Dinosaurs and Mammals

[CRoberson]

The conditions used in many of the lists are nearly the same with the same flaws. Several of the competitors have noticed this and optimized their efforts to those specific conditions.

A new effort in the form of TCEC at least handles longer time controls and faster hardware with more memory, but more is needed.

o Forced openings: bad. Some of us code our programs to play certain styles and are thus crippled by a forced opening. Our books could be geared for our preferred styles and still be not too deep. There are openings where stopping at 8 moves is too shallow. Guaranteed that GM's know their openings deeper than 8 moves. Of course, there should be a depth limit.
o Early adjudication: bad. I've seen programs lose or draw won games in online tournament games due to bugs that were never seen in the rating lists competitions due to early adjudication.
o Too small transposition table sizes.
o Pondering disallowed. I can understand this on single CPU machines.
o Slow computers.
o Lack of long TC's: bad.

Some claim that their conditions lead to finding the best analysis engine, but I disagree. Many times I will let programs analyze a position over night, so several hours on a position. I don't see how fast TC's can predict that type of analysis especially when you consider that some search algorithms that work well in fast TC's don't scale well at longer TC's.

[mvk]

Some claim that their conditions lead to finding the best analysis engine, but I disagree.

The "analysis" argument is bogus. If those who claim this are serious about that, they would play with a set time per move, and not let the players accumulate unused time (and maybe even clear the tables between moves).

I agree with your remarks about openings. They are part of the game. Same as knowing the identity of your opponent and his strength. Also using ponder time wisely is a part of the game.