Hacker News Reader

Comment by ubernostrum | Hacker News Reader

We now have a little more data courtesy of chess.com's computer tournament.

Starting from the final position of Game 12, the engines have played 18 games to completion against each other, with more to come. 10 draws, 6 wins for Black and 2 for White.

Both of the wins for White picked up the game from the point of the draw with

32. Qa3 Rb8

and continued with White, soon after, aiming to play Nd1 and Nc3 (Stockfish went for those on moves 34 and 35, Ethereal on moves 35 and 37).

This is apparently exactly the line Carlsen said he was worrying about.

The position also seems to present some pretty tricky endgames. Laser, which has a rating 300 points higher than Carlsen's and gets to play with tablebase, managed to lose as Black that way.

(Caruana also suggested he might have tried to get a knight to e6; the engines have drawn a few times starting with that line)

View on HN

The major problem with using the chess.com data is they don't take into account the time factor. When Magnus offered, the draw, Fabi had exactly 15min 42sec time remaining to play the next 10 moves. That's less than 2 minutes per move, versus Magnus having between 2x-3x that amount.

It is hard to understand just how significant of a handicap this is unless you've ever played competitive chess at a national or international level. The human mind starts to break down and freeze up when it is forced to make critical decisions in limited time. This was already starting to happen in the game leading up to the draw offer, as Caruana took more and more time as his positioning was worsening.

The end result of this is one player being forced to make a rapid series of moves in a "time scramble" while the other player has the luxury of checking and rechecking his calculations. This gives the player with more time a massive advantage in practice. We're talking at least 100-200 rating points.

If chess.com wanted to do a fair analysis, they should have handicapped the White computer with 3-4x less compute time. I guarantee the results would have been significantly more favorable for Black then they already were.

Caruana's been under time pressure over and over and over in this match, and blundered his way into zero losses as a result.

Time pressure is practically where he lives, because of how much time he spends calculating. The idea that he's suddenly going to completely forget how to play when the clock is running low doesn't hold water. Especially when he wasn't facing any immediate threat on the board; the advantage for Black was there, but it was theoretical/positional, and was going to take significant work to convert into a result, or even into lines that would make Caruana have to sweat through the next series of moves. And the engine games keep showing that all of Black's advantage can be thrown away with one or two inaccurate moves even very soon after the draw occurred; Carlsen's well within his rights to decide not to risk that just to try to toy with his opponent.

Reply on news.ycombinator.com