Into oblivion ?

I think that direction is only one...to solve chess completely! Can you imagine chess in 20-30 years?  All positions will be described and solved ,  thousand  times more powerful computers than today will need nano seconds to find best combinations. Opening books will be  40, 50 moves  or  even longer? Ending databases will  contain all possible variations for 12-15 pieces?     For sure chess won't be the same as they are today....

This question was answered over a year ago by SR here: http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?pid=320945

"Phase 1:
Imagine that we are in year 2040. Since year 2025 super strong +3400 elo engines have been widely available. Users have linked  these engines in clusters and super-clusters and used these clusters to develop a huge public opening data base using a special software SuPerIDeA.
Around 2030 something unfortunate began to happen: No engine loose when playing white in "safe mode". More specifically, single engines stopped loosing when playing white in "solid setting" with access to a solid-for-white online one-million terabytes opening data base.

This solid-for-white database contains mainly critical lines and the lines often run beyond move 40. With an efficient branching factor of 3 (benefiting from transpositions and only considering critical lines )  the number of critical lines in the data-base is roughly 3^40 = 10^19 = 10^6 terabytes.

The "solid-for white" opening book is far from "solid" from a human perspective. In fact at any giving opportunity white choose sharp lines which often leads to huge complications that typically fizzles out after a 10-15 moves in a tough middle game battle. Many lines in the solid-for-white opening book are deliberately inferior since they lead to positions that are easy to hold for white.

In 2035 a special price was offered to ANY chess playing entity that with black can win a game against a single solid engine with access to the special database. By year 2040 not a single cluster or chess playing entity have been able to claim the prize.

Phase 2:
We are now in year 2100. Major wars has miraculous been avoided! The computational power have increased and super computers are now using optics. The prize offered in year 2035 has still not been claimed. The internet clusters in year 2100 play with a staggering computing power of 10^9 of today's super-computers. Yet, none of these clusters have been able to beat a the solid "slow" year 2035 machine.
In year 2100 Computer chess is still thriving. However, computer chess with the ordinary rules starting from the usual standard position has been dead well before year 2050.

Phase 3:
We are now in year 2150. Classical chess has long been for dead for computer chess, however no-one has dared even begin to prove that white can hold the initial position. So some mathematicians decide to use computers to PROVE that its impossible to win the 2035-prize (i.e. its impossible to beat the solid year 2035 engine with black). Lots of computer generated papers are published and researchers are able to use computers to identify roughly 10^10 different types special types of positions. The conjecture is that white can hold any of these special positions unless the year 2035 engine claims otherwise. As an example it is conjectured that white rook +f,g,h pawn against black rook +e,f,g,h pawns is a draw if no clear advantage is detectable by the year 2035 engine.  So in 2150 this and 10^10 other similar conjectures  can be proved to ensure that white holds chess (i.e. prove that white is not in zug-zwang in the initial position).

Theorem: (year 2150) If each of the 10^10 highly plausible conjectures about various general types of positions are correct, then White is not lost in the initial position.

Phase 4:
We are now in year 2200. Many of the 10^10 conjectures had to be replaced by other more tractable ones.  By a combination of ordinary mathematical proof techniques and some very deep strategy stealing arguments(*)  (found by computer)  it has been proved that White is not in zug-zwang and loose by force in the initial position. This has not solved chess, but it is a very important step in doing so.

Phase 5:
We are now in year 2500. By use of much deeper analysis but essentially using the same methods as already discussed it is proved that black can hold the initial position. The new main idea is an introduction of machine generated generalized strategy stealing arguments that  apply to  various families chess positions. Thus chess is solved in the same way as checkers is solved today.

Theorem (year 2500): Classical Chess is a draw with perfect play

Of course even in year 2500 there will be many chess positions the computers fail to understand! These position are however not relevant for the proof.   Phase 4 and phase 5 might of course never happen, but i see no reason why the automated theorem proves of the future could not implement phase 4 and 5 (possibly much sooner than I suggested).

(*) The concept of strategy stealing can be explained by an example. Consider the following version of chess:

Modified chess Played like ordinary chess, but with the following modifications:
a) The player who make the last move wins. If the last move is mate this agrees with the current chess rules. Draws are not possible since the person who make the last move forcing a draw, win the game according to the rules of modified chess.

b) White (the first player to move) can pass on the first move, but otherwise the game is played as ordinary chess. 

Theorem: White has a winning strategy in modified chess

Proof: We use the following two facts:
1) Each actual game has a winner (i.e. draws are impossible)
2) The initial position is symmetric so the special pass rule allow the first player can "steal" any black strategy

It follows from 1) that either white or black has a winning strategy. If black had a winning strategy, while could pass on his first move and then "steal" blacks wining strategy (by pretending to be black). This would guarantee white a win which contradicts the assumption that black has a winning strategy. Thus we conclude that white have a winning strategy.

Notice that the strategy stealing argument says absolutely nothing about how white should play (e.g. pass, c4,d4,e4 etc.)."

Computer chess has reached it's apex and is now in decent. Junior has won the WCCC. Next year it will be Crafty. Soon engine enthusiasts all over the world will be demanding shitty engines! Don't miss out!!!

In what direction is computer chess headed for ?

In the direction that all easily believers (fanboys) still avoid the question of how the naive sheep must be protect against
the ravenous cloner wolves (Rybka, Houdini, Vitruvius) in the little realistic hope that one
day the wolves would discover the benefits of vegetarianism.