So, inorder to test this, I'm going to let different engines play each other for 7 moves with opening book turned off. The engines will be given 3 minutes to make each move. Let's see if the stick to the theory and play some theoretical lines.

So first it is Komodo's turn. Let's look at the moves.

Details of the opening played:

The engine played a very rare but yet a theoretical line called -

D51 Queen's Gambit Declined, Manhattan varition

I'll be testing Stockfish next.

in order to really know, we would have to remove the databases from machines and have them calculate the trees themselves. Part of the reason machines play so well is because they have all of our games to rely on. Chess Engines have massive opening and end game databases built in, but they have to decide which ones to use on their own. In short, the more we play, the better they get.

Are you saying that the 2100kB Komodo.exe file has all these massive opening and endgame tablebases in it? I don't think you can fit so much in a 2MB file. 

This is what Stockfish played.

Opening Details :

I've never even heard of this line. It's called -

C05 French, Tarrasch, Botvinnik variation

I'll be testing Houdini 3 next.

Theory is super massive. Of course they'll play theory... whether or not it's unambitious and bland is another story.

The first opening I think would more commonly be called the ragozin.

Not the mine I presume! :D

Yes, there are so many opening lines, any decent chess engine would happen upon one just by playing sound moves.  Even total patzers such as myself chance upon them.  We don't know any openings, but so long as you stick to the basic principles, at least the first few moves are going to be a part of some opening, somewhere.

There was a recent discussion in another thread speculating on how much an engine benefits from using an opening book and endgame tablebase. Because of that discussion, I decided to run some engine tournaments in an attempt to answer that question. The tournaments are still ongoing, but I've been watching the engines in action. Of course, the engine without an opening book must calculate its responses, and that takes time, but it almost always will come up with a book line at least 5-7 moves deep. Normally the line will be maybe 10 moves deep, and at times the line can be 12-15 lines deep. And that's being done at a relatively fast time control of 40 moves in 4 minutes.

You'd think so, but that's essentially how they program chess engines now adays. The engines don't draw the trees themselves anymore... or so I've been told haha.

The way almost every engine works is that it accesses the opening book in the opening phase of the game, and it accesses the endgame tablebase files near the end of the game. Neither the opening book nor the tablebase files are contained within the engine.

Now, it IS true that opening books are usually derived from a large game collection. But the opening book itself consists only of a move tree and weightings for each branch of the tree. Once the book is generated, I think the bookmaker will usually tweak the book by adding or culling lines and changing some move weightings.

If that's the case then why are these computers all playing recognized openings? Is that coincidence?

It is not coincidence, it is chess evaluation of positions. They use the same criteria to evaluate the Grand Masters use, just that contrary to the masters they don't blunder in calculations (tactics) and they are much faster.

As long as you play reasonable moves, you'll be playing theory.

I've played over 10 moves of "book" without knowing it until after the game. Lots of people do this.