Rate my stacking, no srs Options

[myndzi]

Lately there have been quite a few new players asking for help and the like, and after spending some time with someone in 40lines today I got to thinking, should I even be giving advice? I mean, I could be ruining these poor people! (Okay, I'm exaggerating)

But anyway, I thought I'd upload a replay of a hundred lines and ask if I'm any good I always see way more opportunities in a replay than I saw in the game, but that's the nature of the beast. Some of the awkwardness is because I'm making an effort to use finesse and 180, but I'm not really used to either (well, "proper" finesse).

Note: this is not for speed, it's for "live stacking" - immediate placements rather than considered ones, but not rushed.

(upload directory is read-only apparently, so: here)

[Paul676]

2.0: Why not place the T flat? I find placing the first few blocks flat gives more opportunities for the next few blocks. Not doing so means you get the awkward L situation later. Once you had that situation, you should have fixed it immediately with the other L, because you ended up needing the I.

9.0: With a 4-wide 1-deep hole, with S, Z, and L coming, I'd have held the S, laid the Z down, put the S back on it, and the L vertical on top of the S for a flat stack. It uses hold twice (3 times to get an I back in hold), but for the flatness of stack, I'd do it. I know others wouldn't. It's that kinda pattern which I'm able to see and execute at 4pps+. You didn't get in trouble for it, and you might have had a different pattern in your head (I'm slower at seeing opportunities to put the L/J in the kneeling position leading to the O placement, and so won't do it much.)

12.3: With an L and J coming, I'd have laid the Z down flat to avoid the awkwardness which ensues when you don't do so.

18.0: The flat L is so lovely as a border for the right well. Y U NO PLACE IT THERE, with those Ts coming!

22ish: Here's a case where I definitely would have held, and would advise others to as well. An easy L onto J placement for a kind field was wasted because you didn't hold, and you wasted more time working out that field than you gained from not holding.

24ish-35ish: Nice skimming, but I'm sure you can see that with more thoughtful play about which borders to make, and stacking pieces on their flatter ends, you could have avoided it. e.g. the L really should have gone on the right at 24...you'd have had to hold the Z, but it would have led to less awkwardness, and no need to skim, which imo saves more time thinking than hold spends. Anyway no more banging on about hold henceforth.

28: T should have gone on the left to allow the J to kneel on the right, to make a nice border for the I column.

31: I should have made a Tetris, for the T to go on the left for the nice T-->L pattern, where you see a _-_ stack of 3.

44: O should have gone on the right.

45: If you'd have used the J in kneeling position, rather than for skimming, you'd have got an easy use of the S piece upright on it, for a flat stack.

48: Not sure why you didn't place the T on the left, and then the O in the gap where you placed the T.

51.5: If the Z goes upright on the left first time, rather than 2nd, then you might have got out of that situation you got yourself in due to not holding.

56: I didn't notice this at first sight, but the way to get out of that situation is placing the S 2 spaces to the right. Then you could have used the L and J upright for a flat stack.

1:08: The reason you got into a pickle was the flat L placement, where I would have noticed the Z coming, and placed the L kneeling, leading to a flat Z placement, and an easy way out of the struggle you got yourself into.

1:12: I'd have placed the T 2 to the right, the other way round, to give you space to put the S down.

1:23: Where you've got a 3 high 3 wide backwards gamma hole, with the bit on the left flat (imagine the top side and right side of the O piece), you should instantly see that the J goes belly-down, right in time for the L upright on the right.

1:27-8 - good stacking

1:31 - something you might not see instantly, but if you'd have placed the T with its back to the left, point to the right, you could have put the L kneeling down, creating a nice space for the Z to go into it.

1:35 - you probably see it now, but putting the I on the 2nd to right column leaves easy spaces for the J and L to fit into.

1:44 - that was a nice spot to use the LJO. I didn't see it until you did it.

1:52 - anything but that tempting L placement would have been fine. The Zs would have both gone upright to raise the L hole by 2.

1:58 - if L kneels on the left, you've got the 2 Js sorted by putting one belly down for the other to stand, leaning on, as well as a space for the Z. What you did was fine, but it demanded another L which was not coming in the next 6 pieces. It also meant you used the Js not on the left, but on the top of the stack, where they don't solve anything, and just create an O drought which didn't need to happen.

2:01ish: You now probably see that glaring J placement needed desperately to be filled.

2:14ish: You seem to make a habit of placing pieces which demand other pieces, or the same piece. That L placement now demands an L, which is not in your near previews. It's good play to make a play demanding a specific piece which is in your previews, and that's one of the biggest keys to getting a very fast time, but it's equally bad to demand a piece not in your previews, which you just did. At 2:13 you could have avoided that by putting the S not in a skimming position, but with its back in the 4th to right column, to demand that L piece upright in the 2nd to right column, which then would have left you with an easy stack.

2:17 - Z flat on the left solves your problem. It's an excess piece because that part of the stack is already 10 high, but it's better than forcing a soft drop-->zangi. Then you'd have had an easy J standing up on the 2nd to right column, straight for the O and the L leaning on the O.

2:19 - haste to finish the run meant you messed up where you might not have otherwise. I goes 3rd from the right, and then use the 2 Ls upright in backwards gamma position to finish off the run. I think you kinda realised this at 2:20 and then did it anyway, when laying it down in the centre for the Z to go on might have been a better use, especially with the J-->I coming soon. Anyway I'm sure you can work out what you could have done with your pieces at the end to minimise your piece count.

on the basis that I'm fair game, I played a round with your exact pieces myndzi:

I've uploaded my replay here...in the middle I went on a massive rubbish hold spree, as well as a few other errors which I know happened... I wouldn't usually do those things, but I guess that comes with not having played for a while, and not warmed up...but lol here it is in all its awfulness. I held more than I usually would because I think I was thinking about hold at the time, and how it might solve my stack. Something like that anyway. It might help someone, I guess...

[myndzi]

Hmm, I thank you for the detailed inspection and I'll have to go back and watch it to pay more attention. Like I said in the OP, I always see way better stuff watching the replay, so I bet I noticed a number of the same items. I guess what I'm asking is more like, on a scale of awful to Trance, where does my "average" fall? I'm going to go back through and see if I missed any of your comments and can learn something later, but I was just curious where I'm at when you don't think about speed

benmullen: I C U lurking thar! I bet you have a much different take on this sort of thing.

[Paul676]

My take is that it's pretty decent stacking - seeing certain patterns quicker would help things, as would using hold the right amount (not demonstrated by my video...). They kinda go hand in hand, too.

[myndzi]

That helps You can see some KoS influence in that replay actually; if you were wondering "why the heck is he making these crazy overhangs" - that's why. Most of the time I can fill em in with spins, but I don't always know it when I place the piece. Still, it tends to give a flat top which works for twisting OR skimming and makes your field more usable

[caffeine]

Let me preface this by saying that I believe you're a very skilled stacker. You played fluently with no major errors--something only a small portion of Tetris players are able to do consistently. You've been playing a long time, and it shows. It's easy to forget that when Harddrop has so many amazing players on it, but we shouldn't! That said, I want to take this opportunity to dive into a little Tetris theory and fundamentals. I'm sure what I have to say is stuff you've already "instinctualized," but I thought maybe it would be useful to at least other players to see it spelled out on paper. I also thought it would be useful for myself if I try to take what I've "instinctualized," and turn it into words. I absolutely hate picking apart replays, since I know when I play, I do sub-optimal moves all the time. It doesn't mean I didn't understand why something else was better. It only means that actually playing the game and analyzing the game are two different things. Hindsight is 20/20.

That all said, with a little tweaking, I believe it would be possible to go without the need for skimming. We're playing with bag, after all.

Skimming is usually needed either when

• the stack becomes unstable
• the stack becomes too high and we don't get a I-tetromino in time.

When Tetrising in a bag game, a high stack is the result of either previous field instabilities (which requires using precious I tetrominoes to go towards fixing the field instabilities instead of lowering it by making Tetrises) or is the result of simply neglecting to Tetris with the I and instead using it in the stack without needing to use it there.

Therefore, when we have an I and the field is not unstable and we can clear a Tetris, we should use it to clear the Tetris. If we follow this simple rule, all need for skimming will be the result of field instabilities.

But what is field instability anyway? Generally, it is the shift from a field that is amenable to many piece placements to a field that is amenable to very few piece placements.

Consider the following:

• A perfectly flat field will allow for every piece but S and Z without making a gap.
• Similarly, a "castle top" field (where the height staggers from a height of x on column 1, x+1 on column 2, x on column 3, x+1 on column 4 and so on) will allow for every piece but O, J, and L without making a gap.
• A field that is two columns wide and has a height of x on one column and x+2 on the other will allow for only I, J, or L.
• A field that is two columns wide and has a height of x on one column and x+3 on the other will only allow for I.

Given this, we should generally aim to create a field that is a mix of flatness and x+1 raised surfaces. To account for as many placements as possible, we should have as many adjacent columns within 1 or 2 rows of height as possible. For this reason, it's best not to create "towers" or "trenches" (which create x+3 or greater adjacent columns).

Therefore, the most stable stack that allows for constant Tetrising is the one that has a Tetrising gap off to the side (as opposed to somewhere in the middle, which would create two x+4 or greater adjacent columns compared to only a single x+4 adjacent column). The single 9-wide island should then aim to be somewhere between totally flat to totally castle top (while avoiding both extremes), while making an effort to keep adjacent columns from exceeding an x+1 heights. (Perhaps a slightly more accurate way of saying this is "a flat surface with only enough bumpy surfaces to comfortably fit both an S and Z.) This is basically what I look for in clean stacking.

But enough theory, I'll show you what I mean.

This is what you do with the first bag:


The L placement on the right side only allows for an I or another L. I think we can do better. For example:

This is stabler, since we have now freed up the next L or I to use for any placement we please, opposed to being forced to use it on the right side.

Here we go again:

Let's maybe do this instead:


Here we have the same problem, but it's slightly worse. Instead of having a single x+2 adjacent column, we now have an x+2 and an x+3 adjacent column. This is because the 2-deep gap was made in the middle of the island as opposed to on one of the sides.



Here's something we could've done instead:


All placements being equal, it's better to create a raised column on either of the very sides of the island (as seen above with the first L). By doing this, we're "filling in" what we know will work from the outside working inward. This is good simply because if we create a raised surface in the middle, we might end up dividing our preferred single island into two separate islands by way of a middle tower. This is the same as saying we'd rather have a single x+1 (or greater) adjacent column opposed to having two x+1 (or greater) adjacent columns. (I hope I haven't made this too confusing!)

Note: it is for this reason I believe "pyramid stacking," as advocated in the TGM series (and necessitated by ARS and 20G), is actually contradictory to good fundamental stacking habits.

Here, you've rendered a field instability (dependence on an I-tetromino), and must skim in order to re-stabilize.


Instead of creating an x+3 adjacent column, let's instead raise a side column (as we've just discussed in the above example):


This sequence of moves lead to an awkward situation due to a trench in the middle of the island.


While I don't think I'd play this any better than you did in an actual game, I do see this more stable option when looking at it now:


I could go on with more examples, but that's the basic idea. I hope you find this interesting.

[Paul676]

Caffeine your post is beautiful. It's exactly what I was trying to say, but you used fumen and addressed all types of bad stack, and it is wonderful. Therefore I'm putting this thread on the Most Important Threads List, because of your post.

2 things:

Major: An endnote to caffeine's post: It is fine to create an x+2 column, so long as we can see an L or J to fix it within the next couple of pieces, because it no longer counts as an x+2 column, but in fact a 2 or 3 piece super-stack ending in an x+1 or x+0 column. I think this is one of the biggest things of playing fast: finding 2 or 3 piece super-stacks as described above in a game, so that you're thinking about higher numbers of pieces, rather than 1 piece at a time.

Minor: I think we need a 3rd point where skimming becomes necessary - as a big sub-part of "unstable stack", we should put "Hole becomes 2 wide" and "Hole becomes 3 wide", for example in places where myndzi put the J kneeling down. I know it's part of unstable stack, but I'd say it's a big enough sub-part to warrant putting in, especially as that point is at the top, and we don't realise that it comes under unstable stack.

[Barneey]

Hey, I cant download the replay right now, I'll watch it later.
Quick question though, is this run 25 Tetrises? And is it no-hold?
If the answer to both of these is yes, then I think it would be hard to argue that your stacking is poor.
(I think it would be easier to judge this from a multiplayer game, where you're encouraged to downstack efficiently, keep up APM through B2B, T-Spins, etc. ontop of stacking neatly)

[caffeine]

Thanks Paul.

Major: An endnote to caffeine's post: It is fine to create an x+2 column, so long as we can see an L or J to fix it within the next couple of pieces, because it no longer counts as an x+2 column, but in fact a 2 or 3 piece super-stack ending in an x+1 or x+0 column.

Ah yeah. Good point.

Minor: I think we need a 3rd point where skimming becomes necessary - as a big sub-part of "unstable stack", we should put "Hole becomes 2 wide" and "Hole becomes 3 wide", for example in places where myndzi put the J kneeling down. I know it's part of unstable stack, but I'd say it's a big enough sub-part to warrant putting in, especially as that point is at the top, and we don't realise that it comes under unstable stack.

Hmm, I don't have the replay in front of me right now, but when I get a chance later I'll look at this. I'm having trouble understanding what you mean by "hole becomes 2 wide."

[Paul676]

hole becomes 2 wide, so you need to skim.