Can an engine "understand" an opening ?

[Ponti]

I mean, without an opening book, can an engine make and follow correct plans of a given opening?

For example:

[d] rn1q1rk1/pbp1bppp/1p2pn2/3p4/2PP4/2N2NP1/PP2PPBP/R1BQ1RK1 w - - 0 8

This is a position from the Queen`s Indian Defense. There are many plans that white tryed here, and a great amount of GM games. Some positions like this one requires mainly strategy, not tactics (where engines rule). I know sometimes engines "discover" a new move, and this is being used by many GMs to play "novelties". Kasparov used many "computer novelties" in the opening during his career. Sometimes though, it just seems to be a stupid "computer move" that does not fit the opening`s strategy.

Here's another example:

[d] r1bq1rk1/1pp2pp1/1bnp1n1p/p3p3/2P5/PPNPPNP1/5PBP/R1BQ1RK1 b - - 0 10

This is Karpov's variation, English Opening. As the game progresses, somehow an engine`s move may seem stupid, and if you search the databases you can`t find that "idea". Many engines here after some moves try to play Na4, and Nf3-g1-e2-c3-b5 to get the bishop pair, eliminating black`s dark-coloured bishop. I`ve studyed many games in this variation and I may say, this is kind of a new plan here, as GMs don`t do that (the Knight at a4 seems to be particularly out of play). Common ideas for white here are to play e4 and Nd5, or play for d4 (Kh2, Ng1-e2 to support the advance d3-d4), or try king-side expansion with f4...

A group of masters from my town once told me that they often make mini (100 games) engine matches to try to discover new plans in an opening...

[Nimzovik]

No.

[Nimzovik]

This is why God created opening Books. Until a major paradigm shift occurs where computers can recognize 'Pablo Chess' the answer is no. I have spoken! You have heard. Go forth and teach!

[Uri Blass]

pablo chess is a small minority of all the variaty of openings
and I think that engines can clearly discover good new opening moves so the answer is yes.

[Dann Corbit]

Here is the computer analysis I have for those two positions.

[d]r1bq1rk1/1pp2pp1/1bnp1n1p/p3p3/2P5/PPNPPNP1/5PBP/R1BQ1RK1 b - - bm Bf5; pv Bf5 h3 Bh7 Bb2 Re8 g4 Bg6 Nh4 Bh7 Nf3; pm Bf5 {173} Bg4 {44} Re8 {25} Be6 {12} Ne7 {2} Ra6 {1} ; ce 0; white_wins 100; black_wins 72; draws 72;

[d]rn1q1rk1/pbp1bppp/1p2pn2/3p4/2PP4/2N2NP1/PP2PPBP/R1BQ1RK1 w - - bm Ne5; pv Ne5 Na6; pm Ne5 {170} cxd5 {119}; ce 40; white_wins 164; black_wins 81; draws 81;

[Nimzovik]

pablo chess is a small minority of all the variaty of openings
and I think that engines can clearly discover good new opening moves so the answer is yes.

Pablo's stone wall is not the only anti computer strategy in the chess cloud. There are others. As has been discussed several times closed positions and other Nimzoid like strategies are often effective. Eventually perhaps computers will understand the openings as chess is a closed system. Of course one could say that an opening book could 'understand' chess openings as well as they could contain ' All Possible good moves.' So you see it is a matter of time and approaches. It gets a little tiresome to hear people say Pablo chess is -' relatively irrelevant.' He has demonstrated his techniques against the best of the best repeatedly. A lot of draws to be sure. Then there is the time factors discussed ad nauseum. I assert that Pablo has discovered his niche, an achilles heel, in the current scheme of the computer chess world.