Falling-block guides usually recycle scoring folklore. This one reads the engine. Our browser
game, Block Stacker, implements the modern standard mechanics, and each constant in the source
implies a specific strategic decision. Once you know what the code pays for, the
correct openings and endgame habits follow from arithmetic.
The board and the seven pieces
The playfield is 10 columns by 20 rows. Seven tetrominoes exist, I, O, T, S, Z, L, and
J, and the engine draws them cyan, yellow, purple, green, red, orange, and blue
respectively. Piece distribution uses the 7-bag randomizer: the engine fills a bag with
all seven pieces, shuffles it with a Fisher-Yates pass, and deals from it. When the bag
empties, it refills.
This rule matters strategically. The longest possible drought for any piece is 12
deals: your wanted piece can come first in one bag and last in the next, which puts 12
pieces in between without it. Counting pieces since the last I tells you when the
next I piece is due, which is the foundation of well play.
What the engine pays, exactly
Line clears score from a fixed table multiplied by your current level:
- 1 line, 100 times level.
- 2 lines, 300 times level.
- 3 lines, 500 times level.
- 4 lines, 800 times level.
The multiplier makes clears increasingly valuable as you survive. At level 1, a four-line
clear is worth 800. At level 5 it is worth 4000. Four single clears in one level pay 400,
while one four-line clear pays 800, so the classic advice is literally true in this engine:
holding out for four-line clears pays double per line compared to singles.
Movement also scores. Soft dropping a piece one row earns 1 point per row. Hard drop
earns 2 points per row dropped, then locks immediately. Hard dropping everything is the
fastest way to farm drop points and the fastest way to make placement mistakes. The
engine rewards speed and punishes it through the same key.
The speed curve and when it stops getting faster
Gravity starts at 800 milliseconds per row at level 1. Each level subtracts 60
milliseconds, computed as 800 minus level minus 1 times 60, and the result floors at 50
milliseconds. The floor becomes active at level 14, so from that point on the engine
cannot drop pieces faster. What keeps killing runs at high level is your input budget,
and that is where lock delay and DAS enter.
Lock delay is 500 milliseconds. A piece resting on the stack locks after half a second.
Moving or rotating it resets the timer, and the engine allows at most 15 resets per
piece. That cap exists to prevent infinite stalling, and it means you get 15 saved
inputs per piece to slide or spin it into place. At high levels, spend them on
corrections, and never burn resets early while a piece is still high in the well.
DAS, hold, and ghost
Delayed Auto Shift is tuned to 170 milliseconds initial delay and 50 milliseconds
repeat. Hold a direction key and the piece steps once, pauses for 170 milliseconds,
then slides every 50 milliseconds. Competitors call this the movement rhythm, and
knowing it lets you place pieces with the key held instead of tapping.
The hold slot follows one strict rule: you may hold once per piece. The flag resets
only when the next piece spawns. So the standard pattern is hold an early I for the
well, then let the slot sit occupied, because swapping again mid-piece is refused. The
ghost piece shows the exact landing position by simulating the drop, which removes all
guesswork from column choice. Use the ghost, and your only remaining error source is
rotation state.
Rotation: why SRS changes strategy
The engine implements the Super Rotation System with two kick tables, one for the I
piece and one for the rest. When a rotation collides, the engine tries up to four
offsets, and the I piece uses different offsets than the others. This is what makes
wall kicks possible, letting pieces rotate into slots that look blocked.
The strategic consequence: never treat a blocked rotation as final. A T or L piece
locked against a wall can often spin into a notch one row lower. SRS rewards players
who attempt the rotation before manually repositioning. It also enables T-spin-like
placements, although this engine scores only the line table above, with no spin bonus.
A deterministic improvement plan
1. Keep the stack flat and reserve one clean column, ideally the well mouth at the
far side, for I pieces.
2. Hard drop only when the ghost confirms the landing spot. Otherwise soft drop the
final rows.
3. Hold your first I. Do not touch hold again until the current piece locks.
4. Count pieces since the last I. Prepare the well for it before it arrives.
5. Burn singles and doubles before level 10. After level 10, protect the well, because
each level multiplies the four-line payout by a larger number.
6. Use all 15 lock resets only for corrections near the floor of the stack.
Checklist for your next run
- The well is exactly one column wide and kept empty.
- Hold holds an I, and you swap at most once per piece.
- You read the ghost before every hard drop.
- You attempt rotation before manual repositioning near walls.
- You know your level, and therefore your gravity interval and multiplier.
The game Alexey Pajitnov built in the Soviet Union in 1984 needed none of these
numbers. Modern engines publish them, and players who read them stop guessing. The
code is the strategy guide.
What is your current high score and level, and does it die to stack mess or to speed?
Test the numbers above against the real engine in our
[Block Stacker game](https://webrecast.com/en/tetris) and see which single habit changes your
result first.
