Everyone knows the classic: three in a row on a 3×3 grid, you against an opponent, X and O. But the version I really wanted to build is the Double-Three 5×5 mode, where the board is bigger and a line of four counts as a double — worth more, and far harder to set up. The small game is a warm-up. The 5×5 is the one with real depth.
Play the AI across seven difficulty levels, or pass and play against a friend on the same device. The top levels use proper lookahead, so the easy draws you're used to from childhood don't come for free anymore.
On the 3×3 board, perfect play from both sides is always a draw — the classic grid is a handshake, not a fight. The larger boards are where the ZAP version earns its name, because three-in-a-row logic stops being memorizable and starts being tactical. Seven AI levels run Beginner to Cyborg. Its pencil-and-paper sibling Dots & Boxes stays sharp at every size.
Tic Tac Toe was a notebook game for me — the margins of school notebooks, friends trading Xs and Os while a lesson happened somewhere in the distance. And every notebook player eventually learns the same disappointing truth: 3×3 gets stuck. A lot. Two players who both know the handful of standard responses will draw forever; the classic grid is a solved handshake, a politeness ritual conducted in pencil marks. I kept it in this build because the handshake is tradition, and traditions matter — but the reason this page really exists is what happens when the board grows.
The design call at the heart of it: scaling the win-length with the board size. Bigger boards don't just give you more room to play the same tiny game — they raise how many marks in a row you need, which changes what you're required to build, and suddenly Tic Tac Toe stops being solved and starts being a real fight. The larger boards are the real game. 3×3 is the museum lobby you walk through to reach it.
And this is where the player mistake lives, because everyone walks onto the big boards carrying the 3×3 script: grab the centre, take the corners, block the obvious line. That script is dead weight out here. On a large board, occupying the centre wins you almost nothing — there's far too much space for any single square to dominate. The big-board game is about two things the small game never taught you: double threats and open-ended runs. A double threat is a move that creates two ways to complete a winning line at once — your opponent can block one, never both, and the game is decided the moment you build it. An open-ended run is the raw material: a line of your marks with empty space at both ends, threatening to extend in either direction and forcing an answer every single turn.
So the discipline flips from occupying to constructing. Build runs that stay open at both ends; a run blocked on one side is half as dangerous and your opponent knows it. Watch their runs with the same paranoia — the moment an opposing line has open air on both ends, it's already a countdown. And always be hunting the fork: the intersection square where two of your potential lines cross is worth more than any centre square ever was.
Play 3×3 for the nostalgia — I still do, and it still draws. Then pick a bigger board and meet the game the notebook version was hiding all along. The daily board arrives at midnight, identical for everyone. Bring threats, not scripts.
And a closing tip for the transition: on your first few big boards, spend one full game just watching what the AI builds instead of racing it. The shapes it prioritises — the open pairs, the crossing points — are the curriculum. Steal the shapes, then start winning with them.
First, unlearn the thing everyone believes: on the 3×3 board, the centre is not always the best opening. Against a sharp opponent a corner often gives you more ways to build a threat, because it sits on more potential lines than people expect. Try opening in a corner and watch how many more forks open up.
And forks are the whole game on the small board. You don't win three-in-a-row by trying to make one line — you win by creating two threats at once so your opponent can only block one. On the 5×5 board it goes further: because a line of four counts double, the strong play is to build toward overlapping fours, where one move threatens two scoring lines at the same time. Stop thinking "how do I complete a line" and start thinking "how do I threaten two."
Plain 3×3 tic-tac-toe gets solved by your brain after a few dozen games — every line becomes automatic, and a smart player gets bored. That boredom is exactly why I built the 5×5 Double-Three mode: it reopens the game, with enough ways to win that you have to actually think again. I spent a long time scribbling these grids in the back of my school notebooks, not to win but to figure out how the other person thinks — what they block, what they miss, where they overcommit. Reading an opponent like that carried straight into chess for me; I got better at the bigger game because I'd learned to anticipate a person on the small one. That's what this trains: not just your own plan, but a read on the plan across the table.
The skill is the fork — setting up a situation where you win no matter what the other player does. That's a real strategic idea, and once you feel it here you start seeing it everywhere. The AI difficulty climbs so that the cheap childhood draws stop working and you're pushed to plan properly. No download, no sign-up — play in your browser on any device. If you want a deeper version of the same forward-thinking, try Dots & Boxes.