lkaufman wrote:The resemblance to Halloween gambit is superficial. That gambit is justified by the facd that after kicking the c6 knight away White has e5 to chase the other knight back home. Here the knight is already at home so the sacrifice is totally unsound. Two tempi less for White here. One factor in the GMs favor is that all four positions could happen in real games without the GM playing any strange moves, so the match should feel more like normal chess. He gets four won games at the start. He can probably make draws if he so desires, but winning won't be easy. As usual we offer 3 times as much for a win as for a draw, mimicking the so-called "football" scoring system used in some major chess events.
Some of us would much rather see Komodo actually playing the real Halloween Gambit vs a GM. Any thought to attempting some well known dubious gambits in these odds matches, if only for one game? You've tried so many different handicaps, after all.
lkaufman wrote:The resemblance to Halloween gambit is superficial. That gambit is justified by the facd that after kicking the c6 knight away White has e5 to chase the other knight back home. Here the knight is already at home so the sacrifice is totally unsound. Two tempi less for White here. One factor in the GMs favor is that all four positions could happen in real games without the GM playing any strange moves, so the match should feel more like normal chess. He gets four won games at the start. He can probably make draws if he so desires, but winning won't be easy. As usual we offer 3 times as much for a win as for a draw, mimicking the so-called "football" scoring system used in some major chess events.
Some of us would much rather see Komodo actually playing the real Halloween Gambit vs a GM. Any thought to attempting some well known dubious gambits in these odds matches, if only for one game? You've tried so many different handicaps, after all.
Thanks,
CL
Well, if the GM can win against the much less sound "gambits" planned for this match then it might make sense. But I suspect that with any gambit good enough to have a name, at best the GM might make a draw.
lkaufman wrote:The resemblance to Halloween gambit is superficial. That gambit is justified by the facd that after kicking the c6 knight away White has e5 to chase the other knight back home. Here the knight is already at home so the sacrifice is totally unsound. Two tempi less for White here. One factor in the GMs favor is that all four positions could happen in real games without the GM playing any strange moves, so the match should feel more like normal chess. He gets four won games at the start. He can probably make draws if he so desires, but winning won't be easy. As usual we offer 3 times as much for a win as for a draw, mimicking the so-called "football" scoring system used in some major chess events.
Some of us would much rather see Komodo actually playing the real Halloween Gambit vs a GM. Any thought to attempting some well known dubious gambits in these odds matches, if only for one game? You've tried so many different handicaps, after all.
Thanks,
CL
Well, if the GM can win against the much less sound "gambits" planned for this match then it might make sense. But I suspect that with any gambit good enough to have a name, at best the GM might make a draw.
I think that the interesting question is if there are special positions that GM's still understand better than computers.
I would like to see a fair match of 2 games from the same position when the human can choose the position when the target of the human is to get at least 1.5 out of 2.
1 is easy to get because you can choose a winning position for one side.
1.5 is a different story and in theory if the computer play perfect it is impossible.
In order to prevent preperation that is based on some long tactical line that chess programs cannot see at tournament time control it is possible to decide that the GM has to decide about the starting position
one week before the match so the computer team can prepare a special opening book for it but cannot modify the program except the book.
carldaman wrote:Some of us would much rather see Komodo actually playing the real Halloween Gambit vs a GM. Any thought to attempting some well known dubious gambits in these odds matches [...]?
Me too. I'd prefer see Komodo take the white side of the Cochrane, or the Balck side of the Albin, these are gambits with a name, and with adequate preparation, the GM could actually do something here.
Are we sure those gambits provide a good enough handicap vs a 2500er..? I know e.g. the fried liver is rated around +1.00 for white, but it's still open for debate whether it's a win or loss... surely K will gain an edge quickly regardless of side, no?
I'm pretty sure that Komodo will win or draw but almost never lose from either side of any of the gambits mentioned. None of them are as bad as giving the f7 handicap, and no GM has won a game with that handicap. That's why I had to invent particularly unsound gambits for this match. Regarding Uri's idea of the GM picking a start position, I think it would be rather easy to find positions that no computer understands and that a strong human would beat the computer in a two game reversal match. No one is claiming that computers are stronger than humans in all positions, just in "almost" all positions.
lkaufman wrote:I'm pretty sure that Komodo will win or draw but almost never lose from either side of any of the gambits mentioned...
Yes, I forgot to mention to use "draw odds" for this well known gambits.
Well, we haven't yet tried any version of giving draw odds. The first thing to try might be (against a very strong GM) draw odds plus White pieces plus no (or maybe 2 or 3 move) opening book plus single-core only plus maybe double time handicap. But I don't know whether the present contempt (regardless of the value chosen) is really sufficient to avoid draws at all costs from a lousy opening position which can be expected given the above. To do well giving the above handicaps Komodo might need significant work; after all it's been optimized to aim for draws from bad positions.
lkaufman wrote:...
Scheduled for April 19 and 20 on chess.com/tv, with Perelshteyn commenting his own game just as Benjamin (and Hess and Williams) did, at same TC as Benjamin (45' + 30").
...
