Lets talk about NPS and see if we can create a better in game Difficulty calculator

[Jousway]

so I was bored and messing around a bit so I added an NPS calculator to sm5 default theme



for anyone that wants the code and stuff I got it on my git
https://github.com/Jousway/Stepmania...%20overlay.lua

anyway I wanted to talk about NPS because its not really fail proof and its rounded off quite big, so if you got a file that's like 8.59 nps and 8.99 nps they both get shown as 9, but in sm5 we can display numbers as big as we want, so we can use bigger values.

back to not being fail proof, with NPS its a global value of the notes / chart length so if you got a really hard part and then a really easy part its not really a good way in my eyes to get a proper difficulty value.

also it doesnt support xmod files, I have already made an amod/average mod system so I tough it would be cool to have a system that supports xmod.

so yeah, does anyone have any ideas for a system, it can be advanced like fuck, because sm will calculate it for you aside from the way you have to calculate NPS manually atm

[Dynam0]

I was considering SM calculating localized density first in terms of specific areas of the file in order to eliminate the problem Jousway described (having the song be 12 nps for 90% and then 36 nps at the end or something like in trollstream).

Another factor is pattern choice. The obvious answer would be to attempt to calculate density specific to one column. This way jack-type patterns would carry much more weight into the difficulty equation.

However, going off of density alone isn't enough since you can make a file harder just by the type of rhythmic variation (poly rhythms and odd syncopation really screw this up).



To get it absolutely correct, you'd have to make an index of some preset patterns and give them number values based on how difficult they are. These values will be functions of bpm as well since a running man at 300 bpm is much harder than one at 250bpm for instance. Big question here is can SM actually detect specific patterns that are 5+ notes long? Even granted that you could do this, transitions and stamina matters can't even be taken into account. In addition to the fact that there are thousands upon thousands of poly-rhythmic patterns that the program would have a bitch of a time calculating, for example if you have a colour-note guitar solo or something. It would have to quantize the distances and 'approximate' to a pattern in the index that best matches what was analyzed


So many factors in this. It's more beneficial to simply have a few people play-test the file and rate it manually.

[kaiten123]

I think the coolest possible thing (though probably annoying to make) would be a graph of the note density vs time that includes separate note densities for each column. you could probably add color-coded info about polys and such too.

even if this is impractical to implement I'd like to get some difficulty that isnt just a single number (at the very least, a series of numbers that represent different aspects of difficulty. kinda like groove radar but better)

[DossarLX ODI]

Originally Posted by kaiten123

I think the coolest possible thing (though probably annoying to make) would be a graph of the note density vs time that includes separate note densities for each column. you could probably add color-coded info about polys and such too.

even if this is impractical to implement I'd like to get some difficulty that isnt just a single number (at the very least, a series of numbers that represent different aspects of difficulty. kinda like groove radar but better)

I was thinking for note density the groove radar could be modified to calculate density vs. time. So something for jack density (one column), overall density, etc. rather than the groove radar's stream and voltage.

[NeoMasterPie]

jousway reppin' my files (o:

[tunsz555]

Maybe you could run a simulation algorithm to simulate 'hard pain-ness' which you have a stamina starting at 100 for each finger. Each time you press an arrow stamina will reduce (If solo file the ring finger would reduce more stamina per arrow) and overtime stamina can increase. So when jacks came the stamina will greatly decrease and have no time to recover. By the end of the song you collect lowest stamina on each fingers and if it's kind of low you add up some numbers to calculated NPS.

[Kraezymann]

Is there a system where its NPS AND the older system (I'm talking about the system I saw in the FFRCP4)? What if you made it a combination of both? A combination of an objective and subjective number?

We can show it as x - y where x is the raw NPS and y is the users idea of how difficult it is after factoring easy and hard parts. When you edit the chart difficulty after you finish the song, you have two values to enter. You enter the NPS on the x value, and what you think the difficulty is for the y value.

For example, moches Apple Butter (<3): His easy chart is listed as a 5. Its NPS is about 5.3 (I think?). So it would be displayed as 5 - 5.

My files are much easier than most others, so something like No Such Thing as the Promised Land would look something like 6 - 3.

Since SM5 shows the length of the song in song select, with 2 numbers giving you different pieces of information, you should have a better idea of what to expect from the chart.

[Jousway]

Originally Posted by Kraezymann

Is there a system where its NPS AND the older system (I'm talking about the system I saw in the FFRCP4)? What if you made it a combination of both? A combination of an objective and subjective number?

We can show it as x - y where x is the raw NPS and y is the users idea of how difficult it is after factoring easy and hard parts. When you edit the chart difficulty after you finish the song, you have two values to enter. You enter the NPS on the x value, and what you think the difficulty is for the y value.

For example, moches Apple Butter (<3): His easy chart is listed as a 5. Its NPS is about 5.3 (I think?). So it would be displayed as 5 - 5.

My files are much easier than most others, so something like No Such Thing as the Promised Land would look something like 6 - 3.

Since SM5 shows the length of the song in song select, with 2 numbers giving you different pieces of information, you should have a better idea of what to expect from the chart.

it already displays like that

[Jimmymi1]

I see alot of potential in this system. I would also love to see Kaiten's idea being done.

Mabye we can make a fusion out of this, like the graph showing density (mabye a weight system where it shows where there is regular stream, jumpstream, handstream and jacks?) + an NPS average graph maybe?



Just me having ideas though, doubt it will ever happen.

[NeoMasterPie]

sawa saw it

[Dj_Ossa]

Make the program count the amount of arrows per measure and then average them out maybe... Plus, obviously including the other methods of calculation? Sorry if this sounds dumb, I just don't really know how NPS works