Computer Endgames

Hello fellow intelligent chess.com members... Is it true that computers have extremely intelligent endgames?  If so, then their middle games are weakest, due to their pre-programmed openings and strong endgames?

So here's the question:  Which part of chess are the strongest computers weakest at, the openings, the middle game, or the endgame?  

Of course, your engine needs to be equipped with the latest tablebases and opening book theory money can buy :)

Endgames that are within a tablebase horizon (upto 6 or 7 pieces these days?) are considered "solved" for a computer that uses said tablebases.   What this means is that there's no more calculation/evaluation  ... the computer has all move sequences from here on end mapped to the point of winning/drawing.

Endgames outside this horizon are still up for grabs.

It's also a given that opening books tend to make them invincible at conventional main line theory.

So to list the pros and cons, I could probably say =>

Table-based endgames => Solved

Other endgames => Up in the air / depends on the engine.

In Book" Openings  => Solved until deviations from theory / novelties occur.

dry "tactics free" middlegame positions => Up in the air / depends on the engine.

Tactical / non-quiescent positions => Really strong to the point of near infallibility.

Edit: Of course, your engine needs to be equipped with the latest tablebases and opening book theory money can buy :)

Thanks for the comment, I learned a lot, and also thank you for the edit suggestion!

Engines are generally considered to be precise but unremarkable in openings because they simply read moves from a book and do not calculate anything.

Once a deviation occurs, the engine may be fairly weak early in the game where there are few clear tactical ideas and the strategic decisions rely on deeper understanding of opening theory and experience with games from those positions. This is the second weakest part of an engine's game.

Once the middlegame proper is entered, chess engines are usually very strong--so strong, in fact, that few grandmasters are ever able to draw them with one exception. Engines are notoriously weak in blocked positions. Engines don't understand the significance of having a lever in a blocked position or how to rearrange their pieces for a breakthrough, nor do they know how to achieve draws by closing up the game. These rarely occur (especially in the openings programmers choose for chess engines in their opening books), but they can sometimes be forced. However, the vast majroity of the time, the middlegame is the second strongest part of an engine's game.

As the endgame approaches, the engine's skill decreases dramatically, perhaps by a thousand Elo points or more. If there are more than around eight men left, the engine is unlikely to find reliable ways to simplify to a won subendgame. This is because most endgames are too complicated for brute force analysis while also deficient in typical calculated tactical sequences. The strategy is also intuitive in principled and is hard to code. Engines know some basics, like centralizing kings and getting passed pawns, but they are far from the superGM-chewing monstrosities they were in the middlegame. This is the weakest part of a computer's game.

As the number of pieces drop, the engine begins to become more powerful. WIth seven men, it usually is excellent, and with just six, it is perfect*. Literally perfect. Scary perfect. It will know from any position the exact right move to checkmate you as fast as possible. It is that good. This is because reverse induction has allowed us to construct endgame tablebases for all positions with six men or fewer (and also for certain 7 man positions). Even with seven men, it is likely that the computer can calculate out sequences that result in won six-man endgames. This is, for obvious reasons, the strongest part of an engine's game.

*Technically, engines use Nalimov tablebases which do not include castling, and therefore are not quite "perfect." However, this is an almost totally irrelevant omission.

In summary, engines are strong in the beginning, middle, and end, but weaker in the early middle and late middle.

oh wow, you're seemingly extremely knowledgeable on that topic, thanks so much for your "beyond explanation" response for a low-demanding question.  That was outstanding, I learned so much!!! :)