English Chess Forum

Someone alerted me to a free online endgame database, so out of curiosity, I experimented for a couple of minutes with a few positions of rook plus a and h pawns v rook, knowing that this can often be a draw if one knows the right technique in the right positions.

This position (with Black to move) is a draw:

1k6/7R/8/8/7P/8/PK6/5r2

Slight changes to the position make a difference (e.g. with the White rook on e7 and Black to move, having the pawn on h5 is a White win, but it is a draw with the pawn on h4).

The position below (also with Black to move) is a win for White, even though one's immediate instinct is that Black's rook is better placed and White's rook worse placed than in the position above. (Once you look at it a bit more closely, it is not too difficult to figure out why this immediate instinct is incorrect, but it is instructive).

1k6/8/8/8/R6P/8/PK6/7r

It was an experiment which brought mixed feelings. It reinforced my pleasure at the subtlety and complexity of chess, but it also made me a little sad to see something solved so cold-bloodedly; rather like being shown the technique for a magic trick. It also made me wonder how long it will be before there is a database solving this position (White to move):

rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR

Perhaps neither as far-fetched nor as distant as current scientific opinion suggests.

Martin Benjamin wrote: It was an experiment which brought mixed feelings. It reinforced my pleasure at the subtlety and complexity of chess, but it also made me a little sad to see something solved so cold-bloodedly

Endgame tablebases and for that matter the assessment of the initial position as mostly equal with perhaps a slight edge to White are conditional on the idiots not taking over and declaring stalemate a whole or partial win.

Supposedly the "I go there, he goes there" search approach, whilst successful in chess, doesn't work for Go. What might work for Go is the approach of playing any number of games extremely quickly from the position being analysed. The idea is to find the move with the best chance of success. I would suspect that is the only way that the initial position in Chess can be anywhere near solved. I lack confidence in Chess960 as I conjecture there are certain positions, which if not totally won for the player moving first, set immediate traps where plausible moves just lose.

Martin Benjamin wrote: It was an experiment which brought mixed feelings. It reinforced my pleasure at the subtlety and complexity of chess, but it also made me a little sad to see something solved so cold-bloodedly; rather like being shown the technique for a magic trick.

I tend to agree with you Martin. I have never looked at these tablebases, but at your or my playing levels, would it make any difference in practical play? I mean, we would probably never be able to memorise these nuances (eg drawn with the pawn on h4, won if its on h5), so would not know whether to simplify into a given ending or not. I see it as more of an academic exercise.

If you could post a link I might try to have a look, maybe while I am waiting for some paint to dry!

Graham Borrowdale wrote:I have never looked at these tablebases, but at your or my playing levels, would it make any difference in practical play?

It's a very practical way of covering a lot of ground when you use them to test an actual game. For Rook endings, the same material will likely be in Levenfish & Smyslov and other more recent works, but actually finding the nearest parallel takes some tracking down. It's also a refresher in the actual techniques. You might recall from the books that a particular position is won or drawn, but the method of doing so still has to be uncovered.

There was a game by Keith Arkell against one of the Ledgers in the past year or so where he defended R v R and two passed Knight and Rook pawns. Intrigued by this apparent rewriting of theory, I ran through it on the tablebase. The draw was only available right near the end where a stalemate trick became possible as the defending King had been left with no squares. When I had a similar ending in a recent game, I made sure to not allow this possibility.

The online tablebase I use when analysing games or following games online is this one:

http://www.shredderchess.com/online-che ... abase.html