Spill is a game about edges. Every tile holds exactly as many drops as it has neighbours — two in a corner, three on an edge, four in the middle — and one drop past that edge dyes the neighbourhood. The rules fit in a sentence, but the game underneath was tuned by simulation, and the numbers it left behind are a better teacher than instinct. This guide is built from them.
Corners are an opening, not a plan
The corner is the board's safest real estate: it primes in one move (capacity two) and only two tiles can ever attack it. Taking corners early is right. Staying there is fatal — our corner-only test bot lost 100% of its games, most inside 16 moves, because the middle of the board is where chains live. A centre tile touches four neighbours, which makes it slower to prime but four times as connected when the cascade comes through. The measured shape of good play: claim cheap safe tiles early, then spend the midgame building connected threats toward the middle.
Feeding is the losing habit
A tile one drop from its capacity is primed — in the game it trembles and glows, and it is the single most important object on the board. The mistake that decides most games is placing ink next to an enemy primed tile: when it spills, your tile is dyed along with everything else, so your drop was a donation. The CPU's own evaluation docks a tile almost its entire value the moment an enemy primed tile touches it, and that one idea — prime your own, never feed theirs — beat a bot that simply grabbed the most ink every turn in 92% of simulated games. Before every move, look at what your target touches.
Fuse networks: two primed tiles are one bomb
Chains are not luck; they are built. Two primed tiles side by side are a single bomb with two fuses — triggering either spills both. A line of them is a wick across the board. Because a spill dyes its neighbours and refills them, a chain running through your own primed tiles costs you nothing and converts everything it touches, so the strongest midgame is quietly wiring your territory together while denying your opponent adjacent real estate. One measured move of caution: cascading is a trade — the spilling tile empties. Popping a fuse into two enemy drops is even; into six it is a massacre. The lookahead bot that waited one extra turn before triggering beat the threat-aware bot 83% of the time, and patience is most of that gap.
The seat does not decide the game — we checked
Chain-reaction duels have a folk reputation for crushing first-player advantage. We went looking for it before shipping: across three board sizes, all four stages and every CPU tier playing mirror matches, the first mover won between 42% and 53%. That is why Spill has no komi or handicap rule — the absence is a measurement, not an oversight. If the same person keeps winning at your house, the seat is not the reason. The chains are.
Terrain: holes are accelerators
On the Wells and Ring boards, missing tiles are not neighbours — so every tile around a hole has a reduced capacity and primes sooner, behaving like an extra corner in the middle of the board. Those are the flashpoints: chains ignite there first, and the safe-tile map is genuinely different from the Open board. Quick, the 5×5 board, simply gets everyone to the avalanche sooner — our simulated games there ran about 38 moves against roughly 60 on the full grid. The stage select is on the title screen (S key, or tap the BOARD chip), and which terrain you pick changes the opening more than anything else in the game.
Against the CPU: Easy is a greedy grabber that plays a quarter of its turns at random — beatable by anyone who avoids feeding. Normal plays the full threat-and-safety heuristic. Hard adds a move of lookahead and refuses to hand you chains; beat it by building wider fuse networks than it can afford to answer. All three play by identical rules through the identical move generator. Load Spill, watch for the trembling tiles, and remember: the player who spills first is rarely the player who spills last.
