Joined: 16 Dec 2010
|Post subject: Do not let marketing spoil computer chess Posted: Sat Jun 09, 2012 7:10 am
|It has been a relatively long time now that the landscape froze at the top of the hill.
Houdini 2c is still the king, the most feared contestant, Komodo 4 MP,
is still somewhere in the Aether or in the mind of its creator and, as the top is not far ahead, the other ones considered that their best interest is just to wait in silence keeping their own progresses for themselves till they have accumulated enough to become the new king.
The current situation, with the next Houdini scheduled not before September, has the unfortunate effect to block further progress toward the conquest of the top and the conquest of the top is just a rephrasing of what is the essence of modern computer chess : to have an ecosystem of cross hybridized engines collectively fostering better and better chess.
In the old Rybka 2 days, Vas was issuing a lot of versions and people were efficient in testing them, pointing out the weaknesses and advantages of each one and the pace of progress was steady. But Rybka was so much ahead of the rest of the field that Vas forceful releases policy didn't matter much for the collective advancement of computer chess. Vas just achieved to accelerate his progress by using the collective force of (beta) testers but, as far as I know, didn't receive much back from other engines creators as they didn't decide to play the same game by throwing at him their minute answers to his incremental updates.
Now that situation was a nightmare for the well established ELO rating list testers and they started to complain. Vas answer was predictable:
he ceased to produce almost weakly updates and the field froze till Rybka 3 stunned the world.
With Rybka 4 being only a modest improvement over Rybka 3, the other contestants had hope again to be able to backlash. And, indeed, there was a remarkable race again to become the new king of the hill. This race was helped (some would say spoiled) by the Robbolito/Ivanhoe underground work. But from an external point of view, the result is clear, there was remarkable and steady progresses with all those Robbolito/Ivanhoe themselves being cannibalized by some of the most promising new contestants. This is the power of cross hybridization on an open field, it promotes evolution not unlike genetic evolution among competing individuals.
But here came marketing once again. Recognizing the fact that pushing to the public scrutiny new incremental versions helped the competition to adapt and progress itself, the major players, once again - as did Vas with Rybka 3 - decided to keep their innovations for themselves till they had a reasonable hope to have enough of an advantage to become the new king of the hill. Unfortunately to accumulate such piecewise advantages takes time and the end result is a slow down of the collective progress of chess.
Is it what we want? I say no!
So, here is my proposal: Mr Dailey, Houdart, Vida, Rajlich, please consider to agree to once again release more incremental versions. Not weekly updates, which would be a delight for us but a nightmare for the testers and marketing people but, say, a new incremental version every 2 months, even if it's only for a marginal progress (e.g. 5 ELO).
This will keep the field entertaining, somewhat level off the competition (by no one hiding in the forest furbishing his arms) and above all, this will be the best scenario IMHO to achieve what is the essence of the computer chess art: to foster continuous and steady improvement at the quality of chess games played by computers for the pure enjoyment of every Caissa lover.
Per ardua ad astra