Silly idea for a new rotation system.

This will be an intelligent rotation system that tries to "do the right thing" in any circumstance.


My idea. Before wallkicking, try to use alternate rotation algorithms to place the piece.


I'll give an example with the T piece.


In most rotation systems, it will go

Code:

.T. .T. ... .T.
TTT TT. TTT .TT
... .T. .T. .T.

But in ARS, it will drop down to the bottom row instead of float in the first position.


And atari rotation has a different result, which has the piece pull to the left wall of the bounding box.


So, if basic rotation fails when rotating to pointin up, ti will first attempt to kick down, to mimick ARS. and when rotating to pointing right, first try to kick left, to mimick the atari rotation system.


The algorithm is to try all basic rotations in all rotation systems before starting to try and wall kick.


Then, try to wall kick in all possible positions allowed by the rotation systems that exist, and pick the one that leaves the least holes after line clear, breaking ties by preferring the least squares shifted,and beyond that, prefer down and right to left and up.


Maybe it's a silly goal, but my idea is to letpeople play with this version, no matter what rotation system they are used to, and have the pieces do whatyou'd expect them to do, and when it doesn't it should be a pleasant suprize..

 

Wow, such a system seems so complex I can't even begin to imagine the limitations and exploits. Good luck to whoever tries to sort it out.

 

I can already see a major limitation here: What if multiple rotation algorithms would give a valid and different no-kick position?


Also, for the least holes after line clear, what about maneuvers where you rotate-kick then shift left or right?


Besides that, I could see how to program this, although it'd be far from actually achieving its purpose, and the code would be REALLY clunky. I've written a 9-page algorithm for a Battleship game AI and yet I'm still going to refuse to touch this.

 

In that case you would have to give priority to some rotation system - for example try ARS first and if that fails you try SRS. So the pieces would rotate like in ARS when there is nothing else close by but then switch behaviour all of a sudden if there are obstacles.


It would be really difficult learning to use such a rotation system.


Also the "intelligent" picking of wall kicks kind of defeats the purpose of the player having to place the pieces in a skillful manner when all of a sudden the game places them for you.

 

yeah, i like knowing exactly what's going to happen when i press button x. it's like my phone that tries to be smart when text messaging. i want it to type either "me" or "if" when i press 6 then 3. instead, it picks the one i most frequently use, which varies, and it winds up adding an extra step because i must first learn which it picked, then continue typing. i'd rather just know if i need to scroll or not, and get on with it.

 

sprint LG. the actual technology is called "T9Word." it's good, but i'd rather it always pick me first or always pick if first rather than sometimes/sometimes.

 

I said it was probably going to be a silly idea.


But yeah, the basic rotation systems will have a priority order.


ARS and SRS would be first and second, though it's arguable which one should be first.


And most possible positions are covered by those two.


My thought was to prioritize SRS, then ARS, the Atari rotation system, then the few positions from other obscure ones.


As for the wallkicks, the answer is simple. Try wallkick 1 from rotation system 1, then try wallkick 1 from rotation system 2, etc.


Thinking back on it, i've got a new idea.


Pick a default rotation system, but test all possible rotation systems. When a rotation would fail in any system, eliminate the system from the list to try, until you are down to one rotation system.

 

I think your idea isn't different enough from simply having a huge number of kicks. The only thing that's different is that the order of the kicks changes depending on the context in a somewhat arbitrary manner. If you're going to make a system that "reads the player's mind" I recommend instead taking ARS and then adding special exceptions where you want them.


In other words, human-tuning the kicks in certain situations. Any completely automated system is not going to work. In ARS rotation direction matters when you do a regular t-spin. Use the wrong way and you'll kick out of the hole. Also, with ARS you often want to use this very same kick that fails the t-spin for finesse when placing a piece. No rotation system would pick the best kick in both situations without taking into account screen geometry.

 

A situation where the player presses "rotate left" but the game rotates right because it is better is akin to have the game just play itself, without player interaction.

 

it will always rotate the direction pressed. It would not rotate the other way because that direction is more useful. The player requested a left rotate, a left rotate is what the player will get. Either that, or a rotation failure. The fact that SRS and ARS often reverse the directions of the buttons when included in the same game is an issue. But in this case, rotate left and rotate right will not change functions.


But sometimes the location after kick is closer to the spot it would be without one if you rotated the other way. We see this in SRS all the time with the I piece. Look in the i spin section of the wiki. (http://www.tetrisconcept.com/wiki/index.php?title=I-spins_in_SRS)


Check out the first example. The final position is closer to rotate left than it is to rotate right! Effectively, the game rotated left when you told it to go right!


And the t in the hole case is an odd one.


In SRS, it's not a kick, but resets lock delay anyway.


In ARS, assuming there's no room for kick left or kick right, it IS a floorkick, and resets lock delay, but lock delay reset is allowed only once. Of course if there IS room for "kick keft" or "kick right" that means the 3 corner rule fails and there is no t spin, and the player would just slide it in with a zangi-move anyway.


rotating the other way is identical in SRS and ARS.


Unless i'm missing something, what you describe with kicks during a normal T spin is a non issue.


That said, there IS a situation where srs and ars will rotate diffrently.


that is here.

Code:

XT
TTT
X X

now if you rotate left, in srs, it will stay in the hole. In ARS it will rise up (compared to SRS) and then kick to the right, out of the hole. instead you must rotate right 3 times quickly. resulting in the piece rising up one square, falling back down, and then two rotates identical to srs and the T spin bonus.

However, this case is rare, because an ARS player is more likely to do this.

Code:

XT
 TT
XTX

and rotate right once or twice, which works perfectly in SRS as well.

Are you telling me that an ARS player will slide a piece in, then rotate it to get it out again and slide it back the other way, when they culd have simply avoided sliding it in in the first place?

Or are you talking about this instead?

Code:

XT XXX
 TT X
XT X
XX X

In SRS, you rotate either way twice to get it in.


In ars, this requires you to rotate right to stay in the hole on the left, because if you rotate left, it will kick right and drop down one. You them tap right and it drops into the big gap. The kick saves a tap, if that's where you were trying to get it. In SRS, to do the same thing you must tap away from the wall first. At first glance, that move seems potentially useful.


But as an ARS player, i'd slip a J and an S piece in there, in that order. Or a L and a Z, again in that order. Then there would be a t spin setup that behaves identically in SRS and ARS afterwards. And the same bit is possible under SRS as well i think.


Ah well. The idea is still not very good, aminly because ARS and SRS are quite incompatible, with their reversal of locking rules. In my opinion, ARS has it right.

 

Assuming there is space for it, then yes, initial rotation would be handled ARS style. but as I said, it will never work. The two systems behave very differently when pressing up versus pressing down on the stick, sad to say.


The intent here iwas for it to do what an ARS player expects it to when playing, but if someone is used to SRS, it will do what the SRS player is used to. But the incompatible drop methods kinda ruin that idea.