SDA logo

Released in August 1996 for the SNES, Tetris Attack is a puzzle game with swapping tiles, which does not actually have anything to do with Tetris or attacking. Tetris Attack was previously released as Panel de Pon in Japan, but Nintendo decided that skinning the interface with characters from Yoshi's Island would increase the game's popularity tenfold. They were correct.

 

Note: Runs on stage clear mode are not allowed because they make people fall asleep.

Best single-segment versus mode time: very hard 0:04:19 by Craig 'cyghfer' Gordon on 2010-03-29.

Get Flash to see this player.

Author's comments:

I've been playing this game ever since it came out when I was a kid. I remember being beaten constantly by my older sister until I began practicing when I arrived home in the afternoon, while she was still at school. In a week or so I had surpassed her, and I eventually became known as a master of the game to my family and friends. I never had anyone to play with who could match my skill level. Even as I grew older, I liked the game so much that I continued to play it sporadically. I've probably sunk more hours into the game than any other.

As I watched many of the speedruns on the SDA, I sought to start creating my own runs. Tetris Attack was not one of the games I initially had in mind for speedrunning, but after watching Darkwing Duck's good, though beatable, run many times, I decided to pursue this game first.

This run got enough good luck on enough stages to achieve a final time of 4:19, thanks in part to some strategy changes from the previous run. First off, I try to start my chain at most after raising my stack one row, which both saves time in the standard sense and also increases the chances of the computer having a relatively high stack when I drop my garbage. For the first 6 stages, my ideal execution is to send a x5 - x7 chain block (depending on the exact stage, the accompanying combo blocks I've made, and how big the computer's stack is when I'm completing a chain), a 4-combo, and a 5-combo. As Darkwing noted in his comments, generally a x8 or so is sufficient to just fill the computer's screen, and creating chains any larger is a waste of time; he also noted that a combo will help fill any gap if the chain was one line short. I took this idea even further. Creating a 5-combo is still faster than creating a new chain link. Compare: clearing two extra blocks vs. clearing three extra blocks + time blocks are suspended in the air after clearing but before falling + block falling time. Although creating two combos instead of one doesn't save that much time, it often means the difference between a :14 and a :15. Note that I need a 5'er as the second combo because two 4-combos will only fill one line. I generally found trying to make more than two combos during a chain too strenuous. And if I don't manage to scrounge up two or even one combo, I just add chain links accordingly.

From stage 7 on, I change my strategy so that my ideal execution is a 3x chain block + a 4-combo, and then another 3x - 5x chain block (with extra garbage as needed). Darkwing mentioned this in his comments but chose to continue with the large chains strategy. The later opponents usually do clear the first round of garbage anyway, but there are nonetheless a few reasons why I prefer the segmentated strategy for them. First, it opens up the slight possibility of achieving a 'quick kill,' wherein the computer's stack is near the top and yet it fails to clear the small chain block sent. Unfortunately, this does not occur in this particular run. The second reason has to do with the process of garbage blocks falling into the opponent's playing field. Anyone who's played the game much will have noticed that larger chain blocks create a loud rumbling sound effect and shake the playing field more upon impact. Most people have probably also implicitly realized that larger chain blocks give the receiver more of a buffer time to recover; this means that more time passes between impact (at which time the stack ceases to rise) and the continuation of the stack rising, which wastes more time in terms of the run and, more importantly, gives the computer more time to reduce his field or find a way to clear the garbage. Furthermore, this buffer time can be extended by clearing blocks, clearing garbage or having garbage fall one or more rows. Thus, reducing the buffer time increases the probability of a relatively quick finish in many respects. If the computer is not engaged in clearing blocks/making combos/making a chain at the conclusion of the garbage-clearing process it will pause for some time after the block hits the stack (crucial) and then either reduce the stack or switch blocks to clear the garbage. You can get away with large garbage blocks in the early stages because the computer's cursor moves slowly enough; all of the later computers will clear barring ridiculous luck. But if the block is only of medium size (3x - 4x, I think), it will cause only a medium buffer time, greatly decreasing the computer's chances of clearing.

One more note about sending a second round of garbage: say you complete a x4 chain with an accompanying 4-combo. If you start another chain before all the garbage from the previous barrage has fallen (including the combo), the combo block's drop will be delayed until the second chain is completed and has already fallen. Thus, if this situation occurs, and you complete another x4/4-combo, the following will occur: the first x4 chain will fall, then the second x4 chain will fall, and finally both 4-combos will fall. This defeats the purpose of throwing in a chain with the first barrage of garbage, and is the reason I will sometimes pause or cease chaining after completing the first chain.

Thus, in the later stages, assuming the computer clears the first round of garbage, the ideal situation is 1. drop a x3/4 block 2. computer clears while you are constructing your second chain, and while his stack is 3-5 lines from the top or so 3. your second chain block falls while the first garbage is still clearing 4. the garbage does not fall after clearing and the computer does the after-clear pause. Unless the computer's cursor is sitting right next to an easy clear this situation will usually lead to a high :10s/low :20s finish. The third main reason I send a x3/4 first is because it is too large a block for all the later computers to ignore, and since it takes <10 seconds to send to the opponent's play field, his stack will usually be high enough to achieve the ideal situation.

I execute my strategies effectively in all but two or so stages. On to the individual stage comments:

Stage 1 - Lakitu
God, Lakitu is such a bitch to start a run against. Compared to the next two stages, Lakitu doesn't raise his stack as high at the start, and thus is usually pretty low (maybe 4 lines high on average) by the time I've finished my chain, which I generally opted to go the full x7/4/5. This time I suffice with x5/4/5 and Lakitu fails to clear, netting a very nice :13.

Stage 2 - Bumpty
Bumpty and Lakitu together ended about 80-90% of my runs. It is pretty likely that when your chain drops, the computer will be in the process of making pointless matches around its field, and won't pause the cursor upon impact. It is also rather likely that an easy clear will be available that even these early computers can reach in time. Well, I get good luck in that regard this time around, but I mess up my chain early in the process due to not seeing the purple block that rises in line with the other two. That is seriously the thing that annoys me most during Tetris Attack: making a match you didn't intend because a block from the rising floor helps you out. What's worse is that I fail to see the easy chain continuation with the green blocks, and thus have to start a new chain. Still, I recover and get away with only a x2 plus a few combos, since Bumpty needs to take the time to clear the little 6-combo block I send and keeps his stack high. Final time is :15, which I consider 'solid' for the first four stages.

Stage 3 - Poochy
Similar to Bumpty - doesn't reduce as quickly as Lakitu, still pretty slow cursor movement. I notice his stack remains pretty high and only send a x5/4 block. You'll see that he tries to 'stall' with regular matches when the garbage drops but dies when going for the distant yellow blocks. Again, a nice :15.

Stage 4 - Flying Wiggler
I have similar problems here that I usually have with Lakitu - Wiggler's stack is usually pretty low by the time I send garbage his way. I opt for the full x7/5/4 and luckily there is no easy clear for the computer; however, he makes one match and stalls for just long enough to push the time to :17. My general target time going into the 2nd third of the game is actually 1:00 so I was on point so far. If any of the first four clear I restart.

Stage 5 - Froggy
The next four stages tend to be a tossup in terms of how high the computer's stack will be when the garbage block falls. I underestimate Froggy here and only send a x5/4/5 block, though he actually responds kindly and clears immediately instead of reducing and leaving lines of space near the top of his stack. I send some more garbage his way and he goes down in :22, which is a decent time given one clear.

Stage 6 - Gargantua Blargg
Darkwing mentioned that Mr. Gargantua is tough to take down quickly - it's been rather the opposite in my own experience, often going down without a clear most often of the mid-run opponents. I send a x6 with an excess of combos and he does an inadequate job of reducing, resulting in a stellar :14.

Stage 7 - Lunge Fish
This stage, along with stage 4, tends to wall me for some reason, I don't really know why. This time around Lungefish reduces more than he usually does, both before I send my initial x3/4 block and while it's clearing, so that my subsequent x5/4 doesn't reach the ceiling (seriously, what an ass - ending with only one complete row of non-garabage). I'm forced to send another combo along while the computer slowly lets the final row rise, wasting time that definitely shouldn't've been wasted. I think I wasn't expecting Lungefish to clear as much as he did and I don't make an effort to send more garbage after my second barriage - and in addition I didn't want to delay the combo blocks in the second barrage. The hesitancy in sending the final combo was my first really costly mistake of the run - about 3 seconds lost. The stage ends in a disheartening but still acceptable :26.

Stage 8 - Raphael the Raven
My favorite character in the game : ] (not my favorite music though - Forest Stage for sure there). I believe this was the best luck I ever got on Raphael with speedrunning in mind, recorded or not; he clears very consistently and tends to combo a lot during the initial dozen seconds or so. This time he barely clears anything before I send my initial x3/4 block, which fills his entire playfield. After getting a few easy matches while the initial garbage is clearing he stops near the bottom, enabling the pause after the garbage finishes clearing. He then goes for a clear with the reds at the top, which is unfeasible for Raphael's speed level. Really great :16 here, and a good example of how the segmented-chains strategy is effective.

Stage 9 - Hookbill Koopa
Unfortunately Hookbill busies himself with some small chains and doesn't clear until after I send my second round of garbage. :27 is a pretty standard time for the Cave stages.

Stage 10 - Naval Piranha
Worst stage of the run here in terms of execution. I'm not exactly sure what my plan was at the start of the level, but I stupidly start off by matching and not 4-comboing the purples, which leads to yellows that are positioned in a way deleterious of continuing the chain. It's actually possible to switch the yellow on just the right frame so that the game recognizes it as 4-combo chain link 2, but I'm too flustered to even come near pulling that off. Then later, after dropping the first chain-block, my second chain ends after link 3 due to switching the light blue and green blocks instead of dropping the light blue onto the two greens. Thus my set of chain blocks ends up being too small to cover Naval's entire field. After that my nerves fully kick in and I piddle around accomplishing nothing for a full 8 seconds or so. Eventually I send more garbage and Naval goes down in :34.

I'm honestly not sure whether he would've died at around :21 if I had sent just one or two more lines of garbage; it's hard to tell if enough non-buffer time passes before clearing, as he finds a clear pretty quickly after fully reducing. My guess is that he would've died, but I'll forgive myself since this was only the second instance that I got to the late stage of a speedrun while recording.

Stage 11 - Kamek
Though I make a mistake at :10 it wouldn't've made a difference in the final time, given the same execution by Kamek. I simply get bad luck here, as Kamek has easy clears (I could've gotten a very quick kill if, say, Kamek had had another stack on the right to reduce when my first garbage block fell) and starts a chain from a random block off the garbage (most annoying occurrence while specifically speedrunning the game). I end up with a :36 through no fault of my own, which is quite heartbreaking given the extraordinary pace I had going into the Cave, but not really bad enough to cause a restart so late in the run.

Stage 12 - Bowser!!!!
Bowser clears twice but the after-clearing pause does its job, and the run ends with a sound :24. As you can see I was quite eager for the run to end and dick around praying for Bowser to not pull off a bunch of bullshit chaining.

Though the final time is a large improvement over the previous run, I'm not quite satisfied with both my performance and the time itself. I really get pretty lackluster luck in the Cave, which plain sucks, as I believe my time going into it had been my best ever for that point. In addition, while practicing the game unrecorded (on a TV with HD lag to boot!), I happened to get a ridiculously lucky run ending in sub-4:00. Though I could've kept going after I got this 4:19 run and tried to match the luck of that unrecorded run, I felt that 53 seconds was a good enough improvement for submission. One can only handle so much dedicated puzzle game running before offing oneself. My eventual target time is sub-3:45 - whence I will never look at a stage timer in this game again.

Thanks to the SDA management for running and maintaining the site. Thanks to Darkwing Duck for providing an enjoyable run and the framework for speedrunning it myself. I'll be there to take the record back when you beat this run :W

If you read through all of my long-winded and wordy comments, thank you! I hope you enjoy my first completed run!

Return to the Game List, the FAQ, or the Home Page.