Tetris Slowruns Explained
Vložit
- čas přidán 12. 03. 2022
- "It's too slow" -andrewg1990
This video discusses one of the more overlooked NES Tetris categories, one where the goal is to perform the longest game possible. It might seem simple from the surface, but there's an abundance of strategy involved.
Check out NGC here: / @ngc_man25
Thanks for watching :) let me know what other NES Tetris stuff you want to see next.
#tetris #nes #speedrun - Hry
Correction: at 0:54, I mentioned that a piece falling on level 0 takes "a full second and a half" to fall one unit down the board. The correct value is ~0.8 seconds. I must've doubled the time it took somewhere during the script writing process, and failed to double check that value along the way. Apologies!
I noticed that too, because NES Tetris runs on about 60 fps
Actually since level 0 is the slowest level in the game, it's the most important part of the run, assuming you can't play 29. A bad placement on level 0 could waste a full 48 frames or more if you really mess up, but piece orientation changes the amount of cells it takes to hit the floor. That combined with the need to take pretty much all singles makes it very important to use the time on level 0 to think.
But there is an optimal time for 0-29. We might never get it, but it exists. There is no optimal time for 29+, or at least, we may never get remotely close to it.
isn't wasting frames the point here
@@kaja3932 I meant wasting frames as in saving time instead of losing it
Cheez: "There's a record I *don't* have? Oh. Give me an afternoon."
I've seen some comments about how long specifically to delay transitioning from level to level. The actual answer is complicated, mostly because it depends on board state, your current/next piece around the time of each speed increase, and potentially your risk tolerance.
I did some analysis using conservative assumptions about how a close to optimal game would go, and I have a table with optimal transition delays for each speed. If you want to know more about this category, leave a like and comment on this video to force Cobalt to make a part 2!
Fractals 29 start game lasted 40 minutes :) Nice video, thanks!
great video! love your vids, can't wait to see more in the future, whether tetris or not :)
Here when a level 18 start lasted over an hour (alex t)
0:54 "Pieces take 48 frames to fall just one unit down the board, a full second and a half."
That's not correct. The NES runs at about 60.1 fps. So 48 frames is about 0.799 seconds. NES games have twice the frame rate of normal broadcast TV but half the vertical resolution, using progressive scan.
a correction: 48 frames is not a second and a half its 0.8s
Oh yeesh, I think I accidentally doubled the rate.. can’t believe I didn’t catch that in post. Thanks for the clarification :)
You can redeem channel points for NGC Man to attempt another run. Get in there boys B)
Love the video, glad I could help!
at a glance, wouldn’t the strat to be to only kill 1 row at a time? and have multiple holes per line to prevent a line as long as possible (in the lower levels)
factual description
How about doing this on the PAL version?
Here before Cheez hits 1 hr with Lv29 start
Since lvl 0 is so slow freaking slow, would it be beneficial to build a "terrible stack" with a bunch of holes, maybe fill up half the screen? You can then reduce it with single lines at higher levels. Levels 1-2-3 will be a bit faster, but you could spend a lot of time at level 0.
Weirdly analytical answer from someone who was wondering the same.
Early in the game, one level increase basically cuts the fall time by two rows of level 0. Aka falling from top to the bottom on level 1 is falling 2 rows short of that on 0. Level 2, 4 rows short, etc. This holds up until level 7 or so.
The most pieces you can possibly place and still be on level 1, is 95 (19 lines and 190 filled squares on the board, for 380 tiles total)
The question you asked can be answered by figuring out how many of those 95 pieces, optimally, you place before transitioning to level 1. This is assuming a slow run that ends when you hit level 2, but by the stuff above it should scale all the way to the end of a longer slowrun.
The first 25 pieces have to be placed on level 0. Then you have the first choice. Do you place all 70 of the remaining pieces on level 1 or some of them on level 0. If you place all 70 on level 1, then at best the first 25 all fall a full screen, and the remaining 45 fall on average, half a screen. That's 47.5 full screen falls on level 1 in this approach.
If you wait 25 more pieces before transitioning to level 1, then those pieces all fall, on average, 3/4 of a screen at level 0. This is about the same as falling about 5/6 of a screen on level 1. Then you transition to level 1 and your best option is to dig back down to maximize piece falls. With 9 singles that get to fall about half a stack each, you dig half of what's been left behind. The rest of the pieces stack as densely as possible to the top like in the other case. The average amount each falls is 3/8 of the board. The end result is 25×5/6 + 9×1/2 + 31×3/8. That's 37 full screen falls, significantly less than the other option.
I think this scales to prove that when level speeds are spaced like in the early game, you should continue to stay low. It's a different analysis when considering 18->19 or 28->29 but those matter less.
@@biggiemac42 that is actually a bit surprising. looking at the splits at 2:09 (8 min 08 sec for level 0 and 4 min 49 sec for level 1), I was expecting the pieces on level 0 to fall much slower than at level 1. Thank you for the analysis.
You don't want to fill up the screen because that means there's less time for the pieces to fall. If you watch the 100 lines speedrun, they will fill up the board to make it higher up so there's less time for them to fall.
The strat is to only get singles and keep the board low.
Is it really better to focus on a low playfield in the early levels? I'm thinking that there could possibly be a better balance to be found between taking singles to keep the board low, and letting more pieces fall at the slowest possible speed, even if they don't get to fall all the way to the bottom. From a mathematical standpoint it's not clear to me that playing for low would be optimal.
See my reply to sideshownicful, I went into the math of that and confirmed staying low is best for the first several levels at least
@@biggiemac42 Thank you! Very clear!
@@biggiemac42 this is actually not correct. The first several level transitions demand delaying the transition by roughly 5 to 10 pieces. The only exception being the level 0 to level 1 transition, where the optimal transition delay is 2 pieces
roller that lines out into level 45 after a slowrun:
As a fan of Tetris and F1, I don’t approve that jajaja.
Good video man
So now we only need EricICX to do one of these and *then* take it to the glitched colours after the transition, and beat both records in one, in NTSC 😂
First a joke gets taken seriously by players. Then by analysis. Then by casters????
alex t's 16 mil game lasted an hour lol
watch cheez run 45 mins post 29 and blow this record out of the water.
It's too slow.
no way lvl 29 is allowed :skull:
At the lower levels they should fill up the sides with pieces
bigmacdontcare (czcams.com/channels/LDCJNtQIdbUcHP-5zpTilA.html) explained why they don't in another reply, so I'll paste that here.
Early in the game, one level increase basically cuts the fall time by two rows of level 0. Aka falling from top to the bottom on level 1 is falling 2 rows short of that on 0. Level 2, 4 rows short, etc. This holds up until level 7 or so.
The most pieces you can possibly place and still be on level 1, is 95 (19 lines and 190 filled squares on the board, for 380 tiles total)
The question you asked can be answered by figuring out how many of those 95 pieces, optimally, you place before transitioning to level 1. This is assuming a slow run that ends when you hit level 2, but by the stuff above it should scale all the way to the end of a longer slowrun.
The first 25 pieces have to be placed on level 0. Then you have the first choice. Do you place all 70 of the remaining pieces on level 1 or some of them on level 0. If you place all 70 on level 1, then at best the first 25 all fall a full screen, and the remaining 45 fall on average, half a screen. That's 47.5 full screen falls on level 1 in this approach.
If you wait 25 more pieces before transitioning to level 1, then those pieces all fall, on average, 3/4 of a screen at level 0. This is about the same as falling about 5/6 of a screen on level 1. Then you transition to level 1 and your best option is to dig back down to maximize piece falls. With 9 singles that get to fall about half a stack each, you dig half of what's been left behind. The rest of the pieces stack as densely as possible to the top like in the other case. The average amount each falls is 3/8 of the board. The end result is 25×5/6 + 9×1/2 + 31×3/8. That's 37 full screen falls, significantly less than the other option.
I think this scales to prove that when level speeds are spaced like in the early game, you should continue to stay low. It's a different analysis when considering 18->19 or 28->29 but those matter less.