A swipe contains two different events: the current board slides and merges, then a new 2 appears in an empty square if the board changed. Predict the first event exactly before judging the whole move. Otherwise a surprising spawn can hide a mistaken prediction about the merge itself.
Read toward the edge you selected
For a left move, remove gaps in each row mentally, then read from left to right. Equal adjacent values combine, and the newly created tile cannot merge again during that move. Thus [2, 2, 2, 2] becomes [4, 4, 0, 0] before the spawn. It does not become 8 in one move.
Now inspect [2, 0, 2, 4]. Compression first puts the two 2s together. They merge into 4, leaving the original 4 beside them, so this also becomes [4, 4, 0, 0]. The identical row result hides a difference: the first example made two merges and gained 8 points, while the second made one merge and gained 4.
For a right move, perform the same reading from the right edge and place the result against that edge. Vertical moves read columns rather than rows. Do not carry a left-to-right pairing order into every direction; the chosen edge determines which neighbours meet first.
Count usable space instead of cheering every motion
Let E be the number of empty cells before a changed move and k the number of pair-merges across the whole board. Each merge reduces the occupied-cell count by one. After the single new 2 is added, the empty-cell count is E + k − 1. This is an exact accounting relation for this Double version.
A changed slide with no merges consumes one empty cell. One merge merely offsets the incoming tile and leaves the empty count unchanged. Two merges create one net additional empty cell. The result explains why a board can feel tighter even after a swipe that rearranged it attractively.
Empty count is not a complete evaluation. Two boards with the same free space can have very different access to compatible neighbours. Use the count as one check, then inspect whether the new arrangement brings useful values together or strands a small tile between larger ones.
Use an unchanged move as information, not an escape
A left move on [2, 4, 2, 0] leaves that row unchanged: the unequal 4 blocks the two 2s from merging. More pressure on the same direction cannot push one 2 through the 4. On the whole board, however, other rows may still change, so a single unchanged row does not establish a whole-board no-op.
When the entire board is unchanged, the game does not add a tile or increment the move count. If no legal board move remains, it marks the round lost. Repeatedly selecting an unchanged direction cannot create a better spawn because it creates no spawn at all. Look for a direction that changes relationships elsewhere.
Keeping a large tile near an edge can help maintain order, but treat that as a plan you can inspect. Before an opposite-direction move, identify which useful pairing or empty corridor it creates and which ordered region it disrupts. An edge preference does not override a concrete move that prevents an immediate dead end.
Separate a guaranteed merge from an assumed future
The new tile is always 2 in Merge Double, but its empty location is selected by the game’s random source. The pre-spawn board is determined by your current board and direction; the next placement should not be silently treated as under your control. The move lab stops before this spawn and lists the empty squares where a 2 could appear. It compares the deterministic slide without pretending to choose the next placement.
Suppose a swipe creates adjacent 4s at the chosen edge. That future merge is visible in the pre-spawn result. In the live game, wait for the actual new 2 and inspect all affected rows and columns again. In the lab, keep any later spawn hypothetical and state its location explicitly. Do not claim an exact multi-move route while leaving an unspecified spawn out of the state.
The target is a 128 tile. A high move score does not itself satisfy that goal, and the sum of tiles is not the score. Merging preserves the total tile value; a spawned 2 raises it by two. Score instead records the values created by merges along the route.
Two experiments to try
EXPERIMENT 1
Same row result, different history
Use the deterministic row exercise with a left move. Keep spawned tiles separate from the displayed pre-spawn result.
- Predict the outcomes of [2, 2, 2, 2] and [2, 0, 2, 4] without moving.
- Apply each once and count the number of original pairs consumed.
- Compare the matching row results with their different gains of 8 and 4 points. Then predict a later left move on the two 4s before adding any new complications.
What to observe
A newly made 4 cannot join the other 4 in the same move, although they can merge on a later suitable move.
Why this comparison is useful
The rule applies once per participating tile per swipe. Matching final rows do not imply matching merge histories or score gains.
Your change and observation:
EXPERIMENT 2
Forecast the space budget
Choose a practice board with at least one empty cell and compare two candidate directions from the same starting board.
- Count the current empty cells. For each direction, predict the total number of pair-merges across all four lines.
- Compute E + k − 1 only if that direction changes the board. For an unchanged board, retain E and expect no spawn.
- Apply the move in the lab and recount the empty cells in its pre-spawn output. For a changed move, subtract one to account for the coming 2; its location does not affect that count. Inspect all listed spawn destinations before assuming a future pairing stays accessible.
What to observe
A move can improve tile order while reducing free space, or gain free space while disrupting a planned edge.
Why this comparison is useful
The lab verifies the deterministic move and shows possible spawn destinations; it does not add the new tile. Empty-cell accounting predicts the count after any one spawn, while evaluating a particular arrangement requires specifying or observing its actual location.
Your change and observation:
Check your decision
Before the spawn, what does a left move do to [2, 0, 2, 4]?
Answer and explanations
Predict compression, pairing and score before adding the new 2. That keeps the deterministic result separate from the spawn.
- [8, 0, 0, 0], because all values can combine immediately.
Not supported. The new 4 from the two 2s cannot merge again in that same move.
- [4, 4, 0, 0], gaining 4 points.
Supported. Compression brings the 2s together; their one merge creates 4 beside the existing 4.
- [2, 2, 4, 0], because gaps prevent merging.
Not supported. The row compresses before the adjacent-equal merge scan.
Your next practice session
For your next five changed moves, predict the pre-spawn board and net empty-cell change before swiping. Correct prediction mistakes before judging the larger strategy.
Scope and checks
This guide describes the current Playfield implementation. The experiments are proposed ways to practise, not measured player results. Rules were checked against the game source; the written scenarios and explanations received a separate review.
- Playfield Arcade game implementation — Primary source for this browser version’s board rules, scoring and state transitions.
Playfield Arcade · Updated 4 October 2026 · Original practice design, AI-assisted writing and code.