Lowest KPT challenge

[caffeine]

Caffeine: I count "drop" as any set of moves (or lack thereof) which lock the piece in the stack.

myndzi: Say KPT appeared in a release in the future of a multiplayer game. Would you still say the same thing?

OK, in that case, I would say that to try to play in a way that minimizes KPT at the expense of playing effectively is akin to ignoring the downstack with the exclusive goal of maximizing TPM. Just as how a player might look at APM instead of TPM for more insightful feedback as to how he performed, he might also look to another metric aside from raw KPT (i.e. "KPT for games that I didn't do the soft drop trick"). Otherwise, he'd only be cheating himself, wouldn't he?

[XaeL]

Soft -> hard drop is avoidable. If you use it to slide or twist, it's a justified penalty. The only "unfair" thing here is completely optional.

soft drop --> hard drop = 2 keys, but the wrong (slower move)
hard drop = 1 key, but the correct (faster) move

you can turn on 20g, and ARE to avoid ever pressing harddrop/softdrop. this will reduce kpt.

[Paul676]

Not under the current system

Caff: When you said I had a point, what did you want me to mean?

[caffeine]

Not under the current system

Is this in response to my comment, "Otherwise, he'd only be cheating himself, wouldn't he?" If so, then does that mean you are referring to how this playing style could cheat other players on this thread? If so, then does that mean that what is meant by "the current system" is the set of rules belzebub laid out for this challenge? If so, then that's why I purposed an "infinite lock delay" rule in my previous post. It turns out that this is possible by setting lock delay to 99, which prevents the "soft drop cheat." (Sorry if this my wording was redundant here, but I wanted to make sure I was addressing what you were talking about, since for some reason I'm finding your posts to be very cryptic tonight.)

Caff: When you said I had a point, what did you want me to mean?

I was wondering if we might come up with alternate, more meaningful metric for "how efficiently the player manipulated the pieces." For example, in the past I've used mKPT ("meaningful KPT"), which, was used to compare apples to apples in my thread about what style of play uses the least keys in 40 lines. mKPT = KPT, but keys pressed on the same frame count as one key. So, it may be possible to come up with an even more representative metric for the core principle we're trying to get feedback for.

[myndzi]

you can turn on 20g, and ARE to avoid ever pressing harddrop/softdrop. this will reduce kpt.

It's a bit more complex, but this can be accounted for programmatically (min 1 point per piece), or simply don't care. 20g is its own skill distinct from other skills, and 20g 40lines if it became "a thing" would only be comparable to itself, so we would still be comparing apples to apples.

caffeine: your point holds true, but rather than trying to devise a different metric I say take the metrics that exist for what they are. The measures should be simple, your interpretation of them can be complex. I don't mind the concept of one single "overall performance" stat if you could ever arrive at such a thing, but if you make everything with that in mind it will just be hard to gain meaningful information.

Edit: Another alternative method could be to ignore drop keys entirely. Since a piece has to lock every time one spawns, you can't play "more efficiently" by not dropping a piece, after all. This turns out to be approximately equivalent to the other method, it just moves the limit to approach 0 instead of 1.

[caffeine]

but rather than trying to devise a different metric I say take the metrics that exist for what they are.

But aren't you contradicting yourself? If I'm understanding you correctly, you're wanting to change KPT from it's most basic formula into a new and different metric: xKPT (KPT with +1 for all locks). In other words, from my point of view, you're the one trying to devise a different metric.

[myndzi]

But aren't you contradicting yourself? If I'm understanding you correctly, you're wanting to change KPT from it's most basic formula into a new and different metric: xKPT (KPT with +1 for all locks). In other words, from my point of view, you're the one trying to devise a different metric.

This entire discussion is about the vagaries of implementation of a simple concept. I am only suggesting the most effective implementation I can think of. It counts one for a soft drop, one for a hard drop, none for a gravity lock. I don't really care, I'm only discussing ideas here. Of course they're not consistent, I'm not making a single cohesive argument, just brainstorming along the way.

[Paul676]

Is this in response to my comment, "Otherwise, he'd only be cheating himself, wouldn't he?"

Nope, in response to Xael's comment - it's not correct under the current system for keypresses used by belzebub...which is why our discussion is legitimate.

My view is that a gravity lock is just as much of a key-press as a normal lock...except that the key isn't pressed - but what matters is the time taken is equivalent to more than a key-press, assuming normal lock delay. Hence why I'd have all of them as 1 "press", or all as 0 presses. Probably the latter.

[Paradox]

[!--quoteo--][div class=\\\'quotetop\\\']QUOTE[/div][div class=\\\'quotemain\\\'][!--quotec--]that's fair, but it also penalises someone who uses soft drop then hard drop, as opposed to someone who uses soft drop to lock delay[/quote]

in a challenge like this where we get lowest kpt sure, but its not like you will strictly follow low kpt when playing multiplayer or something. There is no penalty for having a slightly higher kpt because you used soft drop. It simply means the number is higher. it should be keys per tetromino (lol duh).

[XaeL]

try to complete a 40L sprint with as goal the lowest KPT for a fixed piece sequence, following these rules:

- lowest KPT (keys per tetrimino) wins
- maximum 104 tetrimino's used
- maximum time 1:30
- free to stack how you want
- fixed piece sequence is lapsilap WR bags

how to play
[spoiler]use nullpo blah blah blah.[/spoiler]
Winner should provide screenshot or replay file as proof. Any insights in stacking with low kpt are highly appreciated!

is "how to play" part of "rules"?

KPT 1.00
180 rotate - Yes.

also with respect to the whole softdrop/harddrop thing:

KPT = keys per tetromino. That means Total Keys / tetrominos used.
Total keys = keys pressed. If you press softdrop 10x for one piece, you are still pressing 10 keys. its that simple. i dont see why there should be any special cases.

[belzebub]

I was thinking of good definitions, what I'm currently thinking of (brainstorming, sorry if inconsistent):

• raw KPT could be keys per tetrimino placed in the field, so sum(all keys pressed)/tetriminos locked (the tetrimino in hold and a tetrimino that is falling down doesn't count then) - this metric will then be kind of 'useless' to compare unless other parameters are kept into account
• we need something to compare our skills with the pro's, some kind of efficiency metric, thinking of two now:
  • some metric that, if keypress speed is constant and stacking the same, would give as a result the player/game with the lowest value wins, i think this is what mKPT is a bit like. The lowest value will then be the most efficient inputs per (key)frame. Could be like the number of keyframes used, a keyframe would be every frame where some new input with effect occurs.
  • some metric that expresses the efficiency of tetrimino manipulations, like the amount of 'manipulations' needed per tetrimino, a manipulation would than be every move, rotation, hold or soft drop (not hard drop, so if it's locked by hard drop or gravity is no difference). I will call it MPT, manipulations per tetrimino, for reference.

Some considerations,
- mKPT will be difficult to implement for me
- raw KPT will be only roughly an indication of efficiency
- MPT will be easier to implement
- MPT will give every manipulation equal weight (move as heavy as rotate or hold, no matter what)
- other variations of MPT can be thought of, like weighted MPT or MPT with more manipulation definitions (like rotate,move,hold,rotate+move,etc)

I personally think now that MPT will give some good (but not perfect) indication on how efficient your finesse scheme is.

[caffeine]

some metric that, if keypress speed is constant and stacking the same, would give as a result the player/game with the lowest value wins

I agree, and this is exactly what raw KPT does (it's especially nice because you then have this nice relationship of TPM = KPM/KPT (which is something you lose if you use mKPT)). The real problem is that we can use in-game features to exploit this metric. The KPT is still accurate, but it is no longer a representation of the type of game we're interested in. I believe the type of game we're interested in is the type of game Lapsilap played.

So then, the features that change the type of game we're interested in are:

• Letting the piece drop on its own to bypass the harddrop key.
• Slowing down AR so we no longer need to tap and can ride DAS preservation for several pieces.
• Exploit the hold bug.
• 20G in order to take advantage of mobility factors and/or floorkicks (this one is debatable).

If it weren't for those, the records in this thread would be completely comparable to Lapsilap's game. In order to achieve this comparability, we can do one of two things: either change the rules of this thread or create an alternate KPT metric that does virtually the same thing.

If we change the rules, you'd say: "You must hard drop every piece. Your AR must be 0. Your gravity must be 0." And if you fix the hold bug to boot, this would get the desired result of what we're looking for in a efficiently played 40L game that is comparable to Lapsilap's.

If we create a new metric, "xKPT," then it could look something like this:

• All keys pressed count except for locks done as a result of letting the piece lock on its own, as a result of soft drop, or as a result of hard drop.
• All locks count as +1.
• Fixed the hold bug.
• AR counts as a key press, as a function of DAS. So, holding down the left button results in 1 key press. If DAS = 10 and AR = 2, then each AR event counts as 2/10ths of a key press. This is could be represented as (Total discrete AR events)*(AR/(Min(DAS; 12))) = total "AR-adjusted" keys.

So then, if the thread allows for non-0G play, the 20G players will yield the most efficient xKPT. Elsewise, it would be totally comparable to Lapsilap's game.

[belzebub]

I have to leave quickly, but I want to say one thing already, if we set rules, the best is if we are in a situation where we can look at the resulting screenshot and see that the rules are followed, kind of validation by screenshot would be cool

[Paradox]

wow xael you are terrible at 1kpt hahahah

[XaeL]

wow xael you are terrible at 1kpt hahahah

i havent played it for a long time. i played like 10 games and i coudlnt sub 60. I used to be around 40-45 avg.