Why cant the computer on here ever say anything nice ?

#10 Even so, measuring complexity isn't quite as simple as measuring depth (or time/nodes, or hash table size). There are all kinds of factors which make a move difficult to find:

For example, engines with hash tables may find this simple:
http://chessprogramming.wikispaces.com/Corresponding+Squares

There was a study done by Matej Guid and Ivan Bratko on calculating complexity of the position.

The approach is simple and logical, I tried this before and a variation of it, and it is effective.

Check page 67 and 68.

http://en.chessbase.com/news/2006/world_champions2006.pdf

#12 Hm, last time I read that I dismissed it on the grounds, "What is obvious for an engine is not always obvious for a human, and vice versa."

But that aside, the Guid-Bratko technique seems reasonable (though one-dimensional) and I haven't seen any better alternative.

Humans can sometimes match the best move of engines. It is on this move that human deserves to get that exclamation marks. But it is not that easily given, if the engine experienced a certain level of difficulties thru calculation of complexity number to arrive at that best move then give that exclamation mark to the move of human.

Engines today are getting stronger, it is interesting to retry their algorithm. Perhaps the complexity number is lower as stronger engines can find the best move earlier and does not vary its best move much as it gets deeper.

sometimes a move deserves an exclamation even though it is not deep, and it is only slightly better than the alternatives.
It might not even be objectively better, but it leads for example to stable advantage where the alternatives demand from the player to be precise.
Basically you try to answer this question in code: What moves deserve an exclamation mark?
And your answer is: moves that are either deep, or far better than the alternatives
But this is not the correct answer :)
With your algorithm the program would give some exclamations marks a human would never give, it will give some that also humans would give, and it would not give some when a human would
This is a hard problem which I think you do not fully understand, but it is so hard that it might not be completely solvable so do not feel too bad about it:)
One easy incomplete solution that has been used in the past is to give "!" to the correct sacrifices.

@15

The solution I am suggesting here is a system with an engine to auto annotate human moves with ! or ! or !?.

We already knew the engine can auto annotate ?, and ??.

The TS is asking why not ! and !!

When the move of human and the best move found by the annotating engine is the same why not put ! or !!, after all the engine is the boss :-))

Not in all cases as there are easy moves. So we come up with complexity of position calculation, use of multipv analysis to find the score gap between best and second best, iteration score observation and others.

It does not matter much if the human would agree or not on the exclamation marks added by the engine to his move, because it is the engine that is annotating the game, the human should respect the annotator.

Ferdinand Mosca has produced "Chess Game Analyzer" completely free (it gives "!" too), see here : http://www.talkchess.com/forum/viewtopic.php?p=654864#654864

#16 I question lichess' computer analysis use of "?" and "??" which only account for raw evaluation difference. Dropping a pawn in an equal position is more important than dropping a bishop when already up a queen.

sorry But I choose not to respect the annotator, when the annotator is just bad, as your suggested way would be, but anyone is free to do as he pleases. You can even admire paintings created by computers if you like and say to anyone having disagreements about the quality of the painting : respect the creator

hi all, I'm a researcher in physics, complex systems.
chess is complex enough to stimulate our scientific curiosity so that with my team we have been analysing tons of GM games. We spotted a certain kind of moves which deserve the "!!" mark and only with the help of SF5 (that was the best at the moment when we started the analysis). Obviously I cannot spoil the results before our research gets published. What I can say is that whenever these kind of moves happen, the player who did them has the probability of 85% of winning. We contacted two GMs (one of which you can find in one of my imported games, played last october during a simul) for their opinion. If they agree that there is a meaning in what we find then we publish our work, otherwise forget it. stay tuned.
Is there a way to get played games here on lichess without resorting to player search, say, aggregated by month as on FICS?