Thread in 'Competition' started by Amnesia, 22 May 2007.
Sorry TGGC, but you're wrong.
You can block the well with a Z at any height you want. It's just risk/reward management & RNG for when you get the opportunity.
If we were TASing this, you can clear a Tetris roughly every 10 blocks, so that means 9-10 per section. (Edit: Except I forgot that clearing lines advances levels too, so this is wrong. Oops.) That's equivalent to 12-13 triples. That's about 5 seconds saved over 0-300 just in line clear delays. And that's not even counting the fact that in normal play, you'd be clearing some combination of Tetrises & triples, and you save ~10 frames on each Tetris now. So in practical terms, you're saving time not just on being able to clear additional Tetrises instead of triples, but also on the Tetrises you already would have had in your run.
This would all be purely academic, except Kashiwa's run already smashed world record pace by ~10 seconds on 0-500.
I couldn't let you stay ahead of me for long!!
Gm in 10:47:01
Thats possible, I am just unable to find the error. I started to rethink when I saw people doing this and came to the same conclusion as before.
GyRo that does not match with the number I calculated. (But as I said, I might have made a mistake! I cannot explain 10s difference)
No, you cannot at any height. You need 4 lines above to build a tetris and another 3 to rotate the I piece. So the highest would be 12, to stop the I piece at height 13.
If you were TASing this you would realize you finish section 1 after Tetris 5. And Section 3 around Tetris 19 - 21, depending how low you finish the section.
The absolute maximum according to my calculation is 13 * 21 = 273 frames.
And I am not talking about TAS but a strategy a human executes. A good player gets 100-150 frames on average is what I expect. You introduce some problems in the execution which eats away from the theoretical max: You cannot clear with I or L, but only I. You introduce less leeway while stacking for tetrises, as you have less vertical space. Every skim hurts your time even more than before. And in the end you have to fix the hole you introduced.
Oh sorry simonic, I didn't saw that one you were referring..
Actually you don't need that particular "proof of concept" to show you can make a sub 1min at section 0-100... if you want to make a substancial demonstration you would need to play a full game in this configuration (keeping the messy half bottom the whole time attack). But after level 300 it might be annoying to play with a something like a 10×10 playfield.
IMO, some people might be misleaded by Kashiwa's demonstration : the tetris save time but the real deal is the induced high stacking obligation. Without it, player naturally tend to secure lower stack.
I'm being a little bit rough with the sketch of the situation. "Any height you want," for instance, is meant to be interpreted reasonably. You could place it a quarter of the way up the well to be safe or halfway up to be aggressive. It is assumed reasonable parameters are already known. Similarly, we all know that some of the first section is spent building up the stack & that no RTA run will be able to play with pure Tetrises. The rough calculation of Tetrises to triples for a section would still apply to 100-300, and it was only intended as a quick, handwavey sketch of the maximum time savings due to line clears. In practice, only a portion of the time savings comes from the difference in line clears, and the rest comes from the lower frame cost for Tetrises that would already have been cleared as Tetrises.
tl;dr version: Yes, you could model this more precisely, but why bother since we can already watch it in action?
(Except for the mathematical interest, of course. But there's no need for the practical concern now. The principles are understood & have already been experimentally verified.)
I'll have to interrupt the strat-talk really quick...
TGM1: GM in 11:10:40
PB by over 1:20. Isn't that big of a deal since I pretty much haven't been playing TGM1 for a year now. Unfortunately no fancy speed-strats.
Next goal is sub11.
No, 9-10 tetrises is wrong for every section. 10 I pieces clearing a line are 50 levels on its own.
Because IMHO real world games support my claim. E.g. the savings in Kevins 9:19 game are 63 frames max (9 I pieces clear 7 frames early). There is easily 1 additional line clear you could attribute to this strat, as he does 3 doubles to clear the hole. So its 23 frames saved max, or isn't it?
Oh goodness! You're quite right about the Tetrises counting towards levels too. What an obvious oversight on my part.
I think you've already provided an explanation for the missing frames though, in that the saved time on Tetris clears (whether due to # of line clears or fewer frames spent dropping) isn't the only thing contributing. You also save time on all other pieces generally simply by being forced to play so risky with no safer downstacking at all. We already know the saved time must be coming from something, since 777's run is essentially optimal with the more relaxed risk management.
I confess I don't really care about the exact numbers, just the principles at play, although very precise modeling could confirm (or disconfirm) that we've accounted for every contributing factor.
@TGGC, I would agree with @GyRo that there are two ways to see this, the way to adopt would probably be the empirical analysis : @KevinDDR uses this for his TGM1 PB, as well as Kashiwa for his 0-500 start, so that's worth something. The fact that so few records have that is probably because people haven't given it a serious try yet.
If you look at the other way, that's the theoretical one and we have proven that if you're good enough, you end up being faster (that part is obvious). I wouldn't try to make an in-between saying how many frames you get in a given game. Especially your calculation I find a bit too inclusive : I don't think that he does doubles because of the strat. He does doubles because that's TGM1
Went back to this for the first time in quite a while. Big improvement, but need to get to sub 11!
I am all for empirical analysis. I already talked in the chat about this. I assume if done right you could prove its impossible to find a game using this strat purely by time in a number of games, because the variance in times is far larger than the advantage you can gain by this technique.
Just to be clear, I am talking about the 3 line clears after 270. One is done with an I and one is done with an L. I think its highly likely that these pieces would have created more efficient line clears if the well was not blocked at this time. Obviously its impossible to prove, what the time would have been with a different strategy. However I see strong indicators Kevin would beat 9:20 without this. His perfect 400-500 had nothing to do without this strat beyond him talking about it all the time. Its all about the mindset, not the strat itself.
Thats exactly what I am talking about! You can save this time without blocking the well - so this is a different strategy in my book.
Just so I can understand your points a little better : Do you mean "because the time gains are minimal compared to the risks, then it is not worth doing"? Or do you simply mean that this technique should not be used in a marathon? The use of the word "prove" sets me off a little, because we were talking about empirical analysis.
But you lose time on all the Tetris clears & on the tradeoff to using mainly triples. And no one---not even 777---has the intuition to maintain that much stack height throughout without being forced to. Otherwise the world record would be about 10 seconds faster. When we're considering truly optimal human play, which is essentially where the TGM1 world record stands, all those little things add up. It's all contributing.
I think we all agree that this strategy is much riskier (in terms of dying) but the good news is that, if attempted opportunistically (ie only using it when the situation presents itself, not wasting time trying to set it up) you rarely ever lose time just for trying. Even when you don't gain time, you really don't lose anything either.
So close!!! Really need to pay more attention to the section change so I don't waste time.
Everyone can use or not use it however he wants. I just say: do not expect faster times just because of like 10 tetrises cleared some frames earlier.
I used the word "Proving" for empirical analysis, maybe my english is not correct here. I think you could show that with a high probability you cannot distinguish between those games by time alone or to express it more mathematical the difference of times produced by those strats are not statistically significant. The obvious problem here is to produce enough data of games which are comparable with only the strat changed.
I still think its useful to look deeper why this strategy seems to save 10s. Those 10s are composed of different things, some more useful than others.
It's absolutely true that the strat is more risky and hence higher variance. In fact, given some reasonable parameter for how bad dying is, it's probably a lower expected value.
But that's not how optimal play works. Optimal play lies at the top end of that variance, and from a purely mathematical perspective, higher variance would actually support our claim that you have a higher ceiling to push against. In practice, of course, this doesn't guarantee it's humanly achievable, but we finally have empirical confirmation that it is.
The highest level of TGM1 play is a matter of waiting for the randomizer to deal you a good hand and playing it optimally. Variance only affects how long you must wait for this opportunity, but it doesn't limit the possibility on those rare games where the randomizer does go your way more often than not.
Congrats to @KevinDDR for finally surpassing Jago.
I updated from the spreadsheet. If there are any video links missing, let me know. If you don't have your PB on video and you would like me to link your best recorded run, I can do that too.
After about 5 months of playing (about 1 on the NNC), I've gotten GM after several close calls.
Separate names with a comma.