Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / Elo difference between Deep Blue 96 and 97?
- - By rocket (****) Date 2021-03-08 11:23
There hasn't been much discussion on the difference in playing strength between these two versions, only numbers.... After all, Kasparov playing regular chess beat 96 DB easily..

What would the difference in elo be by depth improvement alone?

Very hard to tell from few games in just a few openings.
Parent - - By Carl Bicknell (*****) Date 2021-03-08 12:54
Here are some considerations to factor in:

a) 6 game matches are TINY, we barely have any data. So all elo estimates are very rough. Just because DB1 lost and DB2 won, doesn't necessarily imply DB2 was much stronger (though I think it was)

b) Pretty sure Kasparov was going through a divorce in his match with DB2 so his play may have been affected and it didn't look that good.

c) IIRC DB1 calculated at 100M nps and DB2 200M nps, but I don't remember the search depth difference. I do remember a figure of 12 plies for DB2 + extensions on top.
Parent - - By rocket (****) Date 2021-03-08 13:30
You are thinking of Kasparov-Kramnik 2000... that's when Kaspy was going through a divorce. I don't believe he had two within such a short time span.

For what it's worth, Kasparov considered his 96 win over Deep Blue a great accomplishment.

After he lost in 97, he said that his confidence was an all-time low and his playing results suffered, with a much higher draw rate. But he was back in to shape in 1999.
Parent - By Carl Bicknell (*****) Date 2021-03-10 17:11

> You are thinking of Kasparov-Kramnik 2000... that's when Kaspy was going through a divorce.


Ah yes, I remember now
Parent - - By user923005 (****) Date 2021-03-08 12:55
Here is an article on the hardware and software changes:
https://core.ac.uk/download/pdf/82416379.pdf
The raw speed was doubled from 100M NPS to 200M NPS on average
There were also search and memory improvements.

I guess DB2 was 80 to 100 Elo stronger than DB.

I agree with you that we do not have enough data to do more than conjecture.
Parent - - By rocket (****) Date 2021-03-08 13:08

>I agree with you that we do not have enough data to do more than conjecture.<


We can assuming that the evaluation function didn't fundamentally change. Then we simply compare the same eval with the increased power and depth.

The Ruy Lopez game DB beat kasparov in was very deep into opening book theory. It played a few strong moves but looking at the overall match, the evaluation function did not look much improved...
Parent - - By user923005 (****) Date 2021-03-09 04:40
The eval did change, because they had more memory to store terms, and (specifically) they added opposite bishops information.
They also had a much more extensive opening book.
Parent - By rocket (****) Date 2021-03-09 11:48
Opening book is not eval.

As for more terms, yes, this add new layers to the evaluation in some positions, but far from all. In dry middle game positions it most likely won't show. Also note that expanding the eval does not equate to higher elo automatically.
Parent - - By Carl Bicknell (*****) Date 2021-03-10 17:17
Have you read "Behind Deep Blue" ?
They detail some of the changes there. It's a good read but chess-light.
Parent - By rocket (****) Date 2021-03-12 13:30
No. I'll try and get a copy of it.

I ws quite impressed by Deep Blues play in game 1 of 97, and I don't think Kasparov knew that the line he got forced into playing was winning. I really question If he would have went for it If he knew that this exchange sac was coming

Sometimes you just accidentally end up in winning positions, just as tennis players can sometimes make winners out of lucky hits.
I
Parent - - By MrKris (***) Date 2021-03-08 18:58
Thanks for the link!

> Here is an article on the hardware and software changes:
> https://core.ac.uk/download/pdf/82416379.pdf


Someday I want to compare computer material like this to books for humans that I already have, 'Think Like a Grandmaster' etc.
Parent - By user923005 (****) Date 2021-03-09 04:38 Upvotes 1
Highly recommended is "The System: A World Champion's Approach to Chess" by Hans Berliner
There are some really, really interesting ideas in that book.
- By rocket (****) Date 2021-03-15 18:34 Edited 2021-03-15 18:46
I am going to plug this game into the analysis of different engines and make an estimate.

It's round 4 vs kasparov, Deep Blue is White.

Gonna see what Stockfish thinks about it...

The Nd1, Nc3 back and forth are not exactly 3000 elo level moves:smile:

Yasser Seirawan was very critical of kasparovs 16 f6.  I like it...

Search may well be in the 2900 range... although I doubt it.

1.e4 c6 2.d4 d6 3.Nf3 Nf6 4.Nc3 Bg4 5.h3 Bh5 6.Bd3 e6 7.Qe2 d5 8.Bg5 Be7 9.e5 Nfd7 10.Bxe7 Qxe7 11.g4 Bg6 12.Bxg6 hxg6 13.h4 Na6 14.O-O-O O-O-O 15.Rdg1 Nc7 16.Kb1 f6 17.exf6 Qxf6 18.Rg3 Rde8 19.Re1 Rhf8 20.Nd1 e5 21.dxe5 Qf4 22.a3 Ne6 23.Nc3 Ndc5 24.b4 Nd7 25.Qd3 Qf7 26.b5 Ndc5 27.Qe3 Qf4 28.bxc6 bxc6 29.Rd1 Kc7 30.Ka1 Qxe3 31.fxe3 Rf7 32.Rh3 Ref8 33.Nd4 Rf2 34.Rb1 Rg2 35.Nce2 Rxg4 36.Nxe6+ Nxe6 37.Nd4 Nxd4 38.exd4 Rxd4 39.Rg1 Rc4 40.Rxg6 Rxc2 41.Rxg7+ Kb6 42.Rb3+ Kc5 43.Rxa7 Rf1+ 44.Rb1 Rff2 45.Rb4 Rc1+ 46.Rb1 Rcc2 47.Rb4 Rc1+ 48.Rb1 Rxb1+ 49.Kxb1 Re2 50.Re7 Rh2 51.Rh7 Kc4 52.Rc7 c5 53.e6 Rxh4 54.e7 Re4 55.a4 Kb3 56.Kc1 1/2-1/2
- By rocket (****) Date 2021-03-25 10:33 Edited 2021-03-25 14:17
Wow! Deep Blue blundered in game 1 of the 97 match with the queen trade. Stockfish reports that it's a dead draw (0.00) with h5!

Kasparov was very dismissal of Deep Blues play in that game but it turns out that "it" was correct, until it botched it up.
Up Topic The Rybka Lounge / Computer Chess / Elo difference between Deep Blue 96 and 97?

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill