Author: Dan Homan
Date: 12:53:21 02/16/99
Go up one level in this thread
This is a very difficult issue. As an amateur chess programmer who wrote his program from scratch, I can tell you that this whole issue makes me angry. The thought of some jerk copying crafty's source, making some (perhaps extensive) modifications to the eval and search, and passing the program off as his/her own is very discouraging to those of us who do the work ourselves. The greatest gift crafty gives to someone who wants to have a strong program without doing most of the work themselves is its' infrastructure. It has tablebase support, multi-processor support, opening book/building/learning, an advanced bitboard engine, positional learning, hash tables, winboard support, etc... etc.... etc... All of these components are state-of-the-art, highly tuned and debugged, and represent years of dedicated effort. In addition, crafty provides an equally advanced search and evaluation. In my opinion, producing the infrastructure for a chess program is the hardest part and takes the greatest amount of time and patience. By copying crafty's code, authors (using the term rather loosely) avoid 98% of the work. Sure, they can spend time to modify/test/tune the evaluation. They can even optimize the search and extensions to fit the eval changes they made. The end result might play much different (and perhaps better) than the original they started with, but it is still 98% crafty. The Bionic-Impakt people have admitted from that start that they do this. Honesty is great, and I think that the program might grow over the years to be their own. However, the idea of a programmer doing this in a secret way, such as Voyager is accused of, is particularly angering and discouraging. The fact that this kind of behavior is so angering means that anyone who is accused of doing this will wear a black mark whether they are guilty or innocent. Because of this, I agree that strong evidence is necessary. The author of Voyager admits (according to your post) to using crafty's book and hash code. This, alone, is a significant piece of work taken from crafty. To make efficient use of the book and hash code, however, it is clear to me that a significant amount of the crafty infrastructure was also likely to have been copied. One cannot just lift hash and book routines from crafty and drop them into an otherwise original program. Bob can answer this better than I can, but these routines are usually integrated with other parts of the infrastructure in such a way as to make such a cut-n-paste difficult. (It also begs the question: if the author wrote the rest of the infrastructure of an advanced program, why borrow crafty code for the hash and book routines?) Evidence from game play is problematic because it is so easy to change the eval (and search) to produce different results. Changing them to produce better results (as you note that Voyager seems to produce) is much harder, but certainly not impossible, especially if the author is smart, systematic, and performs careful testing of new ideas. I think that getting to the bottom of the Voyager case is going to be very difficult unless Bob has some pretty clear-cut evidence. - Dan P.S. Another problem with this whole thing is that any new strong program that has a style similar to crafty's may be accused of this.... If an accused author wants to absolutely clear his/her name, they have to cough up their source code. However, if the author has new advanced ideas, they probably will not want to do this.... The end result may be to discourage all new original programs: The weaker ones could feel like they can't compete with what might be "hidden" crafty clones and the stronger ones might be afraid of being accused of copying crafty's source. What a mess.
This page took 0.02 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.