Interview with Robert Houdart

General discussion about computer chess...
Post Reply
Martin Thoresen
Posts: 386
Joined: Thu Jun 10, 2010 5:27 am

Interview with Robert Houdart

Post by Martin Thoresen » Mon May 13, 2013 4:48 pm

TCEC - Thoresen Chess Engines Competition
http://tcec.chessdom.com

lucasart
Posts: 201
Joined: Mon Dec 17, 2012 1:09 pm
Contact:

Re: Interview with Robert Houdart

Post by lucasart » Tue May 14, 2013 3:26 pm

Interesting answers by Robert. I especially like two of them, because they debunk some myths that amateurs really need to get out of their head, in order to understand how chess programms improve.
Some of the other top programs have Grand Masters in their teams, do you work closely with a Grand Master with regards to improving Houdini and what are your thoughts about this?
Robert:
There is no GM in the Houdini team, but I do receive a large number of suggestions from players around the world.
Being a strong chess player can help to point out positions where the engine doesn’t work well, but otherwise the development of a chess engine is more a software challenge than a chess problem – it’s more important to be a good developer than to be a good chess player.
It is typically a widespread notion that you have to be a good chess player to make a good chess program. In reality, there is absolutely no link whatsoever. For example, Marco Costalba is not a good chess player (according to himself, never had a chance to play with him) but he has been one of the most productive chess engine developper.
Please tell us a little bit about your testing environment – what is the typical procedure when you want to test a change in the code and verify that it is actually better?
Robert:
New ideas or code changes are first validated by playing against the current development version of Houdini. If the outcome is promising, the new version plays a tournament against 7 or 9 different opponents. In both stages typically 10,000 to 50,000 games are played, depending on the results. I use 2 servers (16-core and 32-core) to play about 100,000 games per day, with each game taking about 20 to 30 seconds.

From time to time I perform a regression test at longer Time Control against the same 7 or 9 opponents. This allows me to see whether improvements scale well to longer time controls, and to identify possible unexpected side-effects.
To improve a chess engine you need to be a good engineer, which means two things: (i) know your tool well (programming) (ii) develop a rigourous and scientific approach (eg. signature checks to prove the absence of bugs, or using advanced statistics like SPRT for efficient testing etc.)

Anyway, this discussion could be expanded ad vitam eternam, but I hope it will teach some things to the general audience of these forums, most of whom are not programmers and don't know much about how chess engines work.
"Talk is cheap. Show me the code." -- Linus Torvalds.

hyatt
Posts: 1242
Joined: Thu Jun 10, 2010 2:13 am
Real Name: Bob Hyatt (Robert M. Hyatt)
Location: University of Alabama at Birmingham
Contact:

Re: Interview with Robert Houdart

Post by hyatt » Wed May 22, 2013 10:05 pm

Anything dealing with Houdart is better than a soap opera. On CCC he chastised Vida for supposedly having parts of Houdini in Critter, and threatened a "cease and desist" letter. This from someone that copied one of the RoboLito family and claim it to be his original. It just doesn't get much funnier than this, when you think about it... I have to wonder what that letter would say. Something like this:

Richard:

This is a cease and desist request, requesting that you stop using any parts of Houdini you are using, even though those parts were copied from Robolito into Houdini, before you copied them from Houdini.

User avatar
Robert Houdart
Posts: 180
Joined: Thu Jun 10, 2010 4:55 pm
Contact:

Re: Interview with Robert Houdart

Post by Robert Houdart » Wed May 22, 2013 11:49 pm

hyatt wrote:Anything dealing with Houdart is better than a soap opera. On CCC he chastised Vida for supposedly having parts of Houdini in Critter, and threatened a "cease and desist" letter. This from someone that copied one of the RoboLito family and claim it to be his original. It just doesn't get much funnier than this, when you think about it... I have to wonder what that letter would say. Something like this:

Richard:

This is a cease and desist request, requesting that you stop using any parts of Houdini you are using, even though those parts were copied from Robolito into Houdini, before you copied them from Houdini.
There is a big difference between:
1) Using code and code elements that are in the Public Domain and can be used by anyone in any engine (including by Vida in Critter).
2) Using code and code elements that are not Public Domain, were obtained by RE'ing Houdini 1.5 and Houdini 3, and are now being reused in a competitive engine Critter.

Point 1 is OK, Point 2 is illegal in the EU.

I strongly encourage Vida to release a Critter version that does not violate the laws of his country. I want to make it very clear that we may or may not tolerate the n-th Critter version that contains code and code elements directly taken from Houdini by RE. At some point we will have no other choice than to act with the tools that the EU Laws provide. The sole purpose of my current intervention and of a possible "cease and desist" letter is to prevent this from becoming a necessity.

Note that your attitude is remarkably inconsistent. For over a year you have written *literally* thousands of forum posts about the author of Rybka violating some ICGA tournament rules and you have requested the intervention of the FSF. Now you seem to encourage Vida to commit illicit actions that could cost him dearly.

hyatt
Posts: 1242
Joined: Thu Jun 10, 2010 2:13 am
Real Name: Bob Hyatt (Robert M. Hyatt)
Location: University of Alabama at Birmingham
Contact:

Re: Interview with Robert Houdart

Post by hyatt » Thu May 23, 2013 4:06 am

1. I have not been inconsistent. Not seen an entry by Richard in an ICCA event.

2. the rating lists seem to be "anything goes" nowadays, which is their right to do...

3. I suspect one will find significant parts of Rybka in Robolito, if one looks enough, making it not really "public domain" at all...

I actually would like to see consistency from EVERYONE. Looks like that is a pipe dream today...

Post Reply