STRATIC
Home • Created 2020/4/20 • Updated 2023/2/12 • v0.7 • Read time 9min • Discord
______________________________________________________________________________
About
Stratic is a strategy board game designed to stop AI from outperforming humans. I made
Stratic so we could test AI edge cases as well as human cognitive limits. I have also
included a very simple resource management function so Stratic technically qualifies
as a tactics-level wargame and can be used to teach basic war theory to people.
To clarify how this combats traditional AI engines: the board continually grows in size, the number of pieces on the board generally increases over time, and the game is designed to continue potentially indefinitely, with no asymptotic difficulty limit. This means the state-space complexity, game tree size, and decision tree size are all infinite as well as the computational complexity since you can't calculate end-states in finite time. Stratic is not computationally trivial; players have imperfect information as there are no preset starting formations and new pieces can be placed randomly on the board. Genetic algos (demonstrated by Hexapawn +) are thwarted by Stratic since there aren't a finite or closed number of endstates to the games.
To clarify how this combats modern AI engines: we are working on implimenting mechanics to stop DNNs from being able to use scaling effectively by requiring win conditions for the game to be dynamically determined using things like a score or point assessment every round and making the algorithm for determining the score/points cryptographically non-trivial. Something like VDFs can force realtime delays and significantly hinder engines from training at accelerated speeds. If that's not enough, I have also considered making the score calc be dramatically influenced by the amount of time you take to make a move, where the longer you take to make a move increases the move's scoring, incentivizing an AI engine to waste as much time as possible each move. This forces engines to train many orders of magnitude slower than they would normally, on timescales that make it easy for humans to outpace them.
We need help testing Stratic, so if anyone wants to help try it out, @ me. We have a bot for it on our Discord and soon we should have a web-end as well (this page you are on now will become a portal for playing the game) so you will be able to play it online like most other strategy board games. Here is a link to a gdoc showing a recent game.
To clarify how this combats traditional AI engines: the board continually grows in size, the number of pieces on the board generally increases over time, and the game is designed to continue potentially indefinitely, with no asymptotic difficulty limit. This means the state-space complexity, game tree size, and decision tree size are all infinite as well as the computational complexity since you can't calculate end-states in finite time. Stratic is not computationally trivial; players have imperfect information as there are no preset starting formations and new pieces can be placed randomly on the board. Genetic algos (demonstrated by Hexapawn +) are thwarted by Stratic since there aren't a finite or closed number of endstates to the games.
To clarify how this combats modern AI engines: we are working on implimenting mechanics to stop DNNs from being able to use scaling effectively by requiring win conditions for the game to be dynamically determined using things like a score or point assessment every round and making the algorithm for determining the score/points cryptographically non-trivial. Something like VDFs can force realtime delays and significantly hinder engines from training at accelerated speeds. If that's not enough, I have also considered making the score calc be dramatically influenced by the amount of time you take to make a move, where the longer you take to make a move increases the move's scoring, incentivizing an AI engine to waste as much time as possible each move. This forces engines to train many orders of magnitude slower than they would normally, on timescales that make it easy for humans to outpace them.
We need help testing Stratic, so if anyone wants to help try it out, @ me. We have a bot for it on our Discord and soon we should have a web-end as well (this page you are on now will become a portal for playing the game) so you will be able to play it online like most other strategy board games. Here is a link to a gdoc showing a recent game.
Pieces, Movement, & Rules
There are two ways to win in Stratic, either capture the commander (equivalent to
capturing the king in chess) or someone surrenders; attrition is a real strategy we
allowed so the games could continue forever and get exponentially more complicated
over time — it is up to the player to win outright or test their will by simply never
quiting.
There are four kinds of pieces: Fortifiers, Aggressors, Interdictors, and Commanders. There is only one Commander on the board per player and the game is over when it is captured. You must place at least one of each piece and have a total of 19 pieces placed before starting. 19 was chosen for the starting formations since you can't score 19 in cribbage and an odd number of pieces on an even board works out well in the meta. Players do not know each other's placements until starting the game. You can only move onto a space if it is empty, you are capturing a piece that is currently positioned there, or you are occupying a Fortifier (explained below).
FORTIFIER: can move up to 2 spaces in a straight line; can only capture other fortifiers; cannot occupy any other piece's space; can be promoted to an Aggressor or Interdictor if behind the enemy starting line; when occupied, can only move 1 space at a time. Fortifiers are worth 1 point and promotions also gain you a point.
AGGRESSOR: can move up to 3 spaces straight or 2 spaces in any direction; can capture every piece except Fortifiers; can chain one additional capture if the first capture is within 2 spaces of the originating position; can occupy Fortifiers — doing so allows unlimited unidirectional movement but you cannot move through other pieces. Aggressors are worth 2 points.
INTERDICTOR: unlimited unidirectional movement; can capture every piece; can occupy Fortifiers — allows for unlimited unidirectional movement in which you can jump over as many other pieces as you want. You can only have one Interdictor on the board at a time unless you promote a Fortifier. Interdictors are worth 4 points.
COMMANDER: cannot move on its own unless in check or its position is swapped with another piece that had the Commander in its range; can capture any piece within 1 space of it even if not in check; can occupy Fortifiers. When a Commander occupies a Fortifier, attempting to capture that Commander instead results in the occupied Fortifier & the attacker both getting removed from play. The Commander cannot place itself in check, the same as Chess, however unlike Chess the Commander does not have to move when in check. This is due to the way draws occur in Stratic (see rule on draws). If a draw is not possible, then the Commander does have to move out of check. Commanders are worth 8 points.
Occupation means two pieces occupying the same space, stacked in 3D (like Gungi). Any Fortifier can be occupied by any other piece except other Fortifiers. To remove an occupier you must capture the occupying piece, which means you must be capable of occupying that same space after as well. An occupied Fortifier cannot be targeted for capture by any piece except its occupier. An occupied Fortifier can move even if its occupier is an opponent's piece, but it takes the occupying piece with it. The benefits of occupation do not apply until the next movement after you have occupied a Fortifier. The benefits only last for the first movement of a piece off its Fortifier. You cannot start a game with occupied positions.
Turns in Stratic allow the movement of only one piece at a time, with turns alternating between players. There is no zugzwang, meaning you do not have to move if you don't want to, including during your first turn. You can indefinitely pass your turn if you desire. This is to allow the realism of battles of attrition and also means you can wait to see how your enemy develops their pieces, breaking the strictness of strategy around being first to move. Special conditions allow doubled movement — these are called operations. If you pass on two or more turns in a row, you get one extra movement on the next turn you take to move. This does not stack, you only get one extra movement no matter how many turns you passed on.
The board refers to the 10x10 grid the pieces start on. Starting formations can only be placed within a 2-row space on one of the opposing ends of the board and the space this occurs in is demarcated by the colored 'starting lines'. The 'midboard' is the rectangular space between the two starting lines; it is static and does not grow with board expansion. The 'trough' or 'trench' is the horizontal space between the two starting lines which does grow with board expansion.
Expansion refers to the growth of the board in every direction by one unit (inspired by Prosfair from BBB). Expansion starts occurring after the first two rounds pass (four turns total), then again after three more rounds have passed (six turns), then again after four more rounds have passed (eight turns), then five more rounds (ten turns), and so on, growing by 1 each cycle. The reason for this non-constant expansion is to quickly expand the board in the beginning of the game and then slow expansion for the mid-to-late game. Infinite expansion occurs in the official game, but this is obviously not possible on physical boards so we limit physical boards to an expansion range of eight. Physical boards also don't require infinite rulers to identify their grid spaces with, and so using A-Z and 1-26 instead of negative letters and numbers may be preferable.
Reinforcements are new pieces placed after the game has started (like Shogi). There are two ways to gain reinforcement pieces. The first is by having control of the midboard during expansion (you must have more pieces in the midboard than your opponent) and the second is from your opponent taking more than a set time to make a move. The first method gains you a number of reinforcement pieces equal to how many more pieces you had in the midboard (up to 3). The second method is only applicable in games with time controls and can help games end faster; this mechanic may not be applicable to games against an AI engine (we aren't sure yet). If a player takes longer than a minute to move, it skips to the other player's turn and awards that other player a single reinforcement piece. Both reinforcement methods can be stacked for a given turn.
When you gain a reinforcement, it can only be a fortifier or aggressor, but it can be placed anywhere on the board as long as it is not past the opponent's starting line. An exception to this is if your Commander is past the opponent's starting line, as you can also place a reinforcement within one unoccupied square of your Commander no matter the position of your Commander.
You do not get additional time during your turn to place reinforcements, they must be placed before any other movements are made that turn, and you must place allotted reinforcements. If you place a reinforcement Fortifier in a position that would also allow for promotion, you may also promote that piece in the same turn before moving.
There are four kinds of pieces: Fortifiers, Aggressors, Interdictors, and Commanders. There is only one Commander on the board per player and the game is over when it is captured. You must place at least one of each piece and have a total of 19 pieces placed before starting. 19 was chosen for the starting formations since you can't score 19 in cribbage and an odd number of pieces on an even board works out well in the meta. Players do not know each other's placements until starting the game. You can only move onto a space if it is empty, you are capturing a piece that is currently positioned there, or you are occupying a Fortifier (explained below).
FORTIFIER: can move up to 2 spaces in a straight line; can only capture other fortifiers; cannot occupy any other piece's space; can be promoted to an Aggressor or Interdictor if behind the enemy starting line; when occupied, can only move 1 space at a time. Fortifiers are worth 1 point and promotions also gain you a point.
AGGRESSOR: can move up to 3 spaces straight or 2 spaces in any direction; can capture every piece except Fortifiers; can chain one additional capture if the first capture is within 2 spaces of the originating position; can occupy Fortifiers — doing so allows unlimited unidirectional movement but you cannot move through other pieces. Aggressors are worth 2 points.
INTERDICTOR: unlimited unidirectional movement; can capture every piece; can occupy Fortifiers — allows for unlimited unidirectional movement in which you can jump over as many other pieces as you want. You can only have one Interdictor on the board at a time unless you promote a Fortifier. Interdictors are worth 4 points.
COMMANDER: cannot move on its own unless in check or its position is swapped with another piece that had the Commander in its range; can capture any piece within 1 space of it even if not in check; can occupy Fortifiers. When a Commander occupies a Fortifier, attempting to capture that Commander instead results in the occupied Fortifier & the attacker both getting removed from play. The Commander cannot place itself in check, the same as Chess, however unlike Chess the Commander does not have to move when in check. This is due to the way draws occur in Stratic (see rule on draws). If a draw is not possible, then the Commander does have to move out of check. Commanders are worth 8 points.
Occupation means two pieces occupying the same space, stacked in 3D (like Gungi). Any Fortifier can be occupied by any other piece except other Fortifiers. To remove an occupier you must capture the occupying piece, which means you must be capable of occupying that same space after as well. An occupied Fortifier cannot be targeted for capture by any piece except its occupier. An occupied Fortifier can move even if its occupier is an opponent's piece, but it takes the occupying piece with it. The benefits of occupation do not apply until the next movement after you have occupied a Fortifier. The benefits only last for the first movement of a piece off its Fortifier. You cannot start a game with occupied positions.
Turns in Stratic allow the movement of only one piece at a time, with turns alternating between players. There is no zugzwang, meaning you do not have to move if you don't want to, including during your first turn. You can indefinitely pass your turn if you desire. This is to allow the realism of battles of attrition and also means you can wait to see how your enemy develops their pieces, breaking the strictness of strategy around being first to move. Special conditions allow doubled movement — these are called operations. If you pass on two or more turns in a row, you get one extra movement on the next turn you take to move. This does not stack, you only get one extra movement no matter how many turns you passed on.
The board refers to the 10x10 grid the pieces start on. Starting formations can only be placed within a 2-row space on one of the opposing ends of the board and the space this occurs in is demarcated by the colored 'starting lines'. The 'midboard' is the rectangular space between the two starting lines; it is static and does not grow with board expansion. The 'trough' or 'trench' is the horizontal space between the two starting lines which does grow with board expansion.
Expansion refers to the growth of the board in every direction by one unit (inspired by Prosfair from BBB). Expansion starts occurring after the first two rounds pass (four turns total), then again after three more rounds have passed (six turns), then again after four more rounds have passed (eight turns), then five more rounds (ten turns), and so on, growing by 1 each cycle. The reason for this non-constant expansion is to quickly expand the board in the beginning of the game and then slow expansion for the mid-to-late game. Infinite expansion occurs in the official game, but this is obviously not possible on physical boards so we limit physical boards to an expansion range of eight. Physical boards also don't require infinite rulers to identify their grid spaces with, and so using A-Z and 1-26 instead of negative letters and numbers may be preferable.
Reinforcements are new pieces placed after the game has started (like Shogi). There are two ways to gain reinforcement pieces. The first is by having control of the midboard during expansion (you must have more pieces in the midboard than your opponent) and the second is from your opponent taking more than a set time to make a move. The first method gains you a number of reinforcement pieces equal to how many more pieces you had in the midboard (up to 3). The second method is only applicable in games with time controls and can help games end faster; this mechanic may not be applicable to games against an AI engine (we aren't sure yet). If a player takes longer than a minute to move, it skips to the other player's turn and awards that other player a single reinforcement piece. Both reinforcement methods can be stacked for a given turn.
When you gain a reinforcement, it can only be a fortifier or aggressor, but it can be placed anywhere on the board as long as it is not past the opponent's starting line. An exception to this is if your Commander is past the opponent's starting line, as you can also place a reinforcement within one unoccupied square of your Commander no matter the position of your Commander.
You do not get additional time during your turn to place reinforcements, they must be placed before any other movements are made that turn, and you must place allotted reinforcements. If you place a reinforcement Fortifier in a position that would also allow for promotion, you may also promote that piece in the same turn before moving.
Tentative Rules
Direct checks (next-turn single-movement mates) need to be called but indirect checks
(next-turn multiple-movement mates) do not. This means double movements from attrition
or double captures from Aggressor pieces can mate and end the game without warning.
Draws: if your Commander gets captured but you could have similarly captured your opponent's Commander in the very next turn, you may do so and call the game a draw. A draw can also be called if the players agree to end the game without a clear victor. This will usually occur when it becomes clear that the game will go on infinitely and neither player has the will to outlast attrition.
Truces are when both players agree to pause the timer on the game and come back at a later time. This is a necessary function for indefinite and correspondence games.
Time controls: to make Stratic games capable of continuing infinitely, the default game mode for Stratic uses time controls that count up instead of down. There must always be a limit on the amount of time a player can take to move (with 1 minute being an ideal time in our tests), and if that limit is hit, it becomes the other player's turn, with a single reinforcement piece granted from the idleness of the prior player. There will obviously be time controls that count down as well, to force games to complete so the human players can move on, but these finite games are less important to the ranking system.
Draws: if your Commander gets captured but you could have similarly captured your opponent's Commander in the very next turn, you may do so and call the game a draw. A draw can also be called if the players agree to end the game without a clear victor. This will usually occur when it becomes clear that the game will go on infinitely and neither player has the will to outlast attrition.
Truces are when both players agree to pause the timer on the game and come back at a later time. This is a necessary function for indefinite and correspondence games.
Time controls: to make Stratic games capable of continuing infinitely, the default game mode for Stratic uses time controls that count up instead of down. There must always be a limit on the amount of time a player can take to move (with 1 minute being an ideal time in our tests), and if that limit is hit, it becomes the other player's turn, with a single reinforcement piece granted from the idleness of the prior player. There will obviously be time controls that count down as well, to force games to complete so the human players can move on, but these finite games are less important to the ranking system.
The two on the left were the starting formations of the first game ever played, before
the rules for piece movements and board size were finalized (this setup is no longer
possible). The game was played by Snax (top) versus Sphlem (bottom).
Planned Variants
True Endless: capturing the Commander no longer ends
the game and every captured piece is a piece the opponent gets to place back on their
next turn, so the number of pieces each player has never goes down and no win
conditions can be met.
This variant is designed to be truly endless so we can host The Endless
Tournament, in which all users of the site are automatically added to a side in a
single endless game. You can vote for a piece to move on your side and for any
reinforcements to be placed. Your ELO is the weight added to your vote, meaning if you
have 1500 ELO then it is like 1,500 votes were cast for whatever move you selected. At
the end of the one-hour voting period, the movement with the most votes is what
actually gets played for your side. At the end of each week, the side that scored the
most points through captures is the winner of that week's cash prize which is then
distributed evenly between all the accounts that voted on moves that actually got
played.
Antistratic: the goal in this variant is to lose all your pieces first. Just like Antichess, the main mechanic is that if a piece can capture, then it is forced to capture, and the Commander is no longer a special piece since all pieces must be lost in order to win. This is difficult because having a presence in the midboard means you are likely to get reinforcements, and reinforcements must be played.
4D Stratic: add one dimension to the board so it goes from a 2D square grid to a 3D cube grid. Pieces still move and behave the same as normal but have another dimension to act in. We call this 4D instead of 3D because the normal game is technically already 3D. Just as normal, multiple pieces can occupy the same space in the 3D cube grid and so there is technically another dimension of space being utilized, making it 4-dimensional. Plus the joke of 4D Chess. This also works to loosely simulate space combat.
Antistratic: the goal in this variant is to lose all your pieces first. Just like Antichess, the main mechanic is that if a piece can capture, then it is forced to capture, and the Commander is no longer a special piece since all pieces must be lost in order to win. This is difficult because having a presence in the midboard means you are likely to get reinforcements, and reinforcements must be played.
4D Stratic: add one dimension to the board so it goes from a 2D square grid to a 3D cube grid. Pieces still move and behave the same as normal but have another dimension to act in. We call this 4D instead of 3D because the normal game is technically already 3D. Just as normal, multiple pieces can occupy the same space in the 3D cube grid and so there is technically another dimension of space being utilized, making it 4-dimensional. Plus the joke of 4D Chess. This also works to loosely simulate space combat.