Houdini-Rybka (150m + 30s) LIVE! - TCEC Elite Match Season 1

[Laskos]

One observation: Houdini uses in many games clearly less time than Rybka and around move 40 has 1 hour more in clock. Even in some won games it still has 1 hour to use!

Jouni

Yes, Houdini is too conservative with its time usage, I will certainly adapt this for Houdini 2.0. In most middle games of the match Houdini is actually playing with a time handicap!

It's funny to see how a lot of people that follow the match at ChessBomb fail to understand that good TM means that you're supposed to have used up the time on your clock by the end of the game. They see Houdini with a lot of time left on the clock and Rybka with little time left, and conclude that Houdini must have "good TM" and Rybka "very bad TM", when it's in fact the opposite.

Robert

Among the top 3 engines, the best TM (using all the time without getting into time troubles) is
1. Rybka
2. IvanHoe
3. Houdini

Kai

[beram]

incredible that H15 in this position still thinks white is only +0,3 better, because in fact he is lost here. This is a rather common and known won endgame for the rook with extra free a-pawn. The white king can attack on the kingside pawns, while black is bound on the a-pawn.

kind regards, Bram

Houdini doesn't recognize this particular position, and evaluates this as a standard rook ending with an extra pawn.
Most of the 4 pawns vs 3 pawns rook endings are actually draw, so a +0,3 eval makes sense (in the general case).

Robert

Thx for answering.
But don't you agree that this specific eval deserves attention so to speak ?
because at move 47, Rybka4 already on +1,9 and H15 still on 0,3
[d] 8/8/r4pp1/3k3p/P4R2/6P1/5PKP/8 w - - 7 47

kind regards Bram

[Houdini]

But don't you agree that this specific eval deserves attention so to speak ?
because at move 47, Rybka4 already on +1,9 and H15 still on 0,3
[d] 8/8/r4pp1/3k3p/P4R2/6P1/5PKP/8 w - - 7 47

kind regards Bram

Bram, I will certainly take a closer look at this.
Houdini fails to find the winning manoeuvre with 47.a5! in a reasonable amount of time.

Robert

[Martin Thoresen]

The match has now concluded and Houdini is the TCEC Season 1 Champion.

Congrats!

Final result was 23.5 for Houdini to 16.5 for Rybka.

Best,
Martin

[lech]

But don't you agree that this specific eval deserves attention so to speak ?
because at move 47, Rybka4 already on +1,9 and H15 still on 0,3
[d] 8/8/r4pp1/3k3p/P4R2/6P1/5PKP/8 w - - 7 47

kind regards Bram

Bram, I will certainly take a closer look at this.
Houdini fails to find the winning manoeuvre with 47.a5! in a reasonable amount of time.

Robert

Many of engines (Stockfish too) doesn't give a small bonus for x-ray (R, Q) attacks by own passed pawns. I saw many such strange examples.
It is a fact, that it is not easy to set the proper value of this bonus.

[S.Taylor]

Congrats to Houdini!

Houdini seemed to do well, right at the beginning and right at the end.

So the result looks pretty satisfying that Houdini is stronger, atleast as much as earlier matches showed that Rybka 4 had been stronger.

But this is only if you count the beginning and last few games. Otherwise you would not see it his way at all. Maybe we would see they look dead even.

When Rybka was winning vs SF and Houdinis 1.02, 1.03 or whatever, Rybka was winning at every juncture. In a steady fashion.

You say that means nothing at all? maybe. Food for thought.

[Jouni]

23,5/40 means +61 points. I think the difference is even more - like larkin test indicates +75.

Jouni