Thread in 'Discussion' started by user401730, 21 Jan 2020.
please consider these two wonderful videos and tell us: which is best?
this is not an obvious choice. i'll start with the observation that the score attack is more visually stunning. it suggests to me a shmup score attack, another absolute arcade classic.
I think this is a really interesting question with a lot of nuance.
Interpreting the question pedantically, real score attack would involve level stop abuse and despite my Ti Easy scores I think this is gross and bad. So I'd pick time attack.
More seriously though, I think this question is getting at the difference between playing for tetris clears vs speed.
I think of the effect of these objectives in terms of what they do to the balance between tactics and strategy. Most people use these words interchangeably but they mean somewhat different things. Strategy is stuff on longer time horizons, or following general heuristic rules. Tactics is more situational and improvised. The line between the two isn't super formally defined. I personally think of it as a continuum rather than a strict dichotomy. Strategy is when your actions don't care so much about finer details, eg. You could map out your actions with a small lookup table. And with tactics the fine details matter a lot, ie. It would take a much larger lookup table. Put another way, your actions are conditional on low frequency (strategy) or high frequency (tactics) game state information. This is kind of a bootstrapped definition that ignores how you even get to a representation of the game state where the concept of high and low frequency information makes sense, but hopefully you can forgive me and run with it. Humans are quite good at crunching high dimensional states into meaningful low dimensional states so hopeful it just intuitively makes sense.
Chess players talk about the difference between strategy and tactics all the time. If you want a less technical, more example driven description of the difference here's a random article:
In chess, as in Tetris, players find games that are primarily strategy driven as being boring and predictable, both to spectate and to play. This is part of the reason AlphaGo is exciting, because it plays more tactically than Stockfish and gives us hope that galaxy brained chess is tactical and beautiful rather than a strategic slog.
Anyways, back to Tetris. There are various gameplay elements that simplify the complexity of your board position, nudging your action map from a small to a large figurative lookup table. I can think of:
Just don't die vs clear tetrises
Hold vs no hold
Low entropy rng vs high entropy rng
Garbage (perturbed game state) vs no garbage
Many previews vs few previews
Slow vs fast play, or more specifically...
0G vs 20G
Long lock delay vs short lock delay
How this stuff plays out isn't always super intuitive. Adding previews at face value sounds like it gives you more information, so how could that possibly shrink your lookup table? Well, this added information is more than offset by the fact that you have so much extra control over your board, so you will avoid tons of game states in favour of sticking to a narrow safe subset of game states. A similar argument explains why adding Hold or a low entropy randomizer gives you a much smaller lookup table.
This isn't to say that all tactical things need to be maxed and that's the ideal game. High entropy randomizers force you into more game states, but these states could be unfair / have no winning response. You'd lose strategy but not even replace it with tactics. It's all about balance. Modern guideline games are extremely strategic (lol that sounds like a compliment out of context) to the point of being less interesting than they could be. An obvious game balance I haven't seen explored would be to have a completely random rng, just like the bad old days, but then give hold and previews to tame it. Something in this direction is probably the smallest patch that would spice up guideline.
ANYWAYS... Back to death mode. I gave all that background just to say that the two videos, to me, are like choosing between tetrises and speed vs speed alone. To me the choice is clear, I believe playing for both is more tactically rich than just playing for raw speed, despite how the leaderboard would rank the two plays. Just look at all those overhangs, sometimes filled, sometimes compromised into holes. Moments of control, moments of panic.
To me this is Tetris at its best.
Death (score) > TGM2 (in general) > Death (speed) > TGM1 > NES > TGM3 > Guideline VS >>> Guideline (in general)
I'm not sure I agree that speed + tetrises is more tactical than just speed. Watching some of these replays you open up a whole new bag of moves at the cost of losing tactics like overhangs. I don't have the experience myself, so I trust your judgement more!
Very good write up, it's nice to see an interesting thread topic, and some quality posts.
Well I don't even have a sub 6 GM in death yet! So there is a lot of extrapolation going on between my personal lookup table and the hypothetical galaxy brained version. But I hope it's clear, especially from my ranking, that I think speed is an essential part of the mix.
Not sure how I would describe my ideal... Fastest gold SK? Most tetrises with GM under 5:30? Those are getting pretty close! Some sort of blending rule maybe?
I wish reinforcement learning methods were a little more developed, seems like it should be possible to pick a goal, train an AI, and then "measure the lookup table". Whatever that would mean when you're not using Q-tables.
It is interesting to see someone else bring up the chess comparison with Tetris. It's something I've thought about before in relation to guideline, and also the general gameplay elements of NES and TGM.
Something I kind of had a feeling of whilst learning to play NES was that there are two kinds of placements in the game, which could be described as either strategical or tactical. The way I think about it is that when you begin to play a Tetris game, your decisions are all reactionary (tactical) because you haven't banked anything in your autopilot. When you get a general idea of what is good practice, you store it in memory and it becomes something you don't have to think about so much (strategic). Having a solid grasp of strategy in a game is important, and leads to a lot more safety in NES, and a lot more speed with things like TGM or sprint.
In my opinion though, it is those on the fly decisions that separate good players from great ones. It's similar to someone referring to a move made "over the board" in chess.
In response to your comment on Discord:
"I expect my TGM3 placement to be controversial"
FWIW, I agree 100% with above list, and I'm guessing that's not a surprise to anyone who knows me (as well as my general disinterest in TGM3).
Looking past the obvious potential issues with playing Death mode for score (like you point out, level stop abuse), aiming to play as score-effectively as possible (or rather, going for as many tetrises as possible) in this mode is definitely more interesting to me, than simply playing as fast as possible.
Even as a major TGM fan, I've come to finally realise that maybe I'm not really a fan of seeing Tetris as a speed game, with the speed of which you are able to force each placement being one of the primary metrics used to measure skill. I enjoy the challenge, but more than anything else I enjoy the puzzle.
That's not to say I don't enjoy Tetris as a fast game. The high speed forces you to make decisions faster, and it is definitely a good *method* for testing you puzzle solving skills, even if you aren't using it as a metric at all. To me, the forced speed of Death mode makes Tetris more fun. Meanwhile expecting the player to force as high a speed as possible becomes a game of control moreso than stacking - and one of the things I really *like* about TGM in the first place is how its design effectively tries to make controlling the game as little of an obstacle as possible, allowing player to focus entirely on smart and precise placements.
Or maybe I'm just really biased because I suck at playing fast.
Still, playing fast of course still has a bit of a puzzle to it. If nothing else, a cleaner stack should always allow you to place new pieces faster, and of course the line clear delay in Master mode effectively rewards tetrises (though I wouldn't be opposed to rewarding them even higher compared to triples).
TGM3 has hold and multi previews because the guideline said so.
Hence the crazy speed to make up for it.
Those also allow you to be much more consistent.
Hence the insane consistency requirement to get GM.
These great people sums up why I ditched multiplayer Guideline. Guideline, in my perspective, is a game to control high speed controls, do efficient inputs and learning stuff to boost your attacks. There are too many 'you should do this' and 'you should learn this', instead of 'try to achieve this' and 'bear that in mind'. Guideline is basically learn-to-win to me, which I think is too shallow as a game.
Back to the Death problem by OP. I can't really decide on which is a better run. I think the parameter for the best run, especially at high speeds like Death or Shirase, is not how many Tetrises you clear, nor the score you get, but how well you resolve the problems you made during the stacking. When unfavourable conditions occur, the player will have to cover up some areas or create a less preferable stack. It is impressive not because they stack for Tetris or play real fast - like I can stack for back-to-back Tetrises or play in lightning speed if I wish to. Why these runs are impressive is that they recover from mistakes real fast, and in really creative ways. They get right back into what they are doing, either going for the score or the time. I think this is the thing I feel great to watch and play.
Separate names with a comma.