New opponent, new handicaps for Komodo.

[duncan]

3. Komodo plays Black with his king starting on e4 and pawn on f6, White to move!! The pawn on f6 makes the position legal. This looks like it should be a simple forced mate, but Komodo doesn't see one.

so what is the score when komodo plays against itself. ?

I used Fritz 15 instead of Komodo to make use of the Monte Carlo option, with which I could get lots of variety very quickly. Somewhat to my surprise, White scored 84% in 700 games at ten ply; I expected more like 90%. Probably at any realistic time limit it would be close to 90%. I imagine many human GMs would say White had full compensation for a piece if we removed a White knight, at least at first glance. But apparently this is not so. When I removed the White knight from g1 from the above position, Black scored the same 84% that White scored with the knight! So we can say that the Ke4 handicap is "half a knight".

did you ever say each move is worth a third of a pawn ?

Yes, and since it takes four moves to bring the king home and White moves first, it's four and a half moves so 1 1/2 pawns. But Black has also lost casttling rights which brings it up to 2 pawns. But a knight is worth about 4 pawns in the opening, so it is indeed "half a knight".

if komodo was playing carlsen how many moves would you offer him and being confident of a win?

[JJJ]

A program just 274 elo below Komodo did two poor draw at 1min+1 sec and 4 CPU , so my guess is Carslen won't ever win this handicap match. There is too few winning move and one mistake can make Komodo leading quickly.

[lkaufman]

A program just 274 elo below Komodo did two poor draw at 1min+1 sec and 4 CPU , so my guess is Carslen won't ever win this handicap match. There is too few winning move and one mistake can make Komodo leading quickly.

In a long match at FIDE standard TC vs Carlsen, I would bet on Komodo at five move handicap (including the first move as White), I would be on Carlsen at six move. But most of the games would be draws.