ProDeo
Would you like to react to this message? Create an account in a few clicks or log in to continue.
ProDeo

Computer Chess
 
HomeHome  CalendarCalendar  Latest imagesLatest images  FAQFAQ  SearchSearch  MemberlistMemberlist  UsergroupsUsergroups  RegisterRegister  Log in  

 

 nodes equivalence of engines

Go down 
2 posters
AuthorMessage
Uri Blass




Posts : 207
Join date : 2020-11-28

nodes equivalence of engines Empty
PostSubject: nodes equivalence of engines   nodes equivalence of engines EmptyMon Apr 15, 2024 7:40 am

From my point of view there are no fast and slow searchers
There are only smart and stupid engine.

If you start from the same playing strenght(let say 2700 elo) and you double the number of nodes of smart new stockfish you probably get the same improvement that you get from multiplying the number of nodes of stupid engine like Fritz or Hiarcs or my own engine by 10(do not know if it is 7 or 13 but the idea is clear).

I wonder if people test chess engines in order to find nodes equivalence function meaning
n nodes per move of engine x is the same playing strength as f(n) nodes of engine y.

Ghppn likes this post

Back to top Go down
Dann Corbit




Posts : 189
Join date : 2020-11-26

nodes equivalence of engines Empty
PostSubject: Re: nodes equivalence of engines   nodes equivalence of engines EmptyFri Apr 19, 2024 12:04 am

Uri Blass wrote:
From my point of view there are no fast and slow searchers
There are only smart and stupid engine.

If you start from the same playing strenght(let say 2700 elo) and you double the number of nodes of smart new stockfish you probably get the same improvement that you get from multiplying the number of nodes of stupid engine like Fritz or Hiarcs or my own engine by 10(do not know if it is 7 or 13 but the idea is clear).

I wonder if people test chess engines in order to find nodes equivalence function meaning
n nodes per move of engine x is the same playing strength as f(n) nodes of engine y.

A big problem with using node counts is that engines count them very differently.
Some engines do not count singular move extensions and some do
Some don't count hash nodes and some do
some don't count qnodes and some do
The meaning of a node for LC0 is very different from a node for Stockfish, and depth also has completely different meanings.

I think a better way to compare strength between different engines would be to use time handicaps, which are linear multiples of the stronger engine time.

Ghppn likes this post

Back to top Go down
Uri Blass




Posts : 207
Join date : 2020-11-28

nodes equivalence of engines Empty
PostSubject: Re: nodes equivalence of engines   nodes equivalence of engines EmptyFri Apr 19, 2024 4:32 am

Dann Corbit wrote:
Uri Blass wrote:
From my point of view there are no fast and slow searchers
There are only smart and stupid engine.

If you start from the same playing strenght(let say 2700 elo) and you double the number of nodes of smart new stockfish you probably get the same improvement that you get from multiplying the number of nodes of stupid engine like Fritz or Hiarcs or my own engine by 10(do not know if it is 7 or 13 but the idea is clear).

I wonder if people test chess engines in order to find nodes equivalence function meaning
n nodes per move of engine x is the same playing strength as f(n) nodes of engine y.

A big problem with using node counts is that engines count them very differently.
Some engines do not count singular move extensions and some do
Some don't count hash nodes and some do
some don't count qnodes and some do
The meaning of a node for LC0 is very different from a node for Stockfish, and depth also has completely different meanings.

I think a better way to compare strength between different engines would be to use time handicaps, which are linear multiples of the stronger engine time.

I see a problem with using time handicap and the problem is that you probably cannot reproduce the games and if you play the game again you may get different results.
I want results that everybody can reproduce.

My target is not to find which engine is better but to see how many elo points engines earn from doubling the fixed nodes per move that is usually equivalent to doubling the time.

I alrady have a pgn of game from fixed nodes tournament and I would like a tool to check that every game appears even number of times inside the pgn.

Ghppn likes this post

Back to top Go down
Dann Corbit




Posts : 189
Join date : 2020-11-26

nodes equivalence of engines Empty
PostSubject: Re: nodes equivalence of engines   nodes equivalence of engines EmptyFri Apr 19, 2024 12:07 pm

Uri Blass wrote:
Dann Corbit wrote:
Uri Blass wrote:
From my point of view there are no fast and slow searchers
There are only smart and stupid engine.

If you start from the same playing strenght(let say 2700 elo) and you double the number of nodes of smart new stockfish you probably get the same improvement that you get from multiplying the number of nodes of stupid engine like Fritz or Hiarcs or my own engine by 10(do not know if it is 7 or 13 but the idea is clear).

I wonder if people test chess engines in order to find nodes equivalence function meaning
n nodes per move of engine x is the same playing strength as f(n) nodes of engine y.

A big problem with using node counts is that engines count them very differently.
Some engines do not count singular move extensions and some do
Some don't count hash nodes and some do
some don't count qnodes and some do
The meaning of a node for LC0 is very different from a node for Stockfish, and depth also has completely different meanings.

I think a better way to compare strength between different engines would be to use time handicaps, which are linear multiples of the stronger engine time.

I see a problem with using time handicap and the problem is that you probably cannot reproduce the games and if you play the game again you may get different results.
I want results that everybody can reproduce.

My target is not to find which engine is better but to see how many elo points engines earn from doubling the fixed nodes per move that is usually equivalent to doubling the time.

I alrady have a pgn of game from fixed nodes tournament and I would like a tool to check that every game appears even number of times inside the pgn.

I think the problem with this: "My target is not to find which engine is better but to see how many elo points engines earn from doubling the fixed nodes per move that is usually equivalent to doubling the time."
Is that an engine might take one thousand times longer to calculate a node than another. So the real ratio of strength is not calculated.
I think your method only calculates effective branching factor (which is, of course, a very useful calculation)

Ghppn likes this post

Back to top Go down
Uri Blass




Posts : 207
Join date : 2020-11-28

nodes equivalence of engines Empty
PostSubject: Re: nodes equivalence of engines   nodes equivalence of engines EmptySat Apr 20, 2024 8:45 am

Dann Corbit wrote:
Uri Blass wrote:
Dann Corbit wrote:
Uri Blass wrote:
From my point of view there are no fast and slow searchers
There are only smart and stupid engine.

If you start from the same playing strenght(let say 2700 elo) and you double the number of nodes of smart new stockfish you probably get the same improvement that you get from multiplying the number of nodes of stupid engine like Fritz or Hiarcs or my own engine by 10(do not know if it is 7 or 13 but the idea is clear).

I wonder if people test chess engines in order to find nodes equivalence function meaning
n nodes per move of engine x is the same playing strength as f(n) nodes of engine y.

A big problem with using node counts is that engines count them very differently.
Some engines do not count singular move extensions and some do
Some don't count hash nodes and some do
some don't count qnodes and some do
The meaning of a node for LC0 is very different from a node for Stockfish, and depth also has completely different meanings.

I think a better way to compare strength between different engines would be to use time handicaps, which are linear multiples of the stronger engine time.

I see a problem with using time handicap and the problem is that you probably cannot reproduce the games and if you play the game again you may get different results.
I want results that everybody can reproduce.

My target is not to find which engine is better but to see how many elo points engines earn from doubling the fixed nodes per move that is usually equivalent to doubling the time.

I alrady have a pgn of game from fixed nodes tournament and I would like a tool to check that every game appears even number of times inside the pgn.

I think the problem with this: "My target is not to find which engine is better but to see how many elo points engines earn from doubling the fixed nodes per move that is usually equivalent to doubling the time."
Is that an engine might take one thousand times longer to calculate a node than another.  So the real ratio of strength is not calculated.
I think your method only calculates effective branching factor (which is, of course, a very useful calculation)

The same engine may calculate nodes with a different speed but usually searching twice more nodes by an engine means that the engine use twice more time even if not exactly twice more time.

Ghppn likes this post

Back to top Go down
Sponsored content





nodes equivalence of engines Empty
PostSubject: Re: nodes equivalence of engines   nodes equivalence of engines Empty

Back to top Go down
 
nodes equivalence of engines
Back to top 
Page 1 of 1

Permissions in this forum:You cannot reply to topics in this forum
ProDeo :: Computer Chess-
Jump to: