I guess that should have activated my BS detector. It must have been malfunctioning that day.

One thing that made it much harder to see that it was a joke was that they didn't claim to have checked every line to the end. They only claimed to have checked until the computer's position evaluation function assigned a score that's >5.12 or <-5.12. (An evaluation of +1 means that white has an advantage that corresponds to being up a pawn early in the game). It's hard to tell how much work that saves.

This is what Rajlich said about the possibility of actually solving the Kings Gambit:

It's reasonable to construct a search tree of around 10^18 positions using modern technology. The chess alpha-beta tree is thought to have at least 10^45 positions. The alpha-beta tree for the King's Gambit will be at most 10x to 100x smaller than that. So, we're still probably a good 25 or so orders of magnitude away from being able to solve something like the King's Gambit

micromass

Wait, the article was false??? I read the OP, but I never read the follow-up posts here. I guess they fooled me big time

AlephZero

Since the "value" of a chess position doesn't depend on the order of the moves which created it(*), completely solving one opening such as the king's gambit would take a big bite out of solving ANY possible chess game. Hence Rajlich's factor of "10x to 100x smaller".

(*) OK, that's not completely accurate, because each player can only castle once, the 50 move rule applies, etc, but those details don't affect the basic idea of the argument.