Torris rewards reading the board, not memorizing openings — and the read changes with the board size, the mask, and whichever twists a preset turns on. This guide is built from the same geometry, timers, and even the bot ladder's own scoring math the game runs on, not generic tips that would fit any board game.
One name, three jobs: Torris is the game, Torris is also the hardest rung on the Kindling bot ladder, and Torris is what the rated Heat ladder calls anyone who climbs past 2200 Heat. Context makes clear which one a sentence means, but it's worth knowing all three exist before you meet them.
Twenty-five tiles, two flames spawned on opposite corners — a deliberate choice, not a center-opposite one: the corner start narrows the first mover's edge to about 57.5%, versus the roughly two-thirds a center-edge start would have handed them. There's no slow build on a board this small. The middle few tiles decide the match inside your first handful of turns, so your opening slide basically is your opening.
Forty-nine tiles, both flames spawned on the same middle row, opposite sides. The single dead-center tile sees more of the board in one queen-line reach than any other — the flame that claims it decides where the first walls of ash go up, because every tile it burns to get there becomes a wall the rival has to route around too.
Eighty-one tiles, built for three or four flames. It's large enough that the Backburn schedule below is doing real work — without a forced deadline, a board this size can outlast a reasonable game length on its own.
Rule three is blunt: no legal move on your turn and you lose. So what actually matters each turn isn't how much ash you've laid down — it's how many straight lines you still have open, your exits. Every Kindling bot from Ember up (short of the boss tier, which reasons differently — see the last section) ranks its own moves by roughly the same math: closing off one of your rival's exits is worth more than opening one of your own, so a trade that costs them a line and costs you nothing beats one that grows both sides evenly.
Watch for the moment your remaining ground and your rival's stop touching at all — no shared border tile, no unspent Surge that could leap back in, and, in a Draft game, no live Douse or Phase charge that could reopen a wall. Once that's true, the fight is already decided even though the game isn't: it becomes a race to outlast your own floor, not a hunt for a knockout, and the flame with more tiles left to wander wins it. Count your own remaining space, not just your rival's, once the board splits this way.
Backburn burns the board's outermost intact ring to ash on a fixed schedule, ring by ring, whether or not you're ready — the only cell it spares in a doomed ring is the one you're standing on. On the Standard 7×7 board it's ply 14 for the edge, 22 for the next ring in, 30 for the one after. On the Grand 9×9 board (Trio and Free-for-all) it's slower and wider: 16, 28, 40, then 52.
Treat the number as the deadline, not the ring's current look: if a plan depends on an edge tile surviving two more of your turns and the schedule says it's due before then, it's already gone. The upside — Backburn-consumed ground turns to the same obsidian a masked board is built from, and Surge can freely leap it, so a closing ring isn't automatically a dead end if you kept your leap in reserve.
Cross clears a two-by-two block from all four corners. Corners were the safe pockets on Classic — fewer queen-lines cross them, so a cornered flame usually kept one clean escape line. With the corners gone, nowhere on the board offers that shelter, and the knife-fight starts on turn one.
Diamond clips the board down to a rotated square, with angled obsidian eating more of each row the closer it gets to the corners. The practical effect: a diagonal slide reaches the obsidian wall in fewer tiles than a horizontal or vertical one does, so the "long" lines on Classic are the short ones here.
Crucible hollows a molten obsidian donut through the board's center. Both flames spawn on the ring, and the only ways around it on a plain slide are clockwise or counter-clockwise — unless you leap. Obsidian counts as non-intact ground for Surge exactly like burned trail does, so the dead center is a legal Surge target from your very first turn: one leap can put you on the far side of the ring in a single move, which no plain slide can ever do while the center blocks every straight line through it.
Surge lets your flame cross ash in one straight slide, once a game — and only on a line that actually needs it; the game won't let you spend it on one that was already open, so you can never waste it by accident. What it will do is sit there charged and read "No surge line open" the moment none of your current lines cross a single non-intact tile — common in the first several turns of a fresh Classic board, before enough trail exists nearby to leap over.
Obsidian counts exactly like burned trail for this rule. On Cross, Diamond, and Crucible the mask is already non-intact ground on move one, so your Surge can have a real target from your very first turn; on Classic it's dead weight until somebody actually burns something. Either way, the value of holding it climbs as the board empties out — an unspent Surge late in a nearly-burned game is worth defending far more than one on an untouched opening board, so don't cash it in early just because a line happens to be available.
Restores one of your own burned tiles — never a rival's trail, never obsidian, which never un-burns — to playable ground before you slide. It comes back scorched, so it plays like fresh floor but keeps its history visible. It's the direct answer to being boxed in by your own trail, and no help at all against being boxed in by anyone else's.
Burns one intact tile next to wherever you land — but never a tile next to any rival's ember, so it can't snipe ground beside your opponent. Its real job is denying a tile in your own neighborhood before your rival's next slide can reach it.
Lets your slide pass straight through a rival's ember instead of stopping at it — you still can't land on them, just ignore their body as a blocker. It's the direct counter to a rival sitting in your line: on Classic that kills the whole line; with a live Phase charge it stops mattering.
Spends its one charge to burn nothing at all — origin and path both stay intact. It's double-edged on purpose: the corridor you kept open for yourself stays open for your rival's very next turn too, so use it on a line you're prepared to have them use right back.
Kindling deals you a drafted ability but never gives the bot one back — every bot you spar faces you empty-handed, so you learn what your own pick does without seeing what it's like to play against a live Firebreak or Phase. That's a fine place to rehearse the mechanic; the real test of one ability against another only happens in a Draft match against a human rival.
Both run on the Grand 9×9 board with the same Backburn schedule, and both tighten your clock to 18 seconds a turn instead of Duel's 25 — with more rivals, the downtime between your turns is the real pressure, not the total move count.
A flame with no legal move on its turn is out: its current cell burns in its own color and stays exactly as impassable as any other trail — unwalkable, but still Surge-leapable — so even an eliminated rival keeps reshaping the board for whoever's left. One move, or one Backburn tick, can end more than one flame in sequence if it removes several rivals' last exits at once, so a single strong play in Free-for-all can be worth more than any single kill in a Duel.
Kindling's bot ladder runs Spark, Ember, Flame, Blaze, Inferno, Torch, and — at the top — Torris: the same name the rated Heat ladder gives anyone who climbs past 2200 Heat, which is where the disambiguation above becomes worth remembering.
Spark plays uniformly at random among its plain slides, Surge only as a last resort — no evaluation, no punishment for a sloppy line. Ember ranks moves the same way Flame does but picks randomly among its top three, so an occasional weak line slips through. Flame runs that identical ranking with no randomness left: the same greedy read, made exact.
Blaze is the real turning point — the first rung that looks one full move ahead and picks based on your best reply, not just the position in front of it. A tempo grab that looked free against Flame gets punished immediately from Blaze on. Inferno goes deeper again and starts weighing actual reachable space instead of just your immediate option count (that extra depth quietly shrinks in Trio, Free-for-all, or a Grand 9×9 duel, to stay inside a phone's frame budget) — a move that merely looks mobile stops working once something is counting the floor behind it.
Torch matches Inferno's depth but adds discipline: it treats its own Surge as too valuable to spend on a merely-fine move, taking a slightly worse plain slide instead — expect the same restraint from a human at that rating. Torris, the boss, thinks differently rather than just deeper: it grades every tile by who reaches it first rather than who merely can, refuses to spend its Surge in the opening on principle, and prices an unspent Surge as an asset that only grows more valuable as the board burns — never a coin it cashes in early for a small edge.