Monthly Archives: November 2022

Endgame tablebases

Leave a Reply

I’m going to start this post by quoting an extract from the Wikipedia entry, since it says what I want to say and I don’t see the point of paraphrasing it.

An endgame tablebase is a computerized database that contains precalculated exhaustive analysis of chess endgame positions. It is typically used by a computer chess engine during play, or by a human or computer that is retrospectively analysing a game that has already been played.

The tablebase contains the game-theoretical value (win, loss, or draw) in each possible position, and how many moves it would take to achieve that result with perfect play. Thus, the tablebase acts as an oracle, always providing the optimal moves. Typically the database records each possible position with certain pieces remaining on the board, and the best moves with White to move and with Black to move.

Tablebases are generated by retrograde analysis, working backward from a checkmated position. By 2005, all chess positions with up to six pieces, including the two kings, had been solved. By August 2012, tablebases had solved chess for almost every position with up to seven pieces, but the positions with a lone king versus a king and five pieces were omitted because they were considered to be “rather obvious.”[1][2] These positions were included by August 2018.[3] As of 2022, work is still underway to solve all eight-piece positions.

Tablebases sometimes come with a name as a prefix, e.g. Nalimov, Lomonosov, Syzygy. Eugene Nalimov is a computer programmer who generated the first tablebases in the 1990s. Lomonosov is the name of a Russian supercomputer. Syzygy is an astronomical term describing the occurrence of three celestial bodies in a straight line. No, I don’t know what that has to do with chess either. Syzygy 7-piece tablebases are the most recent. These are used by the International Correspondence Chess Federation.

What is the practical use of a tablebase in correspondence chess? In my experience the tablebase is a means of bringing a game to an end quickly and so achieving a draw or a win, when your opponent cannot or will not see this, and particularly when he or she has slowed the game down in order to delay the inevitable. The ICCF rules allow you to claim a win or draw by reference to a tablebase where there are no more than seven pieces on the board. Here’s an example from one of Barking’s games earlier this year.

White: Maurizio Mangiarotti (1958), Black: Rodney Barking (1915).
D45: QGD Semi-Slav.

Yes, I know there are more than seven pieces… This was the position after White’s 41st move. Earlier I had given up my queen for a rook and bishop. As all the pieces are on one side of the board, and there are no passed pawns, Black can hold the draw easily enough and in fact the evaluation of one of the chess engines I use is 0.00. However, White can continue the game for quite a long time just by probing with the queen.

As I’m not winning this position, there’s no advantage to me in prolonging it. What’s the quickest way out? Answer: sacrifice some material…
41…Bxh4! 42.gxh4 Rxg4+ 43.Kh3 h5

Despite the material difference, this position is actually drawn. Black has constructed a fortress. There is no way the white king can penetrate Black’s defences. I offered a draw, which my opponent accepted.

The point that’s relevant to this post, and obviously I checked this before going nuclear, is that the position is a tablebase draw as well. I would have claimed a draw by reference to the ICCF tablebase if my opponent had not co-operated. This is what the tablebase evaluation looks like:

Interestingly, the various engines I use all evaluate this position differently. The strongest of them has White at +0.81 (approaching a clear advantage, but not yet winning). The weakest has White at +4.21 (a trivial win). I guess none of the engines has the 7-piece tablebase programmed in. Maybe that should not be a surprise as the storage size you need for a tablebase is absolutely huge.

I was going to give a further example, another game against my Austrian opponent Herr Wiesinger. We have reached a rook and pawn ending where the Austrian is totally lost. He knows this and has slowed the game down to a funereal pace. Unfortunately there are still eight pieces on the board so we are just out of tablebase range. I expect one of his pawns to drop off very soon (when he finally gets round to moving) and then it will be Goodnight Vienna.

So much for correspondence chess. What about tablebases in OTB chess? Now here’s a thing. Chess leagues are increasingly using incremental time controls. These carry the risk that a game may still be going on at the end of the playing session. Back in the day, if the two players couldn’t agree on the outcome, these games were resolved either by adjournment to a later date or by adjudication. Neither of these is ideal. Adjournments simply add to fixture congestion. Adjudications take the result of the game out of the hands of the players and transfer it to a third party (in Barking’s eyes this is a detestable practice).

The London League used to rely solely on adjournments, following a playing session of 30 moves in 90 minutes. Those days have gone. At some point in the past decade, the League introduced an incremental time control of G75 + 15. So far, there have been no documented cases of games being unresolved at the end of the playing session. But adjudication is available as a backup if needed.

This season, and as far as I know this is the first time this has happened in English league chess, tablebase decision-making has been introduced. If a game is unfinished, and there are no more than seven pieces on the board, the outcome is determined by reference to a tablebase rather than adjudication. This has the advantage of determining the outcome by objective rather subjective means since you know with certainty what will happen with perfect play on both sides, rather than relying on the human (no doubt computer-assisted) determination of the adjudicator. The relevant London League rule is as follows:

C8. UNFINISHED GAMES
C8.1 If any game is unfinished at the end of the session, both players and both captains shall record the position on the board. If the two sides cannot agree a result, then –

(a)  if no more than 7 pieces remain on the board, the result shall be determined by reference to a tablebase such as the Syzygy endgame tablebase at
https://syzygy-tables.info/?fen=4k3/8/8/8/8/8/8/4K3_w_-_-_0_1;

(b)  In other cases, both sides shall submit an adjudication claim to the league secretary together with a fee of £10. If only one claim is made, it shall be upheld without adjudication. Either side may appeal against an adjudication decision, providing supporting evidence, and on payment of a further fee of £10. The fees shall be returned to the relevant side if no adjudication is needed or if an appeal succeeds.

Whoever thought of this deserves a medal. Who says space can’t be conquered?!

On playing for more than one club

1 Reply

Team sport is an activity contested by different teams with their own sets of players in the course of a season which lasts less than one calendar year. This general definition applies regardless of which particular sport is involved – football, cricket, rugby… and chess.

In what circumstances, if any, can players turn out for different clubs in the same season? Each sport has its own rules. Football is one of the most prominent. There is a defined transfer window at the start of each season and in the middle of the season. During the window, players may transfer from one club to another. So it is possible for a footballer to play for two different clubs in the same season, but not at the same time. That’s how the league works anyway.

In the cup competition the rules are stricter. If you play a cup match for club A, and transfer to club B later in the season, you cannot play for club B in later rounds of the cup competition in the same season. You are said to be “cup-tied.” The point is to prevent eliminated clubs from selling their top players to another club with a view to that club increasing its chances of winning the cup. It’s an old rule and pre-dates the introduction of the transfer window, so maybe it’s anachronistic.

I carried out a survey of chess leagues in south east England (plus the 4NCL) to see whether there were any obvious patterns or predominant practice. There are 17 leagues. All the leagues where information is available restrict players moving around, though the extent of the restriction varies.

In six leagues, a player may not play for more than one club in the same season. These are the 4NCL, the Essex League, the Kent League, the London Public Service League, the Middlesex League, and the Portsmouth & District League. This is stated in the rules of most of these leagues. The London Public Service League is silent on the point, but its administrator has said that players are limited to one club and this may be made clear on the face of the rules from next season.

In another five leagues, a player may represent more than one club in the same season, in certain defined circumstances.

  • The Central London League limits you to one club in divisions 1 and 2, but you may play for a different club in division 3 (presumably this means divisions 3 and 4 combined, now that the league is running a division 4).
  • The London League limits you to one club in the divisional league (divisions 1 to 4), but you may play for one different club in the Major and Minor divisions combined, and yet another club in the Eastman Competition (formerly the Eastman Cup). These are effectively three different competitions in one league with some shared and some distinct rules.
  • The Mid-Sussex League allows you to change clubs once a season, so you can play for a second club, but only if you have played fewer than five games for the first club.
  • The Surrey League allows you to play for different clubs in different competitions, and the different divisions of the league count as separate competitions for this purpose, but only if you register for all the relevant clubs before the season starts. I tried to play for two different clubs in the Surrey League last season but was thwarted by the registration rule.
  • The Thames Valley League allows you to play for two different clubs in one season, but not at the same time, and only in exceptional circumstances, which the rules do not define.

Finally, in a further six leagues, either the rules are silent on the issue or the rules are not available online. These are the Berkshire League, the City League (formerly the Combined London Banks and Insurance League), the Croydon & District League, the Hillingdon & District League, the Southampton League, and the Surrey Border League.

I suppose the point is to reinforce club loyalty – although that’s arguably less important these days, with the advent of clubs in the 4NCL and the London League whose composition reflects friendship or common interest rather than geographical ties. Also maybe the rules are there to prevent player anarchy. Imagine if Erling Haaland played for City v United one week, then decided to play for United v City the following week. This is the stuff of nightmares. You couldn’t run a league on that basis.

Even so, there are circumstances where a player has a genuine reason for turning out for two different clubs in the same season. For example the player may turn out a few times for club A, then resign from the club because of irreconcilable differences, and join club B instead. In effect the player is making a new start in the league. It seems unduly restrictive to prevent the player turning out for the new club until the following season.