Pawn odds

[hgm]

Is there a standard value for the advantage conveyed by a pawn-odds handicap? (and does anyone know it?)

[Vempele]

Search rybkaforum for it. I can't remember how much it's been tested, but Larry Kaufman thinks non-edge pawns are about 200 Elo in blitz and 300 at slow time controls.

[George Tsavdaris]

Is there a standard value for the advantage conveyed by a pawn-odds handicap? (and does anyone know it?)

I have saved some of Larry Kaufman's estimations about these. He seems to be involved very much on these handicap games so i would trust him....

Logically you will find more estimations if you search for example for "Pawn handicap" + lkaufman in the Rybkaforum:
http://rybkaforum.net/cgi-bin/rybkaforu ... order=desc

+++++++++++++++++++++++++
L.Kaufman:

It is certainly true that White will have some compensation for the missing pawn in all cases except the "f" pawn. However, in the case of "b" "c" "e" and "g" pawns, the compensation is at best one tempo, since simply moving that pawn would open all the same diagonals and also control some good squares.

As the usual rule (which I think is reasonably accurate) states that three tempi equal a pawn, it is clear that these are all significant handicaps.
The "d" pawn's removal gives slighty more comp, because the queen is on a half-open file, but this should hardly be worth more than the squares that would be controlled by a pawn on d4, so even in this case the compensation is clearly insufficient. Actually removing the "d" pawn is similar to playing a Smith-Morra Gambit a tempo down, except that the "c" pawns are off the board. The Smith-Morra is considered dubious at GM level normally; a tempo down it should be quite unsound.

This leaves the two edge pawn handicaps. Only in these cases can it be seriously argued that the compensation is anywhere near to being enough. Edge pawns are worth less to start, and the semi-open file for the rook is quite nice. In my opinion Black will still have an advantage even in these two cases, but his advantage may be only slightly more than the advantage White normally has in an even game.

As for some other posts, anyone who plays Rybka at handicap on a 32 bit single processor must bear in mind that the 64 bit MP on a quad is roughly a class stronger. As for playing blitz with an average IM at knight odds, bear in mind that "average IM" does not mean "average IM who plays blitz all day on the internet"; clearly among players of any level there are blitz specialists and those who only play blitz occasionally. As for takebacks as a handicap, this is indeed a valid idea, but I'm afraid that if the player is a mere 2600 Elo GM he would need so many takebacks to have equal chances that it would be laughable. Maybe for Kramnik or Anand or Topalov it might be reasonable.



I had Rybka play a thousand games with herself (using the randomizer) at 8 ply at each of the pawn handicaps. These are very fast games, faster even than 1' (bullet) chess, but Rybka is still playing at around the level of a master at 40/2.
The two edge pawns proved to be just moderate handicaps, on average 67 Elo, roughly double the normal White advantage in chess.
The other six pawns were all major handicaps, averaging 171 Elo, nearly a class (f2 was the worst to remove).

It was also clear from the trend (comparing to even shorter searches) that at longer time controls, the handicap would be greater, since the strength of the receiving player counts for more than the strength of the receiving player. I also did some tests against strong s.p. programs like Fritz 10 and Shredder 10 at time controls ranging from blitz to 25' +10", which proved that this handicap (all 8 white pawns one by one) is roughly fair between Rybka 64 bit MP quad and these SP programs (2.2n2 lost narrowly, 2.3 LK won narrowly). Since the rating gap between 2.3 LK and these oher programs is almost exactly a class on the last CEGT 40/20 list, and allowing another 40 point or so for my quad, it is clear that the handicap is somewhat more than one class between computers.
However with a human involved, it should be a greater handicap, since he can modify his "programming" for the handicap. Therefore it should be quite a close match, I feel, though we are optimistic!