Threes

The game that came before 2048

Threes is the original that 2048 was built from, and the differences are exactly what make it trickier. Two of them matter most. First, tiles slide only one square per move: press a direction and everything nudges a single step, not all the way across the board, so you have far more control but far less sweeping cleanup. Second, not everything merges: a 1 and a 2 combine to make a 3, and only after that do equal tiles double up: 3 and 3 make 6, 6 and 6 make 12, and on up. Those little 1s and 2s are the raw material for everything, and mismanaging them is how a board clogs and dies.

The history explains the polish. Asher Vollmer and Greg Wohlwend spent over a year refining that one-step rule before Threes shipped in February 2014; within weeks, 1024 and then 2048 had rebuilt the idea with simpler merge rules, and the pair eventually published their fourteen months of design emails to show how much iteration the original took. Plenty of players still rate Threes above its imitators for exactly the reason it frustrates newcomers: the 1-and-2 economy forces real decisions from the very first swipe.

Feed the ones and twos

Because a new tile slides in from the opposite edge of whichever way you swiped, every move is also a placement decision: you're choosing where the next tile enters. Keep 1s and 2s from piling up by pairing them into 3s promptly, since a stray 1 with no 2 to meet is dead weight taking up a cell it won't give back. Each tile can only merge once per move, so you can't ram a whole row together in a single swipe the way you might expect; combining is patient work, one pair at a time.

A worked chain: three swipes to a 12

Picture a single row reading 1, 2, 3, 6 from left to right. Press left: the 2 folds into the 1 to make a 3, and everything behind steps forward, leaving 3, 3, 6. Press left again and the pair of 3s becomes a 6. Now the row is 6, 6. One more press makes 12. Three swipes, one tidy chain, and it only worked because the numbers sat in ascending order pointing at the wall. That staircase is the shape you're always trying to build: each merge manufactures exactly the tile the next merge needs. Mind the cost, though. Since a tile merges at most once per move, the chain takes a swipe per step, and every one of those swipes also admits a new tile from the far edge. Three moves of cashing in means three new arrivals landing behind you, which is why chains should point into the corner you've committed to, where incoming tiles can't wedge themselves between the steps.

Anchor the big tile

As in its famous descendant, park your largest tile in a corner and build a descending run of numbers alongside it, so merges line up naturally instead of stranding a big tile in the middle of the board. The practical way to enforce that is direction discipline: pick the two directions that point into your chosen corner (say left and down) and make the bulk of your moves with them. New tiles then arrive only from the right and top, in front of your stack where you can process them, rather than sneaking in behind your biggest number. Swipe the other way only when forced, and budget a move or two to repair the damage afterwards. Watch for the checkerboard, too: a board of alternating big and small tiles with no equal neighbours is quietly dead even though it looks half empty. The cure is dull and reliable: pair off 1s and 2s the moment they appear, so the small numbers keep flowing upward instead of freezing in place.

Habits that clog the board

The classic beginner death is hoarding: leaving 1s and 2s scattered around for later while chasing merges among the big tiles. Small tiles aren't savings, they're traffic. Every unpaired 1 is a cell producing nothing until its 2 arrives, and a 1 walled into a corner behind larger numbers can sit there for the rest of the game. Reactive swiping is the other killer: answering each new tile individually, in whichever direction looks locally best, shreds your corner structure within a dozen moves. Slow down instead; there's no timer, and the board is small enough to genuinely read. Before each swipe, work out which edge the next tile will enter from (always the one opposite your swipe) and ask whether it lands as fuel or as clutter. And don't panic when the grid fills up: the game only ends when no swipe would move or merge anything, so a packed board with one live pairing left is still a game you can save.