Btw for the first couple moves we already have bookmoves. They are extensively studied. So we can use those and continue from there.
The Mega Houdini Premove Project
Sort:
Apr 16, 2014
0
#3
Houdini’s a perfectly decent engine but synchs poorly with most interfaces. In the analysis lines you get repetitive best moves and spacing problems and it doesn’t support certain cool features, like the dial evaluator you get in Deep Fritz. Make Houdini your third choice for an engine. Also there’s no support to speak off – just Bob Houdart, over in Europe and not at home.
ElKitch wrote:
Had an idea that will never become reality, yet still its nice to think about it and see what the possibilities are.
The project is called the Mega Houdini Premove Project and the goal is to make a huge database of premoves going as deep as possible into the game.
The way to do that is to let people make their computers available to do houdini calculations. This is called a 'distributed computing project'.
The starting point wil be the five most used first moves for white. Chess.com's game explorer says they are:
1.e4
1.d4
1.Nf3
1.c4
1.g3
For each of those moves Houdini will be calculating the 3 best answers. And in respond to each of those moves Houdini will again calculate the 3 best answers and so on.
Yes, the number of possible positions will go astronomical pretty quickly. Yet still I think that if a couple thousand computers are pondering on it for a couple years the amount of premoves will grow and we will go deeper into creating an ever more immense premove database.
Q: How deep should Houdini think about each position to find the next best 3 moves?
A: I'll leave this one to Houdini experts. When can it be said that chances are 99% that houdini has found the best 3 answers to a move?
Q: What's the point of doing this?
A: Creating a huge 'bookmove' database. You can play the bookmove database and if you make a move that the database cannot answer then you can request to compute the answer to that move.
Q: Impossibru! This will not get anywhere. There are more positions than atoms etc.
A: I do not claim that we will find answers to every position all the way to the endgame. By only picking the best 3 answers we cut away huge branches. And it's not the goal to get to the endgame. The goal is to get as far as possible into the game. If this project continues for 20 years, or even 60 years then we will get far into the game.
Well.. now I'll just brace for incoming comments saying this project will not get anywhere :)
Sounds interesting, it will progress at a faster rate than Stanford Folding@Home (definitely)
If we use another engine rather than Houdini, we may achieve better results. A free master-level engine would make best.
If we get the whole chess.com community to participate, we might get to move 14 or something like that. By the way, can you show us the bookmoves calculated so far?
Apr 16, 2014
0
#5
Had an idea that will never become reality, yet still its nice to think about it and see what the possibilities are.
The project is called the Mega Houdini Premove Project and the goal is to make a huge database of premoves going as deep as possible into the game.
The way to do that is to let people make their computers available to do houdini calculations. This is called a 'distributed computing project'.
The starting point wil be the five most used first moves for white. Chess.com's game explorer says they are:
1.e4
1.d4
1.Nf3
1.c4
1.g3
For each of those moves Houdini will be calculating the 3 best answers. And in respond to each of those moves Houdini will again calculate the 3 best answers and so on.
Yes, the number of possible positions will go astronomical pretty quickly. Yet still I think that if a couple thousand computers are pondering on it for a couple years the amount of premoves will grow and we will go deeper into creating an ever more immense premove database.
Q: How deep should Houdini think about each position to find the next best 3 moves?
A: I'll leave this one to Houdini experts. When can it be said that chances are 99% that houdini has found the best 3 answers to a move?
Q: What's the point of doing this?
A: Creating a huge 'bookmove' database. You can play the bookmove database and if you make a move that the database cannot answer then you can request to compute the answer to that move.
Q: Impossibru! This will not get anywhere. There are more positions than atoms etc.
A: I do not claim that we will find answers to every position all the way to the endgame. By only picking the best 3 answers we cut away huge branches. And it's not the goal to get to the endgame. The goal is to get as far as possible into the game. If this project continues for 20 years, or even 60 years then we will get far into the game.
Well.. now I'll just brace for incoming comments saying this project will not get anywhere :)