Stockfish over >3000 ELO rating!

I am sorry, but almost every one of your points has holes.

quote "1. An average game lasts 37 moves"
Is this from a GM game database, that includes all the early "draws by mutual agreement" including those where the outcome wouldn't change the outcome of the tournament, then yes it is probably 37 moves, but to compare it with anti-computer player games versus chess computers is simply an error, I think even my average moves per game versus computers would be like 45 to 50, since there are many many games in a far away endgame.

quote: "3. Humans generally get tired and tensed up and can't maintain the same quality of play throught the game"
In a single classical 2 hours per game many but not all humans will get tired. They can walk around and not concentrate on board all the time, also true classical chess is becoming more and more rare, with ever faster time controls taking precedence even in serious tournaments, I actually think that humans have better chances in shorter time controls against computers (of course not 1+0 like I like to play often) - and this shouldn't be considered any less of a chess against computer which can find good moves and never blunder in a millisecond.

quote: "With an opening database, an engine plays the opening perfectly until like 20 moves into the game "
This is just misinformation. These are best lines only, and most of them are not suitable for play against computer since they are highly tactical (they were designed for GM vs GM games where you have to make position crazy to avoid draw), the human player can generally steer game into much much less tactical games and that will be quickly out of book (like within first 6 to 10 moves).

quote: "With an endgame database"
End game database can only tell you true value of position with up to 7 chess-man (this includes kings and pawns), and only in some super-computer with many terabytes of storage, scanning which probably isn't lightning fast. 8 chess-man is only SCI-FI currently. Actually you should have learned that end game databases give very little to nothing for an engine, they are more for fun - scanning them in an active endgame position is a serious tradeoff.

quote:
"engine's endgame play is very strong as it searches very deeply like > 60 moves ahead"
This is just pure nonsense - I hope you meant 60 ply not 60 full-moves ahead, and even 60 ply could only be reached in certain endgame positions by best engines, mostly they will think like 20 to 30 ply deep (that's up to 15 moves and no more - but this is only in chess with large time control).

quote: "So, the only practical window where a human might get an advantage is in the middlegame"
I get advantage in around move 6 when I play the exchange variation of the Spanish game - all it takes to win there is to exchange down pieces and in some scenarios even that is not necessary, and that is not the only opening where I can win computer (I have won top engines also in Sicilian playing as white but that is much more rare). The human can get strategic advantage in the middle game or even the opening, force down exchange in to endgame and beat the clueless machine, which will understand that it is losing only when it will be far too late.
It takes to much text to counter your points so I will stop now.

this is one good example how dumb machines are - that was X3D Fritz against Kasparov little more than 10 years ago:
http://www.chessgames.com/perl/chessgame?gid=1269891

The game proved that they did not cheat (centaur chess), as any master player would have known to attack in king-side, but the computer was taught not to moves his king's pawns too much, and even if it weren't it still probably wouldn't know what to do there.

Again Kasparov is only so good against humans, he has spent whole his life learning how to make games more wild and tactical to be able to beat all those GMs since that is almost the only way, when your opponent "knows everything". And no-one is denying that computers are far better in tactics almost always.

Okay :)
The search depth displayed is in plies. Got it. You made some excellent points.

Btw there are sygyzy bases now which improve the engine strength by 10-20 elo. Maximum improvement was found with 5 men bases. They are about 1Gb in size.

I guess theoretically it is possible for a human to win against engines consistently if they train to play computers. Maybe we'll see someone do that in the near future.

There's no comparision between a modern chess engine and fritz X3D though. It is rated at ~2700 while stockfish has a rating of 3300 in ccrl ratings list. Mathematically speaking the probability of fritz winning is less than 1 in a million.

quote: "The search depth displayed is in plies. Got it. You made some excellent points. "
It is displayed in ply in pretty much all chess engines, also it is not true brute-force search, meaning it does not scan everything up to that ply - to put it simple, it is more or less a maximum depth it has reached, so it can really miss something in between.

In the best possible implementation the TB might improve by a few ELO that's true.

quote: "There's no comparision between a modern chess engine and fritz X3D though."
Again this is somewhat untrue. That rating of 2700 can't be compared to these 3300 nowadays the same way you can't compare Bobby Fischer's rating with Garry Kasparov. Also I'm quite sure that Fritz would win one in like 20 to 30 games, but this again misses the point, it does not matter that much really how many times faster an engine searches when playing against human, since times when humans could compete against engine in a pure tactical play are long over and the times when engines beat humans in strategic play are still very far away.

To put my idea as simply as possible - it is very possible to design your play to more or less force the game off tactical grounds and into purely strategic. I with my probable FIDE rating of 2100 shouldn't be able to draw 1 in a million games against top engines too, but if I really try and am in shape then I believe I draw around one in five doesn't matter against which of the top engines, sometimes even better.

No. The engine fritz x3d is (rather fritz 8 which is slightly stronger than x3d) can be made to play against stockfish 5 and other engines today and that's how rating list is generated. Not the same with humans I agree.

http://www.computerchess.org.uk/ccrl/4040/
Look at LOS between #1 and #2. It is 96.7% which means komodo has won 96.7% of the time against stockfish at those time controls.

Just analyze this game with stockfish. It will suggest strategically far superior moves for black.

The reason why you are able to draw a lot of games is because a computer doesn't want to take risks. This can be changed by setting the value of 'contempt' an idea of how strong an engine should assume it's opponent is.

"Just analyze this game with stockfish. It will suggest strategically far superior moves for black. "
I meant the game between fritz and kasparov.

Stockfish has shown excellent positional play even in closed positions. Take a look at this. (An article explaining a point I made in my previous post)

http://aoxomoxoa-wondering.blogspot.in/2014/09/the-myth-of-bad-positional-play-of.html

quote: "No. The engine fritz x3d is (rather fritz 8 which is slightly stronger than x3d) can be made to play against stockfish 5 and other engines today and that's how rating list is generated."
Of course it can be, but I believe it hasn't been released to public in any form, but unless I'am very much mistaken then it was just some modified form of the Fritz engine, you can see the latest Deep Fritz in the 15th position, meaning it will win quite often against the top engines too.

quote: "Look at LOS between #1 and #2. It is 96.7% which means komodo has won 96.7% of the time against stockfish at those time controls."
No it doesn't. LOS means likelihood of superiority. It does not donate win percentage or something like that, but likelihood of one engine being stronger than other.

quote: "The reason why you are able to draw a lot of games is because a computer doesn't want to take risks. "
This is somewhat true, but the contempt value generally just sets ratio of how much engine will try to avoid draw, not "risks". Engines generally don't take risks at all - meaning if it sees one move to be losing in one tiny very hard to find line it will discard that move. I believe there are some engines that have function of taking risks, but currently am not aware of one, and it is unlikely that it is some of the top engines.
In general it is possible to make those engines harder against humans - by making them avoid exchanges, and at the same time try to keep the game as open as possible and as many possible moves as possible, to somehow make the game more tactical, however you will find that avoiding exchanges and at the same time keeping the position open and tense is contradictory, but I would like to find an engine that is harder to draw or beat - actually I can tell you that I know which engines are - these are engines that a GM would beat often, because they don't calculate really deep and thus play more like a human which "takes risks" by playing very aggressively but at the same time taking risk of losing badly. I can give you example against Nagaskaki - it is generally a master level chess program, but on the machine I played against it plays International master level - which is collaborated by it beating Fide masters repeatedly (it plays on my server), and other FIDE master rating players saying it plays at IM level. I beat it very badly in a Sicilian gambit myself playing as white:
http://www.chesspastebin.com/2014/12/27/megawolt-nagaskaki5-1-by-valters/
The time control was 3 minutes and 3 second increment.

The latest version of fritz is 14 which which is way better than fritz 8 which was released around the time fritz x3d existed. So, I compared x3d to fritz 8.

I just said that because setting a high comtempt factor will make stockfish play a slightly inferior move to avoid a draw. This is a kind of 'risk' hoping its opponent to make a bad move later in the game which works most of the time against humans.

That link shows 503 error here. you can http://en.lichess.org/paste