malvasia bianca

I can’t think of another puzzle game like SPL-T: it’s deterministic, but it nonetheless has the “responding to random circumstances” feel of a game like Tetris or Drop 7.

That was especially the feel for the first week or so when I was playing the game, when I didn’t realize how the drops work. When you create a 2×2 (or 3×2) pattern of equally-shaped rectangles, they lock in place with a count; that count is the number of moves you’ve played so far, and when that count goes to zero, they disappear. The rectangles above them fall down; importantly (but not obviously when you’re first playing around with the game), locked rectangles have their count cut in half when they fall.

So, once you realize this, the basic parameters of your play are set: you’re going to want to start from the bottom of the screen and work up, so that, when something disappears, there will be stuff above it that gets closer to disappearing.

The game ends when you’ve got the entire screen locked up; there’s a score as well as a move count, but in practice I ignored the score, and I don’t really think it pulls its weight: you get different amounts of points for locking large rectangles compared to small ones, but it wasn’t a dimension I felt like optimizing along when playing, and just to survive I was going to spend most of my time diving the screen into as small squares as possible anyways.

So, basically, I thought of the screen as divided into four columns, each of which was 8 squares high, and I wanted to divide those squares into 2×2 blocks in a controlled fashion. I would mess up at that sometimes: in particular, every once in a while I would accidentally finish a square while not realizing that I’d already divided the adjacent half of a neighbor square, so I would make a 3×2 block instead of a 2×2 one, messing up the adjacent column for the rest of the game. So I spent a week or two trying to get better at avoiding mistakes like that.

Eventually I decided to think a little more about what was going on. I’d like to maintain a steady flow of 2×2 squares disappearing, so I’ll be creating new ones at the same rate as they’ll be disappearing. Which raises the question of how long it takes a square to disappear: if I create it on turn N, then it has to count down from N in zero, except that if it falls it gets cut in half. If it’s at height H and if I space out the numbers properly beneath it, then it will fall (H-1) times, so really I want to divide N by 2^(H-1); it takes 3 moves to finish a square, so I figure the right spacing between when I finish a block at height H-1 and one at height H is somewhere between 2^(H-1) and (2^(H-1))*3*4. (The 4 is the number of columns.)

I suspect that there’s something useful that you can do with this; the big problem is that drops really screw up your timing, because if you’re targeting a difference compared to the previous K*2^(H-1) and then there’s a drop beneath you, all of a sudden you’re targeting K^2*(H-2) and the number you’re comparing the difference with also drops in half. The first of those isn’t so bad (after all, I’m pretty vague what the correct number is for K, but I think anything between 1 and 12 is workable); but there’s a big difference between being in range of 200 and being in range of 100.

At any rate: that thinking helped a bit, but only a bit. Soon after I started taking the 4-column approach, I had a lucky game where I hit 450 splits; my high score is still only 485 splits. I get over 400 more often now than I did then, but it’s still quite rare; if I get less than 200 splits, I’ve actively screwed up, 200–300 means that I’ve done a bad job, but 300 400 is the normal range for me.

I’m glad I spent time with the game, and I still come back to it occasionally, hoping that I’ll find the magic timing that lets me reliably get past 400 and even reach the unknown reaches of 500. Maybe I should try a more horizontal strategy instead of focusing so much on the left two columns…