Houdini 3 released

[Modern Times]

I expected 50 elo better, as claimed by Robert. I see we are not so far right now.
But +40 would be satisfying, don't forget consumers pay for the product.

Best Regards

There is no absolute number. Robert said "about" 50 Elo, because the number will vary depending on the testing conditions used, which can differ very significantly.

[Lion]

Well for now, he is pretty close on the IPON rating... after 1983 games, shows up at +46 ELO compared to H2c.

A nice jump for Houdini 3 !

rgds

[Michel]

Tactical mode probably works well when there is tactics to be found such as in a test position.

Obviously Houdini cannot know this in advance....

Hi Robert.
Congratulations and many thanks for a wonderful engine!
Houdini is a piece of art.
I have 2 questions:
1. not important, but still interesting to know - what is the pronunciation of Houdini? Is the 'h' actually pronounced or not? It is logical that the name of the engine is patterned after the author's name, and Houdart is obviously a French name.
2. What made it impossible to integrate the tactical mode into the default engine? Would you plan to do so in the future?

Best regards,
Lyudmil

1) In Dutch the "H" is pronounced like in English, but of course I cannot tell anyone how the name Houdini should be pronounced, the name existed long before the engine.

2) By separating the Tactical Mode I could add more focus on tactical moves and some clever search "tricks" that work well for tactical positions but not for normal positions.
The Tactical Mode ended up about 25-30 Elo weaker than the default.

Robert

Thanks, Robert.
So you are actually a Flemish?

When I asked about the tactical mode, I thought that it would not be very reasonable to expect from users to switch back and forth to tactical and normal mode, moreover that most users would not have objective criteria to know with certainty which position is better checked with the tactical mode and which with normal mode. Was not it really possible to introduce yourself such a clever system of criteria so that the engine knows when to use the one or the other mode?
That would certainly make much sense, just my couple of cents.

Best regards,
Lyudmil

[Lyudmil Tsvetkov]

Test positions are normal game positions.

Tactical mode probably works well when there is tactics to be found such as in a test position.

Obviously Houdini cannot know this in advance....

Hi Robert.
Congratulations and many thanks for a wonderful engine!
Houdini is a piece of art.
I have 2 questions:
1. not important, but still interesting to know - what is the pronunciation of Houdini? Is the 'h' actually pronounced or not? It is logical that the name of the engine is patterned after the author's name, and Houdart is obviously a French name.
2. What made it impossible to integrate the tactical mode into the default engine? Would you plan to do so in the future?

Best regards,
Lyudmil

1) In Dutch the "H" is pronounced like in English, but of course I cannot tell anyone how the name Houdini should be pronounced, the name existed long before the engine.

2) By separating the Tactical Mode I could add more focus on tactical moves and some clever search "tricks" that work well for tactical positions but not for normal positions.
The Tactical Mode ended up about 25-30 Elo weaker than the default.

Robert

Thanks, Robert.
So you are actually a Flemish?

When I asked about the tactical mode, I thought that it would not be very reasonable to expect from users to switch back and forth to tactical and normal mode, moreover that most users would not have objective criteria to know with certainty which position is better checked with the tactical mode and which with normal mode. Was not it really possible to introduce yourself such a clever system of criteria so that the engine knows when to use the one or the other mode?
That would certainly make much sense, just my couple of cents.

Best regards,
Lyudmil

[Laskos]

Test positions are normal game positions.

No.

[Houdini]

Thanks, Robert.
So you are actually a Flemish?

When I asked about the tactical mode, I thought that it would not be very reasonable to expect from users to switch back and forth to tactical and normal mode, moreover that most users would not have objective criteria to know with certainty which position is better checked with the tactical mode and which with normal mode. Was not it really possible to introduce yourself such a clever system of criteria so that the engine knows when to use the one or the other mode?
That would certainly make much sense, just my couple of cents.

Best regards,
Lyudmil

Yes, I'm from the Dutch-speaking part of Belgium.

You can compare the Tactical Mode to what a human does when he/she tries to solve a test position in a chess magazine or on a chess forum. He/she will scan for tactical, forcing moves, threats, sacrifices etc. That way a lot more tactical solutions can be found than what the player would actually see and decide to play OTB during a normal game.
The "this is a test position" message changes the search strategy.

Houdini's Tactical Mode works exactly the same, very human-like: it will spend a lot more time looking for tactical moves, threats etc.
In most positions this doesn't provide any benefit - most of the time there is no tactical solution available and all the focus on tactics is just wasted.
But in test positions (selected because there IS a tactical solution) it works extremely well.

Robert

[Lyudmil Tsvetkov]

Thanks, Robert.
So you are actually a Flemish?

When I asked about the tactical mode, I thought that it would not be very reasonable to expect from users to switch back and forth to tactical and normal mode, moreover that most users would not have objective criteria to know with certainty which position is better checked with the tactical mode and which with normal mode. Was not it really possible to introduce yourself such a clever system of criteria so that the engine knows when to use the one or the other mode?
That would certainly make much sense, just my couple of cents.

Best regards,
Lyudmil

Yes, I'm from the Dutch-speaking part of Belgium.

You can compare the Tactical Mode to what a human does when he/she tries to solve a test position in a chess magazine or on a chess forum. He/she will scan for tactical, forcing moves, threats, sacrifices etc. That way a lot more tactical solutions can be found than what the player would actually see and decide to play OTB during a normal game.
The "this is a test position" message changes the search strategy.

Houdini's Tactical Mode works exactly the same, very human-like: it will spend a lot more time looking for tactical moves, threats etc.
In most positions this doesn't provide any benefit - most of the time there is no tactical solution available and all the focus on tactics is just wasted.
But in test positions (selected because there IS a tactical solution) it works extremely well.

Robert

Thanks, Robert, for explaining.
If the tactical mode rates some 25-30 points behind normal mode,
than there must be a fair number of positions (maybe you have some statistics in terms of percentage points) in which the tactical mode finds a better solution than normal mode, without knowing it must be deliberately activated. Maybe there is some pattern about such positions that could be used to activate it while using normal mode by default. This would, of course, improve the overall play of the engine.
I do not think the knowledge that there is a specific kind of solution would help enormously in finding it. Most problems run like 'white plays and wins' or 'black plays and draws' and the solver would just search for an optimal move.
I am sorry if I am a bit stubborn on this, I just think it would be wrong to waste an opportunity to improve the overall tactical strength while keeping the advantages of normal mode.

Best, Lyudmil

[Houdini]

Maybe there is some pattern about such positions that could be used to activate it while using normal mode by default. This would, of course, improve the overall play of the engine.

This optimal solution with highest playing strength already exists, it's the default engine playing style. It provides the best balance between quiet moves and tactical moves.

Most problems run like 'white plays and wins' or 'black plays and draws' and the solver would just search for an optimal move.

The whole point is that this kind of position ("X plays and wins") doesn't occur very often in a real game.

Robert

[Lyudmil Tsvetkov]

Maybe there is some pattern about such positions that could be used to activate it while using normal mode by default. This would, of course, improve the overall play of the engine.

This optimal solution with highest playing strength already exists, it's the default engine playing style. It provides the best balance between quiet moves and tactical moves.

Most problems run like 'white plays and wins' or 'black plays and draws' and the solver would just search for an optimal move.

The whole point is that this kind of position ("X plays and wins") doesn't occur very often in a real game.

Robert

I understand.
But there certainly are positions more appropriately handled by the tactical mode.
Positions that would be better handled by a tactical try might have something in common: for example, large number of pins for both sides, large number of open and semi-open files for both sides, large number of attacked pawns and pieces for both sides, large number of attacks along an x-ray with consideration of pieces of different capacities - diagonal pieces attacking rooks and knights or linear pieces attacking bishops and knights, large number of squares controlled by pieces in between an enemy piece and an enemy piece defended by it (deflection possibilities), unsafe position of the king (square where it is placed and number of ps around it), etc.
At least most positions from tactical tests exhibit similar features.

Best, Lyudmil

[Houdini]

Lyudmil, thank you for explaining, it's indeed an interesting idea.

Robert