Add support for up to 6 players.

README.md

@@ -1,8 +1,8 @@
 # Super Auto Pets: The Board Game — Online
 
-A web app for playing the Super Auto Pets board game remotely. 1v1 for now;
-the engine is built to grow to more players. Play a friend by room code, or
-play solo against a computer opponent (easy / medium / hard).
+A web app for playing the Super Auto Pets board game remotely, at 2, 4, or 6
+players. Play friends by room code, fill any empty seat with a computer
+opponent (easy / medium / hard), or play solo against a table of them.
 
 ## Stack
 
@@ -81,10 +81,41 @@ Six rounds, each with its own shop tier deck. Per round:
 Apples and bees are **temporary**: they leave your deck after the battle.
 Most trophies after round 6 wins.
 
+## 4 & 6 player mode
+
+A game seats an even number of players — 2, 4, or 6 — so everyone has an
+opponent every round. A lobby with an odd number of humans fills the gap with
+a bot; any seat can be a bot, so a table of one human and five computers is
+just as valid as six humans.
+
+Above two players the shop is shared but the battles are not: each round
+splits the table into pairs, and every pair fights its own battle
+simultaneously. The pairings come from the fixed table printed in the rulebook
+(`internal/game/schedule.go`) — a round-robin that runs everyone past everyone
+before replaying earlier rounds to fill out the six. All of it is public: you
+can replay any table's battle, not just your own, and peek at any player's
+lineup afterwards.
+
+Two rules change with the bigger table:
+
+- **First shopper** — at two players the priority token does double duty
+  (shops first, acts first in battle) and passes from a winner to the loser.
+  With more players it's a separate token that starts at seat A and walks one
+  seat along each round, while every battle flips its own first player.
+- **Packs** — the rulebook asks for one pack per pair (2 packs for 4 players,
+  3 for 6). The host picks them in the lobby and their tier decks shuffle
+  together: all the tier 1 cards into one tier 1 deck, and so on. Two players
+  may combine packs too, for a deeper shop.
+
+Ties on trophies are settled by counting back from the last round: whoever won
+round 6 takes it, else round 5, and so on. Players with identical records
+share the victory.
+
 ## Packs
 
-Three card packs are playable, chosen by the host in the lobby. Each pack is
-six tiers, one per round, and every pet ships as two copies.
+Three card packs are playable, and the host combines one or more of them in
+the lobby. Each pack is six tiers, one per round, and every pet ships as two
+copies.
 
 ### Turtle pack
 
@@ -164,12 +195,18 @@ web/                 React frontend
 ## The computer opponent
 
 Any seat can be a bot (`Player.IsBot`); humans and bots are interchangeable
-to the engine, which is what will let future >2-player games mix them
-freely. The AI in `internal/ai` never touches the `Game` — it decides from a
-`game.View`, the same per-player state a human client is sent, plus a
-persisted memory of public observations (battle lineups, the event log, shop
-row changes). It cannot see your deck order, hidden trade picks, or the
-shuffled shop decks. It scores candidate moves by Monte-Carlo battle
-rollouts (`game.SimulateBattle`) against sampled guesses of your deck and
-ordering, blended with a long-term deck-value heuristic; difficulty tunes a
-softmax over the scored moves plus the rollout budget.
+to the engine, so a table can mix them freely. The AI in `internal/ai` never
+touches the `Game` — it decides from a `game.View`, the same per-player state
+a human client is sent, plus a persisted memory of public observations
+(battle lineups, the event log, shop row changes). It cannot see your deck
+order, hidden trade picks, or the shuffled shop decks. It scores candidate
+moves by Monte-Carlo battle rollouts (`game.SimulateBattle`) against sampled
+guesses of your deck and ordering, blended with a long-term deck-value
+heuristic; difficulty tunes a softmax over the scored moves plus the rollout
+budget.
+
+At a bigger table it keeps a model of *every* seat, not just one rival — it
+will face each of them eventually, and every battle is fought in the open —
+but plans each round against the specific opponent the schedule pairs it
+with. Its card counting spans the whole table's known cards and the combined
+contents of the packs in play.