In 2016, DrD2k9 submitted a Battle Chess TAS (#5267: DrD2k9's NES Battle Chess "All Battle Animations" in 18:21.19) with the goal of showing all 35 capture animations in Battle Chess. It played multiple games and edited the board repeatedly to make this possible, and was rejected for failing to qualify as a speedrun. In a single standard game of Battle Chess, the maximum number of distinct capture animations is 25. This run aims to show 25 different capture animations as quickly as possible against the Novice AI, which creates an interesting chess puzzle. For instance, the four rooks on the board must be captured by four different pieces, and four pawns must be turned into queens to show off all six queen capture animations. (Battle Chess does not allow the player to promote pawns into anything other than a queen.)
This TAS completes the game in 55 moves, with a total of 27 captures, and a total of 217 squares moved by the pieces, with a time of 25:30. When a piece moves aside to let a knight or a castling rook go by, I counted it as 1 square of movement. Pieces move very slowly in Battlechess, so a run could save time by finding a game with fewer total moves or a shorter total distance. All moves are executed as quickly as possible.
I played out complete games with help from the Stockfish chess engine, looking for games that would lead to the unique captures and multiple pawn promotions needed for this run. I tried several common openings - the King’s Knight, King’s Gambit, Queen’s Gambit, Zukertort Variation, English Game, and the Halosar Trap - before I found that the Ware opening led to the unusual captures and rapid pawn development that I wanted. I play as White, the default, and win the game by checkmate.
Piece movement in Battle Chess is very slow. I created an encode with the movement sped up 4x, while capture animations play at normal speed. Or the game can be watched on a standard online chessboard using this link: https://lichess.org/AsxJLXZx
An analysis of some key moves:
5. Nxe5 is a knight-pawn capture, which leads to a bishop capturing the knight, and sets up a fork of two of black’s minor pieces.
12. Qc3 threatens to put Black’s king in check, but the AI does not notice.
19. f4 leads to a devastating blunder, where Black doesn’t notice that a knight is about to fork his king and queen, allowing the first of six queen captures.
25. Qb4 leaves the Black knight with nowhere to retreat, and the AI moves a different piece instead of trying to save it.
28. b4 dxc2 29. b5 cxd1=Q+ In this sequence, I allow Black’s pawn to capture my rook and become a queen, because I need my bishop to capture a queen before the game is over.
33. Bb3 Rc5 leaves Black on the verge of checkmate, so the AI starts trying to distract me from finishing the game immediately. At this point, Stockfish is screaming at me to checkmate the king already, but I have other plans. Over the coming sequence of moves, Black gives up both of its rooks for a large loss.
44. Qbf7+ puts the Black king in a pinned position – he is unable to move, so one of the black pawns must advance instead. This is good, because my rook needs to capture a queen.
50. Rh5+ forces the Black queen to capture my rook, and put herself in a position where she can be captured by a pawn.
51. Qg7+ sacrifices a queen because I need the queen-captured-by-king animation, and it allows my pawn to capture a black queen.
54. Qg3 is the penultimate move. I looked for a route that doesn’t involve the queen slowly moving across five squares, but couldn’t find one that checkmates as quickly.
55. Qg6# is checkmate. Input ends as soon as the move is entered, so the distance moved by the queen doesn't matter to the final time.
slamo: Claiming for judging.
slamo: So I can confirm that there are indeed 25 unique battle animations executed in this run. Whether that's the actual maximum or not is incredibly difficult to prove due to the amount of possible chess games, so we'll take this as the best possible for now. Otherwise, this is actually a valid and interesting optimization challenge. I think it's cool that you used a chess engine to set everything up, and your general strategy and capture sequences make sense. The extra 2 captures look like they were necessary to move the pawns forward far enough to get promoted.
As far as entertainment goes...eh. The temp encode does not quite do justice to how glacially slow the movement is. If people didn't receive the temp encode very well, then they will definitely have a hard time watching an encode at regular speed. It's just not entertaining enough for Moons.
Since it's not an eligible Standard category either, it's not going be published for now. We can still put it in Playground though, since it's a valid run and there's probably still an audience for it. Moving to Playground.