🪂 Drop Merge

Drop Merge: One Tile at a Time, and the Chain That Pays for Everything

You are handed one numbered tile at a time, and you have exactly one decision to make: which column. The tile drops to the top of that column's stack and stays there. Land it on a tile with the same number and the two become one tile of twice the value — and if that tile now sits on its own twin, they merge too, and so on down the column. Everything worth knowing about Drop Merge is about setting up those chains before you need them.

📏 The Rules, All of Them

• Tiles carry powers of two: 2, 4, 8, 16, 32 and upward. The tile in your hand is shown as Now; the one after it as Next.
• Drop Now into any column with room. It lands on top of that column's stack.
• If it lands on a tile with the same number, the two merge into one tile of double the value, and that value is added to your score. The merged tile then checks the tile beneath it, and keeps merging for as long as the numbers match.
• Merges are vertical only. A tile only ever merges with the tile directly beneath it, at the moment something lands. Tiles side by side never merge, so 16 | 16 | 16 across the bottom row is perfectly normal here.
• A full column can still take a tile if the tile matches its top: the pair merges, so the column doesn't grow.
• You have three Undos per game, and one Discard to start, with more earned as you build (see below).
• The game ends when no column can take the tile in your hand and you have no Discard left.

Nothing ever leaves the board: tiles only combine, so your biggest tile is always sitting somewhere on it.

🔄 Not 2048, Whatever the Tiles Say

The doubling ladder is borrowed from 2048, and that is where the resemblance ends. In 2048 a swipe moves every tile and merges every matching pair at once, and the new tile appears wherever the game decides. Here nothing slides. Every tile stays exactly where you put it until a merge takes it away, and the only merges are the ones you cause by landing something, so each drop is permanent in a way a swipe never is. In return you get two things 2048 never gives you: you choose where each tile goes, and you can see the one after it. A player who reads Next is playing a different game from one who doesn't.

🪜 The Staircase: the One Technique That Matters

A tile merges only with an equal tile, so a chain needs each tile in the column to be exactly half the one beneath it. That shape is a staircase — 32, 16, 8, 4, 2 from the bottom up — and dropping a 2 on it collapses the whole column into a single 64 in one drop.

Five merges from one tile: 2+2, 4+4, 8+8, 16+16, 32+32. That single drop scores 4 + 8 + 16 + 32 + 64 = 124 points and frees four rows.

So the whole game is this: keep each column sorted big-at-the-bottom with halving steps, and feed the top. A staircase can have gaps — a 64 with an 8 on it is still sorted; the 8 just needs an 8, then a 16, then a 32 before the 64 comes into play — and a gapped staircase beats the alternative below every time. Measured over 300 games on the 4 × 6 board, a bot that does little more than keep its columns sorted merges on 52% of its drops and fires a chain of two or more on 24% of them. Its longest chain was six, a full column collapsing to one tile, which is the most six rows can hold.

🧱 Burying: the Mistake That Ends Games

Put a big tile on top of a small one and the small one is buried: it cannot merge with anything until the big tile above it is gone, and that only happens when an equal tile lands on it. Bury a 2 under a 64 and you need to build a second 64, on top of it, before that 2 ever does anything again.

Same two tiles. On the left the ringed 2 is out of the game until you build another 64 and land it on top. On the right it is live, with a 64 waiting underneath.

Small on big is a staircase. Big on small is a burial. When a tile fits nowhere clean, take a gap in a staircase over a burial, and if you must bury, bury the cheapest tile you have, in the column with the least invested in it.

🎯 Reading the Hand, and What You'll Be Dealt

Next is always on show, so use it. If Now is a 4 and Next is a 4, land Now somewhere clean and drop Next straight on top of it: a merge you made rather than one you were given. An 8 followed by a 16 wants a column whose top is a 16 sitting on a 32, so it can take both in turn.

What you're dealt follows one rule, plus a wrinkle. Everyday tiles run from 2 up to your biggest tile divided by 8, and the top of that range is never below 4, so a fresh board deals only 2s and 4s; own a 64 and the 8s start arriving, a 128 brings 16s, a 1024 means anything up to 128. Every value in that range is equally likely. The wrinkle: once you hold a 32, a bonus tile one step above the range — your biggest tile divided by 4 — is dealt half as often as any everyday value. With a 64 on the board that's a 16 about one deal in seven; with a 2048 it's a 512 about one in seventeen. We checked those odds across 40,000 deals rather than assuming them. Bigger tiles bring bigger help, and bigger tiles that need somewhere sensible to go.

🗑️ Discard and ↩️ Undo: When to Spend Them

Discard throws away the tile in your hand and moves Next up. Spend it when the tile would force a burial in every column, and not before: a 2 that can only go on a 128 is a fair reason; a tile that fits a gap is not. You start with one, and the first 64, 128, 256 and so on each earn another, so a game that keeps growing keeps paying — our test bot earns between four and five a game on the 4 × 6 board.

Undo puts the board back exactly as it was, with the same tile in your hand, and the tiles still to come are not reshuffled, so it is no way to fish for a better draw. It's for the misclick, and for the chain you only saw after you'd let go. Three per game, and they don't come back.

⚠️ Four Ways It Goes Wrong

1. Playing only Now. Every drop that ignores Next throws away half the information you have.
2. Spreading small tiles everywhere. Four columns each with a lonely 2 on top are four half-built staircases that all need the same thing.
3. Letting a column fill with a big tile on top. A full column with a 2 on top is fine: any 2 can still land there. A full column with a 32 on top is nearly dead: only a 32 will ever open it again.
4. Spending Discards early. They are worth most late in the game, when a single unplaceable tile is the difference between game over and a 2048.

🎚️ Three Boards, and What the Numbers Say

The board sizes change how far you can climb far more than how the game feels. To put honest numbers on that, we played each board a thousand times dropping tiles at random, and three hundred times with a bot that does nothing you can't: it sees Now and Next, plans those two drops, keeps its columns sorted and avoids burials. It is not a strong player, so read its numbers as a floor.

Median score and average game length, over 1,000 random games and 300 bot games per board.

Two things stand out. Random play never built anything bigger than a 128 on any board, and on 4 × 6 it usually died with a 32 as its best tile, while the same rules played with staircases in mind ran four times longer and reached 512 in 19 games out of 20. The gap between those rows is the staircase, and nothing else. And 2048 on the classic 4 × 6 board is genuinely hard: our bot gets there about one game in eight. The 5 × 7 board is where 2048 becomes the expectation rather than the prize, and a 4096 turns up in about half of games; 3 × 5 is over in a few minutes and usually tops out at 256.

🎮 Playing on This Page

Tap or click a column to drop into it, or press its letter key — A S D F from left to right, with G for the fifth column on the big board — or the digits 1 to 5. The left and right arrows move an aiming cursor; the down arrow or Enter drops. Z undoes, X discards, N starts a new game.

Daily mode deals everyone playing the same board size on the same calendar day the same sequence of tiles, so scores can be compared honestly; it changes at midnight, your local time. Your best score is kept in your browser, per board size, with today's daily kept separately.

📜 Where Drop Merge Came From

The tiles are the tiles of 2048, which Gabriele Cirulli released in March 2014 as a free web game, itself a take on the earlier 1024 and on Threes! by Asher Vollmer, Greg Wohlwend and Jimmy Hinson. Dropping those tiles into columns instead of sliding them — the falling-piece frame of Tetris, the pick-a-column input of Connect Four — has no single credited inventor. By 2019 it was all over the app stores under names like 2048: Drop and Merge, Drop the Number and Merge and Double, and it has since found a home on the web as daily-board games such as Ponder Club's Double Up, the version that prompted this page. Ours keeps the genre's standard rule that merges are vertical only — several apps offer a mode where side-by-side tiles merge too, and it is a looser game — and adds the visible Next tile, three undos, and discards you earn by building.

🧒 A Doubling Game That Doesn't Feel Like Practice

Every merge is a doubling and the score is a running sum, so a game of Drop Merge is the two-times table run up past the numbers most children ever meet — 512, 1024, 2048 — without anyone calling it arithmetic. There is nothing to read, one decision per turn, and the undo button turns a mistake into a lesson instead of a restart. The 3 × 5 board is the place to start: a game is over in a few minutes, and every rule has shown itself within the first ten drops.

So: start on 4 × 6, keep every column sorted big-at-the-bottom, park the awkward ones on the right, and watch Next. When your first staircase collapses from a 2 all the way down to a 64 in one drop, you'll understand why people keep coming back to this one. 🪂