GM Kaufman v.Novag Constellation 3.6 QR Odds (Moves 1-40)

[lkaufman]

Connie does indeed defend the d pawn sliding the R over with..
15..Rfd8

[d] r2r2k1/pbppnppp/1pn5/4Pq2/2B5/B1Q2N1P/P4PP1/3R2K1 w - - 0 16

personally i would have chosen the other R with Rad8.. but then again.. if i were up a R+PP against a GM i would probably be offering you a draw and writing a book to sell my memoirs


Shameless Regards
Steve

I play 16.Bb2, aiming for e6. I don't expect Connie to overlook mate on g7 though!

[Milos]

I think that's a little high between human vs. human but it may very between engine vs engine and human vs engine.

I know the calculations between human vs human is much lower.

A pawn worth approx. 100 rtg. pts. in the initial position. First move maybe 20 rtg. pts. I'd need to check again. I could see maybe 120 rtg. pts. for a pawn and up to 35-40 rtg. pts. for first move, maybe. So 600-700 rtg. pts. for a Rook? Q-Rook being closer to 600 and as low as 500.

I think also it's lower for human vs. human games. But human vs. engine, if engine has odds looks like closer to engine vs. engine.
And as Larry says it's really dependent on TC and strength of the engines.
For 2000 engine and longer TCs (minute per move), Rook is probably worth 1000 elo, for 1500 around 600, for 1000 maybe only 300.

[Terry McCracken]

I think that's a little high between human vs. human but it may very between engine vs engine and human vs engine.

I know the calculations between human vs human is much lower.

A pawn worth approx. 100 rtg. pts. in the initial position. First move maybe 20 rtg. pts. I'd need to check again. I could see maybe 120 rtg. pts. for a pawn and up to 35-40 rtg. pts. for first move, maybe. So 600-700 rtg. pts. for a Rook? Q-Rook being closer to 600 and as low as 500.

I think also it's lower for human vs. human games. But human vs. engine, if engine has odds looks like closer to engine vs. engine.
And as Larry says it's really dependent on TC and strength of the engines.
For 2000 engine and longer TCs (minute per move), Rook is probably worth 1000 elo, for 1500 around 600, for 1000 maybe only 300.

Yes, it would vary a lot factoring T/C and rating.

[Terry McCracken]

It's very dependent on time limit. I can give rook odds to 1800 human in blitz, but at 40/2 I would think the proper rating would be in the 1400-1500 range for rook handicap. Of course that is about a thousand points below me. But it's not a constant; I believe an 1800 would be favored vs. Anand at rook, and a 1000 would lose badly to a 2000 at rook. I have one young student rated around 1050 whom I play at rook plus White plus 10 to 5 time odds, and I've won thirty in a row. I think I have have underestimated Connie because probably most of the handicap games I've played with engines in the past have been at faster levels, maybe 30" per move average or so. Connie would be rated around 1950 on the USCF scale (based on the 2018 official rating of the subsequent model SuperConstellation), and as my USCF rating is about 2450 I've only got a 500 Elo lead, surely not enough to give a rook in a slow game (knight might be another story though...). But I'll do my best.

Yeah, it looks tough, at least draw it if you can. I wondered if the opening you chose was a bit optimistic, you got good mobility and open lines for your pieces but at a hell of a cost, the Connie needs to do something very wrong. It might happen if you find a way to put it on the defence and if it fails to develope on the best squares. It certainly can suffer from the horizon effect among other things...maybe you'll get that crack and win in due time. I think you're strong enough to pull it off. You were very right on how the Connie might behave to your 14. Rc1 tricking it to castle short.

[Steve B]

Connie would be rated around 1950 on the USCF scale (based on the 2018 official rating of the subsequent model SuperConstellation), and as my USCF rating is about 2450 I've only got a 500 Elo lead, surely not enough to give a rook in a slow game (knight might be another story though...). But I'll do my best.

As a Reminder to folks just now following this experiment.. and not from the beginning which began in another thread ...
http://www.talkchess.com/forum/viewtopi ... 20&t=35125


Eric Hallsworth rates the Connie 3.6 Mhz at 1646 ELO (Selective Search Lists) and the German Dedicated Chess computer forum
(www.schachcomputer.info) rates her at 1648 ELO
http://www.schach-computer.info/wiki/in ... lation_3.6

even a very liberal addition of 150 pts to convert to USCF would bring her to about 1800 USCF
added to this is the fact that you are extremely familiar with PC Engines not to mention the fact that at one time you were the foremost authority on dedicated computers in the world has to be worth 100-150 Elo added to your USCF rating in a match like this
im guessing the difference in playing strength between you and Connie is closer to about 700-750

this game is meant as part of the experiment i started in which i have played 1700 ELO Dedicated computers v a 2700 PC Engine to see If a R can overcome the 1000 pt gap in elo
lesser odds games (minor piece odds) or time odds games do not seem to be able to overcome 1000 Elo but a R does
the real world implications of all of this is that i think the difference between a local club player(rated about 1700 ) and a top GM(rated about 2700) is a R

of course this one game proves nothing but if Connie can manage to draw or win then it does add a little bit more weight to the theory

Steve

[zamar]

First 2000 elo is not a patzer.

Yes he is When I afterwards look through my games (even without computer assistance) and see the horrible number of pure miscalculation and overlooks, "patzer" is the title I deserve Of course there will always be things which are beyond mycalculation horizon, but I mean that ordinary 2000 elo player misses and miscalculates lot of 2-3-4 ply tactics which should easily be in his search horizon. Of course even World Champion sometimes misses these but it happens very rarely. I'd say when player reaches 2200-2300 level then he quite rearly misses 2-3-4 ply tactics (but of course more often misses deeper 5-6-7 ply tactics). Chess is a cruel game. Miss one single tactic in a game and you are toast.

So there is no way a 2000 elo player looses from any GM. Loosing would simply mean his real rating is much lower than 2000.

I disagree. This is perhaps true for computer chess, but not for human chess. With white pieces GM can create so irritating complications (like current position that is on the board) that 2000-elo player loses or draws once in a while, I'd of course expect him to win 10 game match with comfortable margin..

[Steve B]

I play 16.Bb2, aiming for e6. I don't expect Connie to overlook mate on g7 though!

Connie perhaps a bit nervous about the e6 mate threat and her exposed Q plays the timid...
16..Qg6

[d] r2r2k1/pbppnppp/1pn3q1/4P3/2B5/2Q2N1P/PB3PP1/3R2K1 w - - 0 17

After the 16th move Connie leaves the playing hall for a spot of dinner and stops to pose for a photo for the frenzied crowds of Paparazzi waiting outside



Dinner At Tiffany's Regards
Steve

[Milos]

I play 16.Bb2, aiming for e6. I don't expect Connie to overlook mate on g7 though!

Connie perhaps a bit nervous about the e6 mate threat and her exposed Q plays the timid...
16..Qg6

It's far from a timid move. She is starting to attack and Larry can soon be in a big trouble .

[Steve B]

My rating is also not much above 2000 but I believe in material so even if black lose the pawn h7 it does not seem dangerous to me(and black can defend).

The only question is if there is a way to cause the computer to make mistakes.
I guess that after
After g4 Qg6 Bc1 the computer is going to find the best move Na5 with the idea Qc6.

I did not think about it before looking at the analysis of engines but the move that I thought about(d6) is probably also enough.
Uri

i notice Movei is rated about 2700 on the Rating lists
i think it might be a good choice next to continue this experiment and pit a different 1700 Dedicated against it at R odds
i would need someone to operate Movei though and post moves at one move per day much like in this match now
if anyone is interested in doing that please let me know

Steve

[pacifist]

I think now white has good pratical chances but maybe it will be necessary to give up an exchange to profit from greed of black