Perpetual Check

[CRoberson]

Has anybody implemented this? The USCF dropped the perpetual check rule due to it eventually going into 3-fold occurance or the 50 move rule. Don't know about FIDE. I am interested in implementing it, because Ares just lost a game on CCRL that was a draw. It's pv claimed 15
consecutive checks. But not a 3-fold occurance or 50 move rule hit (which are in Ares). Ares is down material but the king can't escape the checks. Stockfish, Dragon and Ares given time think it is a draw with scores of 0.00.

The question is how many consecutive checks to base peretual on. When it hits n then report 0.00, but what is n? It might be safest to not do this in the search, but report 0.00 in the final score if the last n moves by the attacker are checks.

[Joost Buijs]

I've never heard of a perpetual-check rule. It seems to me you can only lose a game on time when you have perpetual-check.
Nightmare only knows about 3-fold repetition and the 50 move rule.

[chrisw]

I've never heard of a perpetual-check rule. It seems to me you can only lose a game on time when you have perpetual-check.
Nightmare only knows about 3-fold repetition and the 50 move rule.

Me neither

[Uri Blass]

Has anybody implemented this? The USCF dropped the perpetual check rule due to it eventually going into 3-fold occurance or the 50 move rule. Don't know about FIDE. I am interested in implementing it, because Ares just lost a game on CCRL that was a draw. It's pv claimed 15
consecutive checks. But not a 3-fold occurance or 50 move rule hit (which are in Ares). Ares is down material but the king can't escape the checks. Stockfish, Dragon and Ares given time think it is a draw with scores of 0.00.

The question is how many consecutive checks to base peretual on. When it hits n then report 0.00, but what is n? It might be safest to not do this in the search, but report 0.00 in the final score if the last n moves by the attacker are checks.

It may be interesting question what is the position with the maximal consectutive number of checks that the losing side can force.

My guess is more than 15.

[ovenel]

According to Wikipedia, the rules for drawing by perpetual check had been dropped by 1967. So it's not surprising that people are unaware of this rule.

In regards to engine development, it seems like it would be difficult to say whether a perpetual check is a draw because it's sometimes part of a king walk to safety. In particular, in endgames there are oftentimes lines where a checkmate is delayed by the losing player continuously checking the opponent's king until they eventually run out of checks, at which point the other player can start the checkmating sequence. So flatly saying that a perpetual check is a draw would be difficult as the king may eventually make it to a safe square. But maybe it would be beneficial to evaluate a draw after what appears to be a perpetual check in quiescent search?

[Ajedrecista]

Hello:

According to Wikipedia, the rules for drawing by perpetual check had been dropped by 1967. So it's not surprising that people are unaware of this rule.

[...]

Just to complete Nelson's find, I found the following:

a) I found an incomplete transcription of FIDE Laws of Chess 1977/1978 (https://www.janko.at/Retros/Glossary/FideLaws1977.htm):

Part I. General Laws

[...]

ARTICLE 12 • THE DRAWN GAME

The game is drawn –

[...]

4. when a player having the move demonstrates that at least fifty consecutive moves have been made by each side without me capture of any piece or the movement of any pawn.

This number of fifty moves can be increased for certain positions, provided that this increase in number and these positions nave been clearly established before the commencement of the game.

FIDE Interpretation Art. 12.4 (1958A)

Question: Can a player lose the game by exceeding the time-limit when the position is such that no mate is possible, whatever continuation the players may employ (this concerns Part II of the Laws)?

Answer: The Commission declares that the Laws must be interpreted in such a way that in this case, as in the case of perpetual check, a draw cannot be decreed against the will of one of the players before the situation foreseen in Article 12.4 is attained.

[...]

There was not such a 'draw by perpetual check' rule then, this is, there were other ways to claim a draw, as already stated before in this thread.

As 1958 is cited above (1958A), I also found the preface of 1958 Rules of Chess at FIDE Arbiters' Commission Arbiters' Manual (September 2024):

1958

"GENERAL OBSERVATIONS. The Laws of Chess cannot, and should not, regulate all possible situations that may arise during a game, nor they can regulate all questions of organization. In most cases not precisely regulate by an Article of the Laws, one should be able to reach a correct judgment by applying analogous stipulations for situations of a similar character. As to the arbiters' tasks, in most cases one must presuppose that arbiters have the competence, sound of judgment, and absolute objectivity necessary. A regulation too detailed would deprive the arbiter of his/her freedom of judgment and might prevent him from finding the solution dictated by fairness and compatible with the circumstances of a particular case, since one cannot foresee every possibility."

------------------------

b) I found a mention to and old rule (sic) covering a draw by perpetual check at page 51 of The Official Blue Book and Encyclopedia of Chess (1956) by Kenneth Harkness, published for the United States Chess Federation:



The excerpt says that this old rule had been abandoned. Again, a threefold repetition is the way to claim the draw:

https://books.google.es/books?redir_esc ... =perpetual

It will be noted that the old rule covering a draw by perpetual check has been abandoned. A player who can subject his opponent's King to an endless series of checks can force a third repetition of the position and claim the draw under Article 12, No. 3.

Regards from Spain.

Ajedrecista.

[Jouni]

Fide 2023 chess laws have 75 moves draw rule:

9.6 If one or both of the following occur(s) then the game is drawn:

9.6.1 the same position has appeared, as in 9.2.2 at least five times.

9.6.2 any series of at least 75 moves have been made by each player without the movement of any pawn and without any capture. If the last move resulted in checkmate, that shall take precedence.

[CRoberson]

Checked some old USCF rule books: it is in the 5th edition (2003) but not the 6th edition (2014). So, it looks like the USCF dropped it in 2014. Statement for history sake only.

My point is that after the PV has n consecutive checks, why not cut the score in half? Don't set it to zero as that could miss 3-fold rep.

Another interesting USCF rule is the insufficient moves rule (14d4): This is one exception to loss on time (just like insufficient material).
14D4. No legal moves leading to checkmate by opponent.
There are no legal moves that could lead to the player being checkmated by the opponent.

So, if player A runs out of time and player B calls it, normally that would be a win for B. But there are exceptions to that such as the well know insufficient material rule. 14D4 allows player A to claim he has mate in N, assuming that can be proven, then the game is a draw.
Don't know if FIDE has such a rule.

[ovenel]

The Chess Programming Wiki page on Check lists a few different threads discussing perpetual check. I'm not entirely clear about the issue that you had where you lost a game that was drawn, but I think that it would be related to the Spite Checks that hgm had posted about back in 2014. It sounds like the resolution in that thread was in regards to being selective about which checks to extend in the search and relying on a deeper search to sort out the issue.

It looks like the same thread was referenced here in 2015 with another explanation of the idea that hgm had for resolving spite checks. In essence, he proposed classifying checks as potentially pointless if every other possible move is bad.

However, if I'm understanding your question correctly, it may be the case that these "spite checks" are the only way to survive. That is, if a check is not played, then the opponent can proceed with a forced checkmate, but you are able to push that checkmate out indefinitely with a series of checks (possibly losing material in the meantime), thus turning a loss into a draw. So in that case, we'd want to evaluate the position as a draw (or a potential draw) before we continue down the full line that leads to a draw by repetition. In that case, maybe you can implement something akin to what hgm was suggesting in these threads to classify checks as suspect and then adjust the score as you were suggesting if the best move is a suspect check.

I could also be misunderstanding these other threads, and I know that hgm is still a very active member here (and certainly much more knowledgeable about the theory behind chess engines than I), so he can probably give a better interpretation of his thoughts on this matter.

[CRoberson]

Nelson,
Yes, it is common even for humans to use a series of checks to draw especially against a king in the open field.
Here is the game. You can see the final position: the white king is in the open and enough black pieces are there to produce a long series of checks. Thus the no progress rule could be invoked in USCF events, if the TC didn't have an increment. But no USCF or FIDE TD/Arbiter would call this game a win for white.
Stockfish, Dragon, Ares and others (John Merilino said CM9 sees the draw) all see the draw in 20 ply or less.
Interestingly enough, a superficial position inspection looks like white has a mate in 1 in the waiting, but will not get to pull it off.

[pgn]
[Event "?"]
[Site "Swiss 2025-IV"]
[Date "2025.04.08"]
[Round "?"]
[White "Francessca MAD"]
[Black "Ares5.45avx2q"]
[Result "1-0"]
[PlyCount "75"]

1. d4 {(Book)} d5 {(Book)} 2. c4 {(Book)} c6 {(Book)} 3. Nf3 {(Book)} Nf6
{(Book)} 4. Nc3 {(Book)} a6 {(Book)} 5. g3 {(Book)} dxc4 {(Book)} 6. e4
{(Book)} b5 {(Book)} 7. e5 {(Book)} Nd5 {(Book)} 8. Bg2 {(Book)} e6
{(Book)} 9. O-O {-0.68/28 1744131362 (O-O)} Be7 {(Be7 h4 Ra7 a4 O-O Ng5 h6
Nh3 Rd7 Nxd5 exd5 axb5 axb5 Nf4 Na6 b3 Nc7) +0.15/19 43} 10. Ne4 {(Ne4)
-0.35/30 36} Ra7 {+0.15/21 35 Ra7 h4 a5 Bg5 O-O Bxe7 Qxe7 Nd6 Nd7 a4 Ba6
axb5 Bxb5 Re1 N7b6 Qd2 Nc8 Nxc8 Rxc8)} 11. Bg5 {+0.45/34 19 Bg5 f6 exf6
gxf6 Ne5 fxe5 Qh5+ Kd7 Bxe7 Kxe7 Qxe5 Rf8 Qxb8 Qc7 Qxc7+ Rxc7 Nc5 Rd8 Rfd1
Kf6 Rd2 e5 dxe5+ Kxe5 f4+ Kd6 Ne4+ Ke6 Bh3+ Kf7 Bxc8 Rcxc8 a3 a5 Re1)} O-O
{+0.62/21 44 (O-O)} 12. Bxe7 {-0.35/25 18 Bxe7 Rxe7 a4 f6 Qd2 Nd7 Rfe1 Qc7
exf6 gxf6 Nh4 Rg7 Nc3 Nxc3 bxc3 Nb6 Bh3 f5 axb5 axb5 Re5 Nd5 Bg2 Kh8 Bxd5
cxd5)} Qxe7 {(Qxe7) +0.54/21 63} 13. Nd6 {(Nd6) -0.07/27 27} Rc7 {+0.85/19
47 Rc7 Re1 Nd7 h4 h6 b3 c3 a3 c5 Rc1 b4 dxc5 bxa3 Nd4 Nxc5 Bxd5 exd5 Rxc3
Ne4)} 14. a4 {-0.28/26 15 a4 Nd7 Qe2 f6 axb5 cxb5 Nxc8 Rcxc8 exf6 Rxf6 Rxa6
Nb8 Ra8 Nb6 Raa1 Nc6 Qe4 Rf5 Ra6 Nd5 Ne5 Nxe5 dxe5 c3 Rc1 b4 bxc3 bxc3)}
Nd7 {(Nd7) +0.45/19 73} 15. Qe2 {(Qe2) -0.22/27 16} N7b6 {(N7b6) +0.85/19
26} 16. Nd2 {(Nd2) -0.39/27 16} c3 {(c3) +1.95/19 33} 17. bxc3 {-0.57/29 19
bxc3 Nxc3 Qd3 Nbxa4 Nb3 f6 f4 Nd5 Rac1 Bd7 Be4 g6 Rf2 Rb8 h3 Kh8 Kh2 Qf8
Bf3 Ra8 Qd2 Raa7 Qa5 Ne3 Re1 Nf5 Nxf5 exf5 exf6 Qxf6)} Nxc3 {(Nxc3)
+1.63/20 25} 18. Qd3 {-0.52/28 15 Qd3 Nbxa4 Nb3 f6 f4 Nd5 Rac1 Bd7 Qd2 Rb8
Rf2 g6 Bf3 Qf8 h3 Ra8 Qa5 Raa7 Be4 Kh8 Ra2 fxe5 dxe5 c5 Kh2 c4 Nd4)} Ncxa4
{+1.09/18 41 Ncxa4 Rfc1 Nd5 N2e4 Rd8 Qe2 f6 Nc5 Nxc5 dxc5 fxe5 Qxe5 Qf6 Qe2
h6 Re1)} 19. Rfc1 {-0.48/25 20 Rfc1 Bd7 Nb3 Nd5 Be4 Kh8 Qd2 Ra8 Ra2 g6 f4
f6 Qa5 Raa7 Nc5 Nxc5 Rxc5 Qf8 Re2 f5 Bg2 Qb8 Nf7+ Kg8 Nd6)} c5 {+1.87/17 28
c5 dxc5 Nxc5 Qe3 Nca4 N2e4 Rxc1+ Rxc1 Bd7 h4 Nd5 Qd4 h6 Re1 Rd8 Rc1 Be8)}
20. dxc5 {-0.72/29 19 dxc5 Rxc5 Rxc5 Nxc5 Qd4 Nca4 Rc1 f6 f4 fxe5 fxe5 Qg5
Rc7 Bd7 Bf3 Qd8 Ra7 a5 Rxa5 Nc8 Ra8 Qb6 Qxb6 Naxb6 Rb8 Nxd6 Rxf8+ Kxf8 exd6
Nc4)} Rxc5 {(Rxc5 Qd4 Qc7 Nb3 Rc3 h3 h6 Kh2 Nd5 h4 Rxc1 Rxc1 Ndc3 Qf4 Bd7)
+1.40/17 21} 21. Rxc5 {(Rxc5 Nxc5 Qd4 Nca4 Rc1 f6 f4 fxe5 fxe5 Qg5 Rc7 Bd7
N2e4 Qd8 Ra7 Nd5 Rxa6 Qc7 h3 b4 Ra7 Qc6 Kh2 Qb6 Qxb6 Ndxb6 Rc7 b3 Nd2 b2
Be4) -0.48/32 14 14} Nxc5 {(Nxc5) +0.70/20 39} 22. Qd4 {(Qd4 Nca4 Rc1 f6 f4
fxe5 fxe5 Bd7 Rc7 Qd8 Ra7 a5 Nb3 Qg5 Ne4 Qh6 h4 Bc6 Ng5 Bxg2 Kxg2 Qg6 Nxa5
Nd5 Nb3 Qb1 Nxe6 Qxb3 Nxf8 Kxf8) -0.63/31 14 14} Nca4 {+0.07/18 24 Nca4 Rc1
Rd8 Rc6 Bb7 Rxb6 Bxg2 Rxa6 Ra8 Rxa8+ Bxa8 Nxb5 Nc5 Qd6 Qxd6 Nxd6 f6)} 23.
Rc1 {(Rc1) -0.43/32 14} Rd8 {+0.23/24 34 Rd8 Rc6 Bb7 Rxb6 Bxg2 Rxa6 Bd5
Nxb5 Nc5 Ra7 Nd7 h4 h6 Nd6 Ra8 Rxa8+ Bxa8 Qa7 Qd8 f4 Bc6 N2e4 Bd5)} 24. Rc6
{-0.26/32 14 Rc6 Bb7 Rxb6 Bxg2 Rxa6 Bd5 Nxb5 Qc5 Qxc5 Nxc5 Rd6 Rd7 Nc3 Kf8
Rxd7 Nxd7 f4 Bb7 Kf2 f6 exf6 Nxf6 h3 Ke7 Nc4 Nd5 Ne4 Nb6 Nxb6 Bxe4 Ke3 Bg2
h4 Bd5)} Bb7 {+0.23/23 34 Bb7 Rxb6 Bxg2 Rxa6 Bd5 Ra7 Rd7 Nxb5 Rxa7 Qxa7
Qxa7 Nxa7 Kf8 f4 f6 exf6 gxf6 Nb5 Nb2 Kf2 Ke7)} 25. Rxb6 {-0.46/32 12 Rxb6
Bxg2 Rxa6 Bd5 Nxb5 Qc5 Qxa4 Qc1+ Nf1 Bc4 Qa1 Qxa1 Rxa1 Bxb5 Ra5 Be2 Kg2 h6
Ra7 g5 Ne3 Rd2 g4 Kg7 Kg3 Rd4 f3 Rd2 Rb7 Kg6 Rc7 Bd3)} Bxg2 {+0.15/23 20
Bxg2 Rxa6 Bd5 Ra7 Rd7 Nxb5 Rxa7 Qxa7 Qxa7 Nxa7 g5 Nb5 Kg7 Nd6 Nc3 f3 f5 Kf2
Kg6 Nf1 f4)} 26. Rxa6 {-0.41/34 20 Rxa6 Bd5 Ra7 Rd7 Ra5 f6 N2e4 fxe5 Qxe5
Nc5 Nxc5 Rxd6 Rxb5 Rd8 Rb8 Rxb8 Qxb8+ Qf8 Qxf8+ Kxf8 f4 h6 Kf2 Ke7 Ke3 Kf6
Ne4+ Kf5 Nc3 Bb7 Nb5 Bd5 Kd4 g5 Nc3 gxf4)} Bd5 {+0.15/23 17 Bd5 Ra7 Rd7
Nxb5 Rxa7 Qxa7 Qxa7 Nxa7 g5 Nc8 Kg7 Nd6 Kg6 N2e4 Nb2 Nc5 h6 f4 gxf4 gxf4
Nd1)} 27. f4 {(Ra7) -0.32/34 20} Qd7 {+1.79/19 27 Qd7 Kf2 h6 Ke3 Ra8 Rxa8+
Bxa8 Nxb5 Qxb5 Qd8+ Kh7 Qxa8 Qb6+ Kd3 Qb5+ Ke3 Nc3 Qf3 Qb2 Kd3 g6)} 28. f5
{(f5) -0.83/30 16} h5 {+1.95/16 29 h5 Ra7 Qc6 f6 Qc1+ Nf1 Nc3 Rxf7 gxf6 Qf2
Qd1 Qa7 Qc2 Rg7+ Kh8 Ne3 Qb1+)} 29. f6 {(Kf2) -0.70/33 24} Qc7 {(Qc7 Qf4
gxf6 exf6 Qc5+ Kf1 Qc1+ Kf2 Qc5+ Ke1 Qc1+) 0.00/15 36} 30. Qf4 {(Qf4 Qc1+
Kf2 Qh1 fxg7 Qxh2+ Ke1 f5 Qg5 Qh1+ Kf2 Qh2+ Ke1) -0.01/33 14 14} Qc1+
{(Qc1+ Kf2 gxf6 exf6 Qc5+ Ke1 Qc1+ Kf2) 0.00/16 8} 31. Kf2 {(Kf2 Nc3 fxg7
Nd1+ Ke2 Nc3+ Kf2) -0.01/34 13} Nb2 {(Nb2 Qg5 g6 Qh6 Nd1+ Ke2 Nc3+ Kf2
Nd1+) 0.00/17 14} 32. Qg5 {(Qg5 Nd1+ Ke2 Nc3+ Ke3 Nd1+ Ke2) +0.01/37 13} g6
{(g6 Qh6 Nd1+ Ke2 Nc3+ Kf2 Nd1+) 0.00/21 21} 33. Qh6 {+3.73/35 12 Qh6 Nd1+
Ke2 Bc4+ N6xc4 Nc3+ Kf2 Ne4+ Kg2 Rxd2+ Nxd2 Qxd2+ Qxd2 Nxd2 Ra8+ Kh7 Rf8
Nc4 Rxf7+ Kg8 Rg7+ Kf8 Rxg6 Nxe5 Rg5 Ng4 Rxh5 b4 Rh4 Nxf6 Rxb4 e5 Kf3 Ke7
Rb5 Ke6 Rb7)} Nd1+ {(Nd1+ Ke2 Nc3+ Kf2 Nd1+) 0.00/20 7} 34. Ke2 {(Ke2)
+4.15/37 10} Nc3+ {(Nd1+ Ke2 Nc3+ Kf2 Nd1+) 0.00/20 18} 35. Kd3 {+6.58/33
13 Kd3 Be4+ Kd4 Rxd6+ Rxd6 Qg1+ Kxc3 Qc5+ Kb3 Bd5+ Kb2 Qd4+ Kc2 Be4+ Nxe4
Qxe4+ Rd3 Qe2+ Qd2 Qe4 Qf4 Qxf4 gxf4 Kh7 Kb3 Kh6 Rd7 Kh7 Rxf7+ Kh6 Rc7)}
Be4+ {-5.54/16 21 Be4+ Kd4 Rxd6+ Rxd6 Qg1+ Qe3 Qxe3+ Kxe3 Bd5 Rd8+ Kh7 Kd4
Ne2+ Kc5 b4 Rd7 Kg8 Kxb4 Nd4 Rd8+ Kh7 Kc5 Nf5 Rd7 Kg8)} 36. Kd4 {(Kd4)
+6.58/38 11} Qg1+ {-5.23/15 19 Qg1+ Kxc3 Rc8+ Nxc8 Qc5+ Kb3 Bd5+ Kb2 Qd4+
Kc2 Be4+ Nxe4 Qxe4+ Kb2 Qb4+ Kc2 Qc4+ Kd1 Qf1+ Kd2 Qf2+ Kd3 Qf1+ Kc3 Qc4+)}
37. Kxc3 {+6.42/39 11 Kxc3 Rc8+ Nxc8 Qc5+ Kb2 Qd4+ Ka3 Qc5+ Kb3 Bd5+ Kb2
Qd4+ Kc2 Be4+ Kd1 Bf3+ Kc1 Qc3+ Kb1 Qd3+ Kb2 Qd4+ Ka2 Bd5+ Nb3 Bxb3+ Kxb3
Qc4+ Kb2 Qb4+ Kc2 Qc5+ Kb1 Qb4+ Ka2 Qc4+ Ka1 Qc3+ Kb1 Qxc8)} Rc8+ {(Rc8+)
-4.91/20 15} 38. Nxc8 {+6.54/43 9 Nxc8 Qc5+ Kb3 Bd5+ Kb2 Qd4+ Kc2 Be4+ Kd1
Bf3+ Kc1 Qc3+ Kb1 Qd3+ Kb2 Qd4+ Ka2 Bd5+ Nb3 Bxb3+ Kxb3 Qc4+ Kb2 Qb4+ Kc2
Qc4+ Kd2 Qb4+ Kd1 Qb3+ Ke1 Qb1+ Ke2 Qc2+ Qd2 Qxc8 Rb6 Qc5 Rb8+ Kh7 Qe3 Qc2+
Ke1 Qxh2 Rxb5) (Nxc8 Qc5+ Kb3 Bd5+ Kb2 Qd4+ Kc2 Be4+ Kd1 Bf3+ Kc1 Qc3+ Kb1
Qd3+ Kb2 Qd4+ Ka2 Bd5+ Nb3 Bxb3+ Kxb3 Qc4+ Kb2 Qb4+ Kc2 Qc4+ Kd2 Qb4+ Kd1
Qb3+ Ke1 Qb1+ Ke2 Qc2+ Qd2 Qxc8 Rb6 Qc5 Rb8+ Kh7 Qe3 Qc2+ Ke1 Qxh2 Rxb5)
(Nxc8 Qc5+ Kb3 Bd5+ Kb2 Qd4+ Kc2 Be4+ Kd1 Bf3+ Kc1 Qc3+ Kb1 Qd3+ Kb2 Qd4+
Ka2 Bd5+ Nb3 Bxb3+ Kxb3 Qc4+ Kb2 Qb4+ Kc2 Qc4+ Kd2 Qb4+ Kd1 Qb3+ Ke1 Qb1+
Ke2 Qc2+ Qd2 Qxc8 Rb6 Qc5 Rb8+ Kh7 Qe3 Qc2+ Ke1 Qxh2 Rxb5)} 1-0
[/pgn]