Randomizers

[Kitaru]

Technically it wasn't featured in TOJ at all. In TOJ's 8-bag each bag consisted of a 7-bag plus one random extra piece, which was then dealt fully. By contrast, in the proposed 1+7-bag only 7 pieces from the bag are dealt before replenishment, which provides a correlation of sorts not seen in 8-bag.

Ah I see, it's a difference in terms with the non-standard Bag sizes. I guess it depends on whether you qualify it as 7+x Bag (a full bag and then x pieces from a qualified Bag but the rest discarded, which is just a random piece for x=1) or a separate Bag from which x pieces are interleaved without discarding the remainder.

[Okey_Dokey]

what RNG is used here?

note: ^ this is not the same old n-blox

It's still the same randomizer. All they changed is the spawning orientations and sounds (and maybe some other small things). I also wonder how the randomizer works exactly. It's a well-designed randomizer, but all in all it's not difficult enough for Marathon.

It seems the randomizer thinks in 7 pieces like bag randomizer. However, a "bag" won't always contain all 7 different piece shapes. Sometimes, it's only 5 different shapes. It also seems, that the randomizer counts how many pieces of a shape has been dealt in total, and will deal the lesser ones more likely.

The first 7 pieces contain at least 6 different piece shapes. The next 7 pieces contain again t least 6 shapes again, whereas the missing shape (if there was one in the first 7 pieces) will be dealt, too. Afterwards it gets more complicated. A shape will show up after around 30 pieces at latest. You can get a shape 4 times in a row.

Flood and drought frequency is a recurring topic in randomizer discussions. Why is it so important we avoid them? Back in the day, getting a long run of Z's or not getting an I piece was frustrating, but also exciting. It was a natural challenge to the game.

Can only speak for droughts: If an I piece doesn't show up for a very, very long time or is distributed too unevenly, then it will ruin every record attempt. It's frustrating, if you are near the end of a Marathon or 5-minutes-Ultra game, and then you fail through no fault of your own.

I think floods shouldn't be reduced too much. Floods are one of the few situations where you can't stack that flat and must use techniques like platforming and skimming to stabilize the stack.

[zaphod77]

The reason drought and flood avoidance is so important in TGM games is because they are arcade games, and it sucks to lose because you got a horrible sequence that you were helpless again.  and it suck seven more to be denied the tetrises you need because the randomizer starved you of your I pieces too long.

Since every game is it's own quarter, the randomizer has to not screw the player over too badly, to ensure the GM grade is available every game.

A game that isn't based on tetrises to the extent that the TGM series is may not need that stringent protection.

[caffeine]

But kbr is posing this question to an English-speaking Tetris forum. The quarter argument doesn't really work here. How many of us can honestly say that we regularly are dropping  ¥100 into TGM?

@Okey_Dokey, NES Tetris can give droughts that can ruin a max out run. I wonder if NES Tetris enthusiasts, when confronted with the possiblity to change NES Tetris so that it never gives droughts, would say that it would be an improvement.

[zaphod77]

well it wasn't designed for us emulation using people, was it? it was designed as an arcade game, and thus certain decisions were made.

[Aluce]

Aren't there various ways to solve the bag problem such that the techniques that are super bag-dependent (which are exactly those that arguably kill part of the game) no longer work, but have no large impact on other strategies?

For example, you could make a 5-bag system: 5 random different pieces per bag, such that the third bag contains the two pieces that weren't dealt in the first bag (and the fourth bag has the two pieces not dealt in the second bag, etc.). After two bags, three out of five pieces per bag will be randomly selected out of five remaining options, meaning that to some degree it's quite similar to normal bag, but the distributions are different. There are no big floods ([...SZ][SZ...][SZ...] being a worst case scenario) and the droughts ([I....][.....][.....][....I], maximally 18 pieces in between) aren't too huge, but can still form a challenge at times.

I think the 7+1-bag proposed earlier would have a too small effect and that some 7-bag-dependent strategies could still work with some adaptation.

[pwn_by_numbers]

Flood and drought frequency is a recurring topic in randomizer discussions. Why is it so important we avoid them? Back in the day, getting a long run of Z's or not getting an I piece was frustrating, but also exciting. It was a natural challenge to the game.

A lot of people are talking about single player, but in multiplayer a drought will just straight up make you lose. If you get stuck in a bad position and can't downstack a hole while receiving pro-level APM you will lose in seconds. This means if you want to not lose, you have to play incredibly clean and stay safe will doing riskless t-spins and tetrises (read: boring).

This isn't just the case in real-time games. Anyone who has played Memoryless on KOS knows that a flood of Os or Ss or Zs can ruin a game. Playing with 4-History randomizer has always felt a lot more "fair" to me in terms of luck. (In KOS players don't get the same piece sequence.)

There are a couple of ways to balance this though IMO, the best being rewarding all-spins. Allowing more pieces to attack balances out the unlucky-death-causing ability of droughts. Overall though games would still probably go quicker, since all-spins allows you to send a lot more garbage, and a worse randomizer gives you less defensive options. Another option is getting rid of normal change on attack garbage and sending either entirely messy garbage, or doing something weird and unusual like just raising the field without any garbage (like Air mode in KOS) or having things like tspins appear in the garbage to make defense easier.

Personally, I think t-spin only + 7bag is the most balanced for multiplayer. A decently balanced history randomizer (something like 3 or 4-hist) will lead to similar game balance but with less options for static setups (which is not a good thing IMO). I think the only way to make a memoryless or similarly harsh randomizer fun is to do away with spins entirely and have an arguably simpler game, like Cultris (combos only) or like Classic mode on KOS (only rewards Tetrises, no spins and no combos, no garbage cancelling). These games make up for defense being hard by making attacking harder too.

As far as improving the most popular ruleset (eg. the on used on TF), I think a lot more balance could be achieved by adjusting the combo table than probably any other thing. You could also probably achieve more by limiting piece previews than by changing the randomizer.

[caffeine]

Yeah, of course there will be problems when you have memoryless and then give the players different piece sequences. This is fixed by giving both players the same sequences.

[Kitaru]

But kbr is posing this question to an English-speaking Tetris forum. The quarter argument doesn't really work here. How many of us can honestly say that we regularly are dropping  ¥100 into TGM?

I think it's still something worth considering in the design — players may not pay with money pet attempt, but with their time. Variance is OK, but success should be available in most if not all attempts.

The effect of variance is also accentuated of there isn't a maximum score or goal to achieve. Also, it's important to consider the space within which variance is spread. It's may be more devastating to be droughted at the end of a 40 Line Sprint or 25 Line NES B-Type, whereas it may pose an acceptable obstacle in a longer term NES A-Type where a missing out on an I-Piece here or there doesn't necessarily destroy an attempt to reach the score cap if the situation is handled well.

[bigwig]

color thief has done a lot of research into different randomizers (he's the one that made most of the charts people link to). I PMed him and here are his thoughts

"I think if randomizers are too random, then it's possible to waste the player's time by killing him despite no bad decisions on his part. This is a mathematically proven fact (with that SZ paper). Even if you don't outright kill a player, the same effect can ruin the record setting potential of a run which still wastes the player's time. Ain't nobody got time for that.

At the same time, if a randomizer is too predictable, then there is less variety in decision making and gameplay becomes more bland. It loses improvisation which is too bad. Playing forever strategies illustrate this, though even if those are not possible the effect still exists.

The goal in my opinion is to hit somewhere in the middle. Enough to consistently challenge the player with uncomfortable situations and force them to deal with it. But not severe enough to be beyond the tools they are given to solve them.

7 bag is pretty garbage in my opinion, it's significantly easier than anything else used in a commercial Tetris game. The problem is compounded with move reset, numerous previews, hold, and SRS. Solutions to problems are overly plentiful and you don't have to get as creative as I would like.

TGM is pretty awesome. There is a deep skill curve in creative problem solving. But I'm not without criticisms there too. TGM3 introduced an improved randomizer, but it would have been more appropriate in TGM2. The extra previews and hold in TGM3 change things significantly, and as a result problem solving is blander than TGM2. With these tools, it would be more appropriate to have made the randomizer more random not less random.

Guideline should change randomizer. But if that's the only thing that they are changing, and they want to keep it appropriate for single and multiplayer, then I don't really know what to recommend. Anything more random will be an improvement, but more needs to change. There are limits to how much a randomizer can improve things.
"

[zaphod77]

Variance is OK, but success should be available in most if not all attempts.

This is the reason TGM has it's current randomizer.

Memoryless is really only an appropriate randomizer when the main goal is survival, such as in old school tetris games, where you can help mitigate the randomizer's effects by stacking for survival. The original tetris games (pre gb/NES) actually scored no points for clearing lines. they scored points for placing pieces, so stacking for survival was not harmful to your score. Nintendo introduced the concept of a huge score bonus for the tetris, as far as i know.

But with TGM, if you try to stack defensively, you will NEVER get that GM grade.  so there needs to be a kind enough randomizer that you CAN stack for tetrises the entire game.

it has also been determined that hold and infinite previews will not save you from memoryless killing you off with an impossible to deal with sequence.

[Sisu]

NES Tetris can give droughts that can ruin a max out run. I wonder if NES Tetris enthusiasts, when confronted with the possiblity to change NES Tetris so that it never gives droughts, would say that it would be an improvement.

FWIW: http://www.reddit.com/r/Tetris/comments/1n...c_tetris_world/

I'm a firm believer in the idea that a good player should be able to withstand any sequence.
...
I like to close gaps with Tetrises. There's something very anti-climactic about chasing someone down with lines alone. I might've been a little too aggressive but it just means I need to practice more.

[Rosti_LFC]

But with TGM, if you try to stack defensively, you will NEVER get that GM grade.  so there needs to be a kind enough randomizer that you CAN stack for tetrises the entire game.

I'm very sure from TAS work that there are TAP randomiser seeds that will produce a string of pieces where you literally can't stack purely for Tetrises the entire game. Even using every I piece to clear a Tetris there can still be drought periods where you don't get them often enough to clear the field.

[zaphod77]

skimming doesn't mean you stop stacking for tetrises.

and while you may get a drought and have to skim some, i don't think there are any actual unwinnable seeds.  You may not be able to clear the field with only tetrises, but you should always get enough I pieces to get the needed grade points.

If you don't keep the stack high enough to use most of your I pieces on tetrises, you won't get the GM grade.

With crueler randomizers you may get the I piece before you need it, and it's nearly always the right action to place it in the tetris hole anyway, because elsewhere will destabilize the stack too much, and you might not get another. getting too greedy for the tetris in the older pre gameboy games will kill you, but it's actually required in newer games to do well.

remember. most early games use gravity lock (a piece that would overlap after applying gravity instead locks) and memoryless randomizers can throw truly evil stuff at you, requiring you to stack high and dig you way back down.  If you were waiting on an I piece, you may be topped out when you would have survived with a lower stack at the start of the nasty sequence.

[dehinrsu]

Why am I not able to see my previous FASTREPLY????