Are 5 grade points in the TGM3 fading roll achievable?

After discussing it a bit with @Qlex, here are some calculations and thoughts on the matter:

Let's assume the game runs at 60fps and the fading roll lasts 54 seconds = 3240 frames (if anyone has got a more accurate frame count, let me know).

Let x1, x2, x3, x4 be the number of singles, doubles, triples and tetrises respectively and let x5 be the number of remaining blocks at the end.
Clearing a line takes 2.5 pieces, so let P := 2.5x1 + 5x2 + 7.5x3 + 10x4 + 0.25x5 be the amount of pieces used in the credit roll.
Also let's define S := x1 + x2 + x3 + x4 to be the amount of line clears.
Now we want to account for ARE and line clear delays. At the fastest speed (1200+) ARE and line ARE are both 4 frames and the line clear delay is 6 frames.
Ignoring the fact that ARE doesn't and the line clear delay may not matter for the last piece, we get the following function to determine a piece/second value (pps):
f(x1, x2, x3, x4, x5) := P / [(3240 - 4P - 6S) * 1/60]

The grade point formula is 0.04x1 + 0.08x2 + 0.12x3 + 0.26x4 + 0.5 = number of grade points (the 0.5 are the clear bonus which you're obviously going to need to have a shot at 5 grade points unless you're a cyborg).
It's fairly easy to see that the ideal credit roll would consist of exactly 17 tetrises, 1 double and no remaining pieces (i.e. ending it with a bravo), since that nets you exactly 5.0 grade points with the minimum amount of line clears.

f(0, 1, 0, 17, 0) = 4.3174 pps

So even in a perfect roll you still need to place roughly 4.3 pieces (175 in total) a second on average. In a real roll you'll usually have about 30 blocks left and have cleared maybe 3 singles and 3 doubles and at that point you still need 16 tetrises (16*0.26 + 3*0.04 + 3*0.08 + 0.5 = 5.02) which bumps you up to ~4.86pps. Now let's look at a few instances where players have gotten over 4 grade points:

KAN's fading roll MV (video):
7 singles, 2 doubles, 1 triples, 12 tetrises, 34 remaining blocks => 166 pieces placed, 4.22 grade points, ~4.0753pps

Kevin's fading roll MV (video):
13 singles, 2 doubles, 0 triples, 11 tetrises, 10 remaining blocks => 155 pieces placed, 4.04 grade points, ~ 3.7744 pps

That means that even KAN would need to play quite a bit faster and get a bit luckier (regarding non-tetris clears and remaining blocks) or both. In addition to that the ideal roll above requires you to not lose any frames through HOLD. It's very easy to lose 50-100 frames total if you change your mind about a piece and hold it while it's active a few times. That may not sound much, but 100 frames are already over 3% of the roll.

Now, I hear your counter-argument: "C'mon buddy, the top 40L players easily go over 5pps, it's not that difficult "
I think there are a couple of reasons why this comparison doesn't hold up:
1. TGM3 limits you to 6 frames of DAS, whereas I assume the top 40L players have set it lower than that (I know that Microblizz uses 4 frames of DAS). Depending on how low you can go without not being able to tap anymore, that saves a few frames per piece that needs to be DAS'd which I'd wager is at least 30% of all pieces.
2. TGM3 has a less generous randomiser than bag. This one is hard to quantify, but I do think it matters.
3. The credit roll requires you to place 175+ pieces, however, 40L players aim for 100/101 pieces, which means you have to keep up the pace for a 75% longer game.

Another point is that if you want to get the MO rank through the fading roll in the actual game, you have to first play a draining 5-7 minutes of master mode (and not get messed up by the short break before the credit roll). The only upside is that you can sneak an I piece into the roll thanks to the HOLD box

In conclusion I strongly believe that it's not possible to get 5 grade points in the fading roll, but I'd love to be proven wrong. I'd also like to hear @Amnesia's thoughts, because as far as I know he's of the opinion that it's doable. If I messed up the formula somehow, miscounted the line clears in one of the videos or if there's some other factor I'm overlooking, let me know.

 

Possible by TAS, probably; possible for a human player? Very very unlikely. If it was, KAN would likely have already done it.

 

Does anyone know the absolute maximum PPS in Ti?

 

Here's a naïve lower bound for a TAS without fancy techniques like finishing with an "anti-bravo" to save the frames from the line clear delay:
Let's assume each piece takes 4 frames of active time on average (it's probably less). A tetris then takes 4*10 + 4*10 + 6 = 86 frames. That would get you floor(3240 / 86) = 37 tetrises and 7 more pieces, so a total of 377 pieces / 54 seconds ⩯ 6.98pps. That's also more than 10 grade points.

 

Personally I think an MO (5 grades) is still possible, yet not right now. Even our highest skilled TGM3 players aren't perfect so there is still a fair amount of room for improvement. Though if it were to happen it would definitely be a one-off thing as it'd probably revolve more around good RNG from the pieces than someone who'd be that fast they could consistently get said grade.

Just like the 'secret grade' in Sega Tetris, maybe that could be a nice challenge for TGM3 GMs to try whilst TGM4 slowly yet surely comes into existence.

 

I think you would have to rephrase your question! Everything you wrote only proves that you can achieve 5 grade points. But that is pure theory. Much more interesting would be how likely it is for someone with the right skill to get a piece sequence and play it perfectly enough to really get it. So isn't it that what you really want to know?


From all the information you showed here I would conclude: 5 grade points are achievable! Will it happen within the next 20 years? Almost certainly not.

 

I agree, in hindsight I should have asked how likely people find it to be.
Unfortunately I think it's very difficult to calculate a good estimate for that probability without having more data. In particular I really want to see ZAB, the current shirase WR holder (by a 20 second margin) attempt this challenge.

 

Well...There might be a problem in your calculation but I am lazy now to re-develop it.
First, what we can see on Kevin's video is definitely not at 3,77 or it means that I am playing at max at 4,5+ tps today...
Second, my calculation gave me something around 3,25 tps to reach 5 grades. I am constantly thinking about this since my only one chance to get MM is 5 grades in Fading. So I know what I am talking about.

 

Now I am a bit drunken to read your whole calculation sorry....

 

hmmù
mm yes exactly

 

3,3 tps... That's my result.

I must place a minimum of 178 tetriminos within 54 sec. 178/54 = 3,2963

Kevin and I's top speed is far above 3,3 so I think Kan's is probably around 4,5. He was just lazy to report it but he probably made it already.

 

How did you get 178? 17 tetrises and one double should be 175 pieces. And yes, that would be the real TPS, I just subtracted ARE and line clear delay to make it more comparable to 40L.
But let's use your way: In Kevin's fairly recent Shirase 0-500 in 1:52 video in the 300-500 segment he places 117 pieces in about 38 seconds (3,08tps) (there's also less line clear delay in those sections than in the credit roll, but here it's only a 64 frame difference). Since that 0-500 is his record, I'd assume that segment would come close to his top speed, so I doubt it's far above 3,3.
I also had a look at your Texmaster 4 grades video where you place 158 pieces in 54 seconds (2.93tps). How much faster have you become since then? What's your current grade point record in Texmaster?

 

Ah....Maybe I was overestimating it
I was refering to a very old Max speed than HEBORISC7 had recorded at 3,21 and this was in 2009...So I thought I had reached 4+ (on a short 10sec rush) in 2016. But maybe I have just improved from 3,21 to 3,4 in 6 years...Possible.

I had counted 17 tetrises + 1 triple to get 178 tet

For the fading record, I think I have made something 0,1 more but that's all..

 

Yeah, but your argument already is a statistical one if you bring in the games of the best players to be far below the necessary speeds. That argument is only valid if we agree that those games are already very extreme outliers which makes it very unlikely to see even more extreme ones. If the only data you give to someone are those two games it is more like "do you believe someone ever beats Kan/Kevin by that margin?".

 

If we accept the unreal result of 4,88 tps which is wrong for me, it is above the maximal speed ever reached on TI by japanese players on 1 section. I am pretty sure after the hundreds of videos that I have seen of Kan, that he already performed more than 4,8 grades in the fading, my maximum is at 4,14 and very slow compared to Kan, with way more mistakes.

I will stay on my result of a 3,3 tps with a clean field.

 

@TGGC: Yes, that's what my argument basically boils down to, as well as the fact that 40L players who purely go for speed and tetrises also would be hard pressed to do this under the same conditions. So in a way I'd also count that as data, because 40L is a very similar discipline and the world record has been fought so hard over that we can't expect huge improvements anymore.

I also do really want to know that, but like I said I don't know how to determine this accurately. That's why I settled for making my case and having a poll, but I realise that this is still somewhat subjective and that I haven't actually proved anything.

@Amny: To reiterate, what I called "pps" in the OP are not pieces per second, but pieces per second adjusted for delays. I just measured the speed after taking out all the delays in order to draw a comparison to 40L where there's no delays. It's essentially a measure for the amount of active time being wasted. Another point for this was that you could play at a certain speed clearing only singles or clearing only tetrises; with the formula I suggested the former would get a worse "pps" value. So yes, your 3.3 are correct.
That said, could you point me to the video of KAN doing 4.8 grades?

 

Mathematicans already studied how to model this, an example can be found here: http://info-100km.web.cern.ch/info-100km/files/Statistics_of_records.pdf Problem would still be to get enough data for even a rough estimate.

And there is basically no way to factor in a new technique or even bug, someone might fight someday which might make it a bit easier.

 

could you point me to the video of KAN doing 4.8 grades?


I have not said that there is one video of him doing it, but if me, kevin, DIGITAL and nahu can do 4,10 - 4,15, then I think KAN can do 4,8. It is obvious, his maximal speed today is far above us.

 

Why don't you just TAS it first...

 

@K: The title should say "...achievable by a human player?". It's my bad for not wording it like that, but otherwise there really wouldn't be a question about it. It's obvious that you could TAS it, so there's no point in making one.

@TGGC: Thanks for the link.

@Amnesia: Ok, that's how I initially interpreted it, just thought I'd ask.