stockfish 10 vs. Mephisto III S Glasgow

[Michel]

Black is worse but not clearly losing before 16...Nc5 so I analyzed the reason that stockfish did not play 16...cxd5 and the problem is that after 17.Nxd5 stockfish prune all the queen moves so it does not see that the queen can escape and believe that 17.Nxd5 wins the queen(only at depth 6 it can see that the queen can escape).

The analysis also suggest that stockfish developement version prune 17.Nxd5 for some reason and need depth 3 to see it(also seems strange).

The reason is well known. SF does late move pruning in PV nodes. This causes substantial elo loss at low depths but is elo neutral at higher depths. So it was admitted as a simplification.

[mclane]

Yes a design decision.

[mclane]

Experiment:

People have told you so many times that those are not relevant because what is a node for Mesphito is nothing like what is a node for Stockfish?

I hold that any decent programmer could make an engine that completely destroys Mephisto while searching even less nodes than Mephisto, but it wouldn't be an engine that makes plans and it wouldn't resemble smart AI. And it would be much weaker than Stockfish so my question would be, what would be the point of minimizing nodes searched?

You think any decent programmer could make a chess engine that does less then 1 node per second and play decent chess ??


Mephisto III was a commercial dedicated chess computer made in 1983
https://www.schach-computer.info/wiki/i ... phisto_III

The programmer decided to use a very small tree to beat the opponent dedicated chess computers. At that time brute force was a big topic and the opponent machine did 500-1500 NPS mainly doing A strategy with extensions.
At that time the programmer decided that this is boring and also stupid and came with an engine that is doing very very few NPS.
A Node is the amount of evaluated positions the computer generates.
Mephisto III had so few nodes in the tree that the programmer was able to print out the whole search tree after 3 minutes computation.
This way he could analyse very good why the machine was mistaken or failed.

Todays chess programs and their programmers cannot do it because they decided to do millions of NPS. No chance to find out what was going wrong in the tree.

So this dedicated chess computer from 36 years ago beats todays stockfish 10 . Both engines use the same amount of nodes.

[Uri Blass]

Black is worse but not clearly losing before 16...Nc5 so I analyzed the reason that stockfish did not play 16...cxd5 and the problem is that after 17.Nxd5 stockfish prune all the queen moves so it does not see that the queen can escape and believe that 17.Nxd5 wins the queen(only at depth 6 it can see that the queen can escape).

The analysis also suggest that stockfish developement version prune 17.Nxd5 for some reason and need depth 3 to see it(also seems strange).

The reason is well known. SF does late move pruning in PV nodes. This causes substantial elo loss at low depths but is elo neutral at higher depths. So it was admitted as a simplification.

I disagree that it is a simplification.
it seems to me the opposite because doing it make it harder for chess players to understand what stockfish is doing.
i thought that it gives elo but if it does not give elo but only elo neutral then it means
there is no justification to do it from my point of view.

I think that there is a problem in the order of moves of stockfish.
If the knight has a threat against the queen it does not make sense to search first a move that does not save the queen
and I wonder if it is not productive to spend slightly more time in ordering the moves in order to have a better order of moves.

[mhull]

My point was that higher nps is not needed for chess.
Because there are ways to rely on knowledge instead of search.

Why should I now increase nps ???

Of course Stockfish is the master of bean counting.
I have no doubt that it will Win.

But my point was that all those nps are irrelevant if you want to play chess and UNDERSTAND what you are doing.
Stockfish is not understanding what it is doing. It is more a pocket calculator.
It outcomputes all possible moves with these high amount of calculation.

In earlier years fritz was such an example of search concentrated chess program. And in the years of dedicated chess computers
It was mm2 .
What has changed is the speed of the hardware used.

At least factor 2000 from the 6502 days.

The reality is that Stockfish 8 (Cfish) does 300-600 NPS on an MC69030 @20 Mhz, not much faster machine than the Mephisto. So limiting Stockfish to 1000 nodes is "not fair", IMHO.

[mclane]

What is MC69030 @20 Mhz ??

And what is your point. Mephisto III uses an 8 Bit cpu with 3.5 or 6 MHz.

[PK]

If human matches were organised the way you organise Stockfish-Mephisto matches, one opponent would have an incompatibile liver transplant, and the other an axe in his head.

[syzygy]

This is really not AI but instead artificial stupidity.

The progress you've made in about 8 months in this thread disproves the existence of HI.

My part on this posting is mainly to quote the source:
https://glarean-magazin.ch/2019/04/18/k ... ch-report/

Neither your first post in this thread nor the post to which I responded has anything to do with Leela. I'm afraid you are not making any sense. Which proves my point that this thread disproves the existence of HI.

[mclane]

I have no idea what you talk about.

I sum up:

Mephisto III S Glasgow 68000 (5-10 NPS) wins against stockfish 10.
Even worse
Mephisto III 8 bit 1802 1-3 NPS with 6.1 MHz wins against stockfish 10.
Mephisto III 8 bit 1802 1 NPS 3.5 MHz on 4 AA batteries wins against stockfish 10 too.

Ic0 plays better chess then Stockfish.

Stockfish has only a chance if it has a hardware advantage of 1000-1500 times better hardware and 1000 times better ram and rom and core advantage of factor 4-64.

So the question is: is there any software progress at all ??

If a 36 year old engine beats it having the same NPS, has computerchess made any progress at all ??

There is no doubt hardware made progress.

[Michel]

I have no idea what you talk about.

I sum up:

Mephisto III S Glasgow 68000 (5-10 NPS) wins against stockfish 10.
Even worse
Mephisto III 8 bit 1802 1-3 NPS with 6.1 MHz wins against stockfish 10.
Mephisto III 8 bit 1802 1 NPS 3.5 MHz on 4 AA batteries wins against stockfish 10 too.

Ic0 plays better chess then Stockfish.

Stockfish has only a chance if it has a hardware advantage of 1000-1500 times better hardware and 1000 times better ram and rom and core advantage of factor 4-64.

So the question is: is there any software progress at all ??

If a 36 year old engine beats it having the same NPS, has computerchess made any progress at all ??

There is no doubt hardware made progress.

This is really exasperating. As I explained above, SF's search algorithm is explicitly _not_ designed to play sensibly at very low node count. It will simply blunder. So it makes no sense to force it to play in this way. This is a choice of the developers. It has nothing to do with "understanding chess".