stockfish 10 vs. Mephisto III S Glasgow

[mclane]

There are no engines that have a plan in chess.

Imo it would increase the quality of the games.

Dont you think that THIS is a reason LC0 was such a sensation?! Because it played better chess.
And as a result of this todays AB thinkers use nnue.

Do you not consider a principal variation "a plan"? Why?

A prinzipal variation is NOT what i understand as plan.

[Madeleine Birchfield]

There are no engines that have a plan in chess.

Imo it would increase the quality of the games.

Dont you think that THIS is a reason LC0 was such a sensation?! Because it played better chess.
And as a result of this todays AB thinkers use nnue.

Do you not consider a principal variation "a plan"? Why?

A prinzipal variation is NOT what i understand as plan.

Well, then, I don't know what to tell you.

Neural networks like the ones used in Leela do not create plans either. They are merely universal approximations of any function, which in chess, the function to be approximated is an evaluation function which takes in a board representation and whose output is a value from the set {W,D,L} and usually approximated by real numbers or integer centipawns and a sigmold/tangent transform. No different than handcrafted evaluations, only that they are much better at their job at approximating the actual function than any handcrafted evaluation ever could be.

https://en.wikipedia.org/wiki/Universal ... on_theorem

[Sopel]

There are no engines that have a plan in chess.

Imo it would increase the quality of the games.

Dont you think that THIS is a reason LC0 was such a sensation?! Because it played better chess.
And as a result of this todays AB thinkers use nnue.

Do you not consider a principal variation "a plan"? Why?

A prinzipal variation is NOT what i understand as plan.

Could you elaborate? Or are we in a belief territory?

[mclane]

For a plan you have to give up the maxime to prefer the best move in a given position.
Because the best move in a given position creates no plan.

[amanjpro]

For a plan you have to give up the maxime to prefer the best move in a given position.
Because the best move in a given position creates no plan.

It is not the best move that engines find, but the best line! And this is a plan.

The only caveat is, engines (actually not engines but UCI), is stateless.
That means tell the engine find the best line, it does. Then play the first move and ask
the engine to find the best continuation, it searches all over again. But this cannot be
fixed by the engines, but by the protocol itself (that is IF we want to fix it)

[Ras]

The only caveat is, engines (actually not engines but UCI), is stateless.

The protocol, that is - not necessarily the engines.

But this cannot be fixed by the engines

It can. My engine is UCI and does register a PV hit so that it starts off right with the PV continuation as likely best move. It also notices if the position is actually the same as last time, i.e. without a move being made, in which case it uses the last PV, not the continuation.

[mclane]

That UCI is not taking care about computational results and behaves as if chess is a Position solver is part of the wrong paradigm.
Winboard was not any different concerning this, or ?!
The main line is the result of search until ply X.
But this is not a chess plan !

I give you an example.

You want to drive into holidays by car.

You begin in Dortmund and want to drive to Spain.
Now the plan.
Drive from Dortmund to frankfurt.
Next iteration.
Drive from Dortmund to luxemburg.
Next iteration.
Drive from Dortmund to austria.

Now the plan: drive from Dortmund to spain. Question: whats the shortest or fastest or most beautiful travel ?!
Then consider which motorways or streets.
Usually navigationsoftware asks you how you wish it to compute and then creates a who,e journey.

Thats a plan.

Driving from town to town without seeing the whole is how engines work today.
Doing it like today’s navigation software is doing is the plan.

A plan sometimes sacrifices a small range target for a long range manoeuvre.

Any human beeing knows what a plan is.
How plans are generated. How they fail and they successfully work .

A successful plan increases quality of play.
This is very difficult to measure because both cars reach the target.
But car A drives from town to town while car B PLANS the whole journey.

Similar with a life plan.
That is very long range. You go to kindergarten. School. Army. Job. Wife. Children. House. Tree.

[Ras]

I give you an example.

You're anthropomorphising chess. That's not how it works, as evidenced by strength of play.

[amanjpro]

That UCI is not taking care about computational results and behaves as if chess is a Position solver is part of the wrong paradigm.
Winboard was not any different concerning this, or ?!
The main line is the result of search until ply X.
But this is not a chess plan !

I give you an example.

You want to drive into holidays by car.

You begin in Dortmund and want to drive to Spain.
Now the plan.
Drive from Dortmund to frankfurt.
Next iteration.
Drive from Dortmund to luxemburg.
Next iteration.
Drive from Dortmund to austria.

Now the plan: drive from Dortmund to spain. Question: whats the shortest or fastest or most beautiful travel ?!
Then consider which motorways or streets.
Usually navigationsoftware asks you how you wish it to compute and then creates a who,e journey.

Thats a plan.

Driving from town to town without seeing the whole is how engines work today.
Doing it like today’s navigation software is doing is the plan.

A plan sometimes sacrifices a small range target for a long range manoeuvre.

Any human beeing knows what a plan is.
How plans are generated. How they fail and they successfully work .

A successful plan increases quality of play.
This is very difficult to measure because both cars reach the target.
But car A drives from town to town while car B PLANS the whole journey.

Similar with a life plan.
That is very long range. You go to kindergarten. School. Army. Job. Wife. Children. House. Tree.

Can you give me an example of a plan in chess? And try to explain to me how an engine doesn't do the same?

[amanjpro]

That UCI is not taking care about computational results and behaves as if chess is a Position solver is part of the wrong paradigm.
Winboard was not any different concerning this, or ?!
The main line is the result of search until ply X.
But this is not a chess plan !

I give you an example.

You want to drive into holidays by car.

You begin in Dortmund and want to drive to Spain.
Now the plan.
Drive from Dortmund to frankfurt.
Next iteration.
Drive from Dortmund to luxemburg.
Next iteration.
Drive from Dortmund to austria.

Now the plan: drive from Dortmund to spain. Question: whats the shortest or fastest or most beautiful travel ?!
Then consider which motorways or streets.
Usually navigationsoftware asks you how you wish it to compute and then creates a who,e journey.

Thats a plan.

Driving from town to town without seeing the whole is how engines work today.
Doing it like today’s navigation software is doing is the plan.

A plan sometimes sacrifices a small range target for a long range manoeuvre.

Any human beeing knows what a plan is.
How plans are generated. How they fail and they successfully work .

A successful plan increases quality of play.
This is very difficult to measure because both cars reach the target.
But car A drives from town to town while car B PLANS the whole journey.

Similar with a life plan.
That is very long range. You go to kindergarten. School. Army. Job. Wife. Children. House. Tree.

Can you give me an example of a plan in chess (that humans are capable of doing)? And try to explain to me how an engine doesn't do the same?