Showing posts with label followups. Show all posts
Showing posts with label followups. Show all posts

Sunday, 8 April 2018

Openings: the T-block Ace

Continuing in the specific opening discussion, here's the T-block Ace:

  

This is an alternate way to all-clear at piece 10, with lower probability but better fallbacks than the standard Ace. Given the first deck build, we get the piece 10 all-clear two times out of seven, but when we don't get it, we usually get to T-spin with both our first two Ts. Not bad.

I call this the T-block Ace both because when I originally wrote this I'd only seen it played against me by a player named T-block (#1 in Canada at the time) and because it's built around two T tetrominos. Let's look at how it's built.


1. The first deck



Constraints:
L < Z, J < S

We see from the picture that we can do this whenever we can play L before Z and J before S. This is much like the Justice/Elvis and somewhat like the Left-O in its constraints, and significantly rarer a situation than the basic Ace, so this opening does not add much option coverage. But if our opponent plays way-too-many multispins, it may be worthwhile to do in lieu of other openings which don't counter multispins as strongly.

I've faithfully reproduced the first deck as played by T-block. But I should comment that this does not appear to be the fastest variation. For example, we can swap the positions of the I and O:



This allows us to place both the I and the O by hard-pulling to the right edge. For the O we just hard-pull and hard-drop, for the I we hard-pull, hit B (counterclockwise spin), then hard-drop. This naturally places us two spaces away from the wall. Of course it might be easier to just move the I to the right twice and hit A.

Additionally, we could instead play the mirror of this:



This allows us to instant hard-drop the J, since it starts in the correct position, and gives us an easy placement for the S on top; B + hard-drop.

Your mileage may vary, of course. Playing T-block's version has advantages, such as being able to place the second-deck J/L in the O-hole if we're not all-clearing.

Note: since writing the draft of this two months ago, I've seen this opening used against me by Maguro ( まぐろ ), then ranked #3 in the world. He used this variant, matching the logic above, but curiously didn't do an all-clear even with applicable pieces:




2. The second deck: all-clear at 10

      

It turns out we can all-clear at 10 whenever we have both T and O in the first four pieces of the second deck. The above show all-clears with the S, Z, J,and L; every remaining piece except I. After playing the T in-hand there's no ordering constraint, and even if we get an I in our first four pieces, the other piece we get is guaranteed to be one of those 4. The probability of getting both the T and O in the first four pieces of a deck is 4/7 * 3/6 = 2/7.

It should also be noted that given the S or Z, we can do a two-line clear with our first T and still all-clear. This will put a line of damage on their screen at piece 7, which should zone out an opponent's basic Ace opening--or against Puyo, it's a nice four blocks of damage and some animation delay. Having T, O, and S or Z in the first four happens 20% of the time.

For those interested in squeezing every last percentage point out, it's also useful to know these ones, which use J, L, and O:
  

These require holding the T while playing the J and L, and add another 1/35 chance to our all-clear, bringing us to 31.4%.

3. The second deck fallback: T-spins

The general rule here is that if we get an S or Z before our next T, we get an easy two T-spins. If the T happens first, we may have to take just the one. With the S or Z, we could set up a quick two T-spin combo:

  

Of course, doing our first T-spin in row 3 and 4 requires a few pieces before we play the T, with which our second deck might not cooperate. If we just have S,T or Z,T as the first two pieces as our second deck, we need to start by T-spinning in row 1 and 2 before we are forced to play our next T:

  

    

If our second deck starts with a T and we can't all-clear, we'll just take the one T-spin here:




4. Beyond

Beyond the all-clear is much like the Ace, but we know a lot more about the pieces we have remaining in the deck: we'll have used the T and O, and not used the I. Given that we'll always have the I to play, we can usually set up a variant on the Izzy with our four remaining pieces. For example, after Acing with TOS:



In the next deck, we'll take our T-spin. This will leave us with (4+7)*4 - 2*10 = 24 squares of material, which is about right for starting a repeated T-spin pattern. For example:




Beyond the T-spin fallbacks, we can again start repeated T-spin patterns, or do a Tetris Tea pattern in column 7 or 8. Our worst variation, where we get an early second-deck T and just get the one T-spin, will probably see us switching into a Column 7 Tetris Tea pattern at least for a cycle.

      

        

We can also all-clear at 20 in our most common T-spin fallback in various ways:

  

      

I haven't done much experimentation on the frequency with which these can be achieved, but it does lend us some hard-counter potential against openings like the Long Combo. The second, third and fourth look to be pretty darn achievable, with O < J the main constraint if we place the inverted L first.


Analysis

While we all-clear less than the basic Ace, when we don't get the all-clear, we do almost as well as more standard quick T-spin openings. However, unlike an Izzy or WumbOJ, we sometimes end up needing to waste our first T. Our first attack also happens later, but this doesn't matter much: if we see them doing an Ace, we can just do a two-line clear with our T at 7 to stuff it. Since this two-line clear happens in pure hard drops, it should always be fast enough, unlike a move 7 T-spin. And of course it's often followed up with an all-clear of our own.

Speed-wise, we're quite excellent. Our all-clear can always be achieved in pure hard-drops, unlike a basic Ace. In our recommended first-deck variant, our piece placement is optimized to not need any particular finesse--almost everything is either an instant drop, or a drag-and-drop. We don't have any particularly difficult constraints, and don't need any fancy spins to get our setup.

Counterpunch-wise, we're well behind the Izzy and WumbOJ in our T-spin fallback. We generally end up blocking any damage blocks we might wish to clear throughout our first few T-spins, not opening up our board until we do a Tetris or a weaker clear.

Enough talk! Let's grade:

Damage: B
Speed: A
Followups: B+
Counterpunch: C-
All-clear: A-
Anti-all-clear: A+
Flexibility: B+
Explosiveness: B

Now let's look at TvT opening matchups, theory-Tetris-wise. + is "slight advantage", +++ is "major advantage".

Tetrises: ++
Izzy: - The quick T-spin zones out our initial all-clear possibility, and we end up being a worse T-spin opening head to head.
OJ: - See above.
WumbOJ: - See above.
Ace: +++  They should not be able to all-clear us, but we can all-clear them.
Long Combo: ++ Here, speculating that the 20 all-clear is sufficiently achievable to make the occasional hard counter in the T-spin variant as well as seriously denting the Long Combo game-plan with the piece 10 all-clear. If the 20 all-clears are impractical, count this as a mere single +.
Left-O: ++ Multispins can always get hit by our all-clear, and our T-spins are at least as good as theirs. Furthermore, even the fastest Left-O piece ordering doesn't really have a speed advantage versus our fastest variant.
Jail: ++ See above.


Conclusion

This is a very middle-of-the-road opening, soft-countering multispins with its occasional fast all-clear possibility but being slightly weaker head-to-head against other quick T-spin openings. As such, it's worth doing against those who do too many opening multispins--most of the current online ladder--and theoretically as a feeling-out opening against an unknown opponent. However, against someone you have a hard read on, it's likely more viable to use a harder counter. In an optimal metagame sense, middle-of-the-road openings are not required since we can just mix it up with hard counters. Thus I don't feel this opening a necessary addition to one's toolkit and will probably not add it to my usual repertoire at the current moment.

For those confident in their greater speed and skill who just want an opening that won't leave them too far behind and might put them well ahead, though, this opening could be a good choice to lead to consistent wins. If you end up adopting it, let me know how it does for you.

Wednesday, 28 February 2018

Formulaic Midgames: Repeated T-spins

    

Why?

Obviously, I favor a dynamic, improvisational midgame, adapting to one's opponent and the randomness of the garbage we're given. But, speed is the name of the game, and our dynamic improvisation can have building blocks in it we've practiced to be played fast, fast, fast. By knowing formulaic patterns we can use, we can plug them in and run with them when the situation is right.

Additionally, we should note that the attrition-based gameplay we often get into in late midgame is more characteristic of Tetris vs Tetris than Tetris vs Puyo. In Tetris versus Puyo, Tetris is often a player of rapid repeated attacks, while Puyo is a player of large bops. When we're not doing a massive garbage clear as Tetris, it behooves us to be able to continue our attack repeatedly--and a formula may help us do this.

Why not?

In long term attrition, we often don't have the space to play formulaically. Formulae must stand aside for on-the-fly solutions for our board and garbage, lest we die. So, as with many techniques, these have a time and a season.

How?

1. The Repeat Seat

Suppose our board looks like this, and also suppose our decks were just S, T, and the occasional I:



We could then just use S and T over and over like so:

    

Every time we do the pattern, we clear 2 lines for 2*10 = 20 squares. We leave a row behind with 8 squares of material, for 28 squares total. We know a deck is 7 pieces, meaning 7*4 = 28 squares of material. So, this pattern matches up exactly with one clear per deck in real Tetris; we could theoretically keep it going forever.

Once we've built up too big a hole in column 5, we can just use our Is and add on the occasional Tetris:

    

Of course, this loses us two tetrominos / 8 squares from our repeated pattern, but if we have a lot of spare material, or just want to kill our opponent, that's just fine. We could've also played a J at an opportune time to fill up column 5, and gotten back into the repeated ST using an L:

      

Cool, we have a formula that gets us T-spins over and over and mixes in Tetrises. That sounds pretty darn good. But, how would our board end up looking like that? Well, if you've read the previous posts, you might've come across this 3x3 square with a notch lopped off all over the place, in openings, all clears, combos, and the like. Because it fills 8 out of 9 squares of a 3x3, I call it an 8/9:

Top Corner:
      

Bottom Corner:
      

We can create 8/9s, 3x3s with any corner lopped off we want, though in the setup above, it'll either be repeatedly the top-left or the bottom-left. The available piece combinations for your basic top-left are: JO, OL, JZ, and LS. Though, if we're repeating the pattern already, we'll likely not use LS and save our S for the ST part.

We then just have to fill up columns 1-4, and perhaps the occasional column 5 if we want to Tetris. Let's start at the end of a strong opening we know: the Izzy-dubdub after a couple T-spins:

  

We know we have a full deck to work with other than the T. Let's try to transition into a Repeat Seat and get it going. First, we take our T-spin. We can set up the next block in our right-hand tree while we do it:

  

Our first notch is only one off the ground, so we can't take it with an S. Let's use the L instead:



Now, with each deck, we add another block to our tree, with JO, OL, or JZ, take a T-spin with ST, and build up our left-side 4-wide block with the remaining three pieces:

      

In the above example, I found it the Z piece most convenient for my left-side build, so I ended up taking either JO or OL on the right. We adapt our tree-build block based upon whatever pieces are most convenient for the left-side, allowing us to improvise to the situation. And if we have an I we don't know what to do with, we just drop it in column 5 or 7 to set up and take Tetrises.

Awesome. With a bit of improvisation, we can probably just keep doing that in Marathon. In a real match, of course, we'd eventually be called to do garbage clear combos lest we be attrited to death.

Now, you might say, all this improvisation is hard! Can you give me an even more boring formula to do my 4xN column? Well, you are in luck!

Suppose we always use ST for the middle. We'll always use JZ or OL on the right for an 8/9, and always use I + the remaining JZ or OL on the left for another 8/9. If we start in a position where this works, can we just keep it going forever? Let's try. Here, I'm using grey blocks for my 8/9s and alternating their highlights.

    

    

        

    

We'll always have either 4*n, 4*n+1, 4*n+2, or 4*n+3 squares of material, and we have a repeated pattern for each, though 4*n+2 requires breaking down the OL specifically. So all we need to do is improvise until we have a board flat enough for one of the above patterns to apply at deck start, then just repeat it over. And over. And over. Here's an example transition, using bottom-corner 8/9s:

    
    

We see in the sample we have a measure of flexibility--we can use a deck to mostly tree-build, or mostly column-build, if either one is more convenient to our deck ordering and board. And we nevertheless are able to keep to the repeated pattern on both sides.

It's the Repeat Seat because we get our T-spins with the repeated S, T. On the right, we have a tree-shape--we'll call it the three-wide tree. But what about a narrower tree?


2. The two-wide tree

Let's say we start here. You may recognize this as the beginning of the third deck of the WumbOJ: after we play our L and T for a third T-spin on piece 16:

  

We can set up the first couple T-spins of a Repeat-Seat pretty easily with the J and Z:

    

If we wanted to keep the exact same pattern going forever, we could add an O, S, and L. Incidentally, this theoretically could be used to do four back-to-back T-spin triples--fun to do in Marathon or Free Play, though it's unlikely anyone would try this in a real game:

    


But adding an S is lame, since we prefer to use it for the repeated S,T combo. We could just use Z,T in place of S,T. This gives us a T-spin single and leaves two lines on the screen, giving us a Tetris every 2 T-spins instead of every 4. And hey, against Puyo, 4 singles and 2 tetrises theoretically does slightly more damage than 4 doubles and 1 Tetris:

        

If our 5xN isn't fully built, we can substitute for the S with an O plus a piece straddled across the 5xN and S columns:

  

We could, of course, use JL instead of an S for a cycle:

    

As long as we're willing to use J and L here, though, why not use them slightly differently? We can just use S,L on the left, and J,Z again after J,Z on the right, and we set up a 2 T-spin combo. Heck, add a Tetris on this, and we can do a quick 15 damage bop:

        

Excellent. We have the theory of the right side of the screen. Now let's try to fill in the left and right at the same time with actual decks:

      
      

Looks like we can keep this one going too, though it's hard to give a truly boring repeated pattern since we must use different pieces to build the right. Given an already-built tree, there is a repeated pattern for the 5xN with the 5 pieces other than ST. This works for all five of the 5*n + k cases by adjusting the I-column, but is less useful in practice than the repeated 4xN which allows building the tree at the same time:

  

We see in the two-wide tree we get a moment of choice after playing the Z on our tree: we can continue the pattern of a notch every 3 rows with an O, or get a notch 2 rows up with a J, and allow a 2 T-spin combo. But there's another possibility here: what if we instead form the notch on the left, either with the Z or the J on the right?



If we just keep doing that, and keep T-spinning on the right with a Z and a T, we switch into:


3. The Central Tree

      
      

Above is an example of the Central Tree build coming out of the same situation we used for our two-wide tree: a T-spin double on the third deck of the WumbOJ followed by T-spinning one away from the edge. In general, when we do a T-spin one away from the edge, whether on the J or Z on the right or the L or S on the left, we can start this pattern.

I've intentionally used a wide variety of ways of building the tree in the example. If you look at the size of the tree, it's 7-wide. We could divide this into a 5xN plus a 2-wide tree, or a 4xN plus a 3-wide tree, or be more fanciful. Now is a good time to go through more fancy arrangements that be used to build the tree beyond the 2-wide's JZOSL and the 3-wide's repeated 8/9s. Some of these are also applicable to the 3-wide tree, however, and can substitute for two 8/9s:

                  

Of course, once we get going, we likely just want to treat the Central Tree as a 4xN plus a 3-wide tree as we have fast, braindead patterns for repeating this over and over. But all the different possible shapes can help one transition in.


4. Theory of Getting Started

Notice that in all these setups, I branched off an existing opening after a number of T-spins. We don't just open with these formulae, because they require material on the board. Ideally, we should have from one to three notches in our tree ready at any given time, and our 4/5 wide section built up to the proper height by the time we get to the S/T in a given deck, which might be at the beginning. Thinking about that, for our first ST in the three-wide tree, we'd need 4*3 squares for our column and 8*2 squares for our tree (given two layers), making for 28 squares or 7 tetrominos worth of material. Every deck we do a T-spin double with, we play 28 squares of material and spend 20, so we expect to be able to do thee three-wide tree once we've done four T-spins. And indeed, in the the first above transition example, our first ST is after 4 initial T-spins. In the second example we do T-spin single instead of double for our third T-spin, getting us 10 extra squares of material, and start after the third T-spin.

In the two-wide tree example, we got started after only 3 initial T-spins, giving us 24 squares or 6 tetrominos worth of material to work with. This was a little dicey, but we made it work. Once you take a few Tetrises, though, your material is reduced even further. You may need to transition out, or skip an ST cycle to get more material. Theoretically, we need two pieces in our tree to get our notch, for 8 squares, and 5*3=15 squares for our 5x3 initial column. So, 23 squares! We can transition in after our first three T-spins, but should be ready to transition out if we run into issues. How can we transition out?


5. Transitioning out: clearing, comboing, and bopping

We'll likely reach a point where we want to stop doing our repeated T-spin and start clearing our board: perhaps because we're having trouble maintaining the pattern, perhaps to clear garbage underneath, perhaps to send our opponent big, potentially lethal damage, perhaps to defend ourselves, and maybe all of the above. How shall we go about it? The most obvious way, given built up sides, is to haphazardly shove pieces into the hole with reckless abandon and hope for a big combo:

        Sequence: TJILLIJ for 6-combo.

        Sequence: TILSZJTSOLJ for 10-combo.

This works best if you've already put an I into one of your empty columns. But with shorter combos, the damage can be lackluster, especially against Puyo with its lesser combo count bonus. We could just take a T-spin plus Tetris for a high damage short combo:

    

Of course, we've seen above that we can build a double-double on top of our column. This works well in both the central tree and the two-wide tree. It's not as good in the three-wide tree since it requires using either two Ls (for top-corner only), two Zs or Ss (for bottom-corner only), or an extra I, which is already in short supply:

2-wide tree:   

Central tree:    

3-wide tree:     

Bottom-left:        

Against Puyo, we might want to get both our T-spins and the Tetris in a together in a combo due to the way damage scaling works. If we do this, we get at least 14 notches on our green bar for 8+ lines of garbage. We can generally do this by holding the T and getting a deck with T and I next to each other or one piece apart; this happens more than half the time. And when it doesn't happen, well, doing them separate isn't too bad either, and we can sometime interpose a one-line clear to combo anyway.

What about other ways to do multispin + Tetris here? Well, we've already seen the back-to-back triple with the two-wide-tree, which generally works whenever the inverted L is at the top. How about with the three wide tree, or with the central tree?

3-wide tree:     

Top-left:        

Central tree:  

We see that we can do it with the three-wide tree by using an I, in both the bottom-corner and top-corner variants. With the central tree, we still kind of need the I to smoothly fill in the S-columns with stuff. The trip-trip is more often seen as a standalone multispin, but it's nice to keep in mind for the Repeat Seat in case we end up filling in the S-columns.

Against Puyo, we may wish to combo our two T-spins with the Tetris. This is actually easier in probability than the above dubdub: we can do it by holding the T with any deck in which the I is either immediately before or within a few pieces after the T, via clearing the hook with an interposing piece: (Sequence: TTOI)

      

The above example uses an O for simplicity, but you can generally improvise something. I randomly generated a deck pair just now (losjzti,lzjosit) and lo and behold, we can clear it as follows: (Sequence: TTZLI)

      

This clear does 20 lines of damage against Tetris and 16+ damage bar against Puyo for 10+ lines of garbage blocks so is generally lethal in either case unless defended, though the setup is usually not the most practical.


We can also do a multispin via the Left-O:

  

This is often much faster to set up than the above trip-trip multispin, and does barely less damage. We get a T-spin double and a T-spin triple in addition to our Tetris. The same clearing logic applies, allowing us to combo the T-spins with the Tetris via an interposing clear most of the time: (Sequence: TTOI)

  

Randomized deck (tijlozs): (Sequence via piece swapping: TTJLI)

  


Conclusion:

We've seen how to transition into a formulaic attacking game from basic openings, ways to execute our repeated attacks, and ways to transition out to combos, improvisational clearing/counterpunching, and high-damage bops. We have the tools needed both to repeatedly hit Puyo players and to play our Tetris versus Tetris game beyond the opening. In later posts, we can look at the general theory of midgame and dynamic use of these tools in a game of play and counterplay.