Flash Flash Revolution - View Single Post

MinaciousGrace · 08-26-2016, 12:00 AM

yeah the error bar script isnt meant to be turned off at the moment since there are dependencies on it, refer below (or above i guess i still use the backwards ffr forum viewing bs)
however you can turn off the bars themselves by going to bganimations/screengameplay overlay/errorbar.lua and doing this

this won't actually disable the actors if you're looking to further improve performance but you can do that relatively easily if you're somewhat savvy with lua, otherwise i'd just wait until the next update since it'll be even more performance oriented and customizable anyway

also oops the judgecounter is slightly misplaced

i think i'll have an update where all of that's resolved within a few days that will be compatible with the existing 5.0.10 builds before making xwidget fork exclusive releases

as for the score tracking you have to enable that in the player options, avg will show you your percent score on what you've played and ghost will show you offset from a target

not sure about notes/best score stuff missing under general that shouldnt happen

keep in mind that I'm very much in the "oh shit i better just rewrite everything myself" phase rn

@Arcorann yeah the eval bug screen was a silly oversight on my part, it's been fixed. As for your other concern, the vast majority of the time (90%+ for files without <-significant-> influence from hold dp) your score will look nearly identical to what it would be on MAX DP. The rest of the time it isn't necessarily that it's more strict than percent scoring. In certain situations it's less strict (lots of flubby misses, or, if you just happen to get a TON of greats on the very inner edge of the great window), and in others it's more strict (you mash something to combo it and get a bunch of boos). Even so the variance tends to be rather small for files outside of the consideration of hold influence (1% for anything but extreme outliers). Although this is for the j5 variant, I have -not- tested the j4 variant in any capacity.

That being said what exists now was the result of the minimum amount of effort required to come up with something that isn't based on arbitrary groupings at arbitrary intervals. My 2nd tier priorities at the moment are to implement and test alternative methods that have been suggested by others as well as revisit more complex methods that I considered before deciding the superstructure to allow all of this to work is completed first. The end goal is to create a system that players can look at the results of and say "yeah, that pretty much sums up how well I played" to the extent possible with the information provided. That doesn't necessarily stop with a curve function designed and tested in the span of under 6 hours, and indeed we may never get there collectively as a community. However we aren't ever going to cross that finish line if we just sit down and stop moving (aka the last decade).

Since the timing data/replay data is being saved to absurd levels of accuracy when I've played enough stuff I can go ahead and make use of my actual programming skills and do some larger scale analysis of many different methods applied to the same set of offsets given varying situations. The important thing is to dissociate how you are scored by the current system with how you should be scored. If you want everything to match what DP would give you the best option is to just stick with DP. The current systems remain in place, unchanged, and are accessible for people who may want to use the other features of Wife without committing to a new scoring system.

well rather than editing my previous post I might as well finish what i was going to say in there, here

my current priority list goes more or less like this (not fully complete, when I really reach a good stopping place to sit down and consider everything I'll update this and write it out in more detail)

tier 1 (stuff im actually gonna do)
- Fix song search so it isn't somehow causing tidal waves in alabama or whatever its doing right now. Not because i particularly care about having this feature. Not because it's important to any other development, and indeed it's quite possibly the most independent feature of anything I'm considering. But because I want to show that someone with 0 game design or development experience prior to about a sum 2 weeks of doing applied scripting in lua while learning it can write a fast, efficient, fully integrated while also independent search functionality into a game that hasn't had it in the official build for almost 15 fucking years now. It's honestly just a disgrace at this point.
- finish writing new implementations of the existing gameplay screen features. The unfinished ones are currently the nps display/graph and the percent/target tracker. Reorganize them in a way that isn't clearly cobbled together after 25 hours of continuous programming. Further optimize those that are mostly completed, and make sure they're customizable and independent from one another to a degree that's reasonable for the amount of time that i want to spend on that.
- Finish the final integral component of the wife system which is to recalculate existing scores in the database depending on whether or not the parameters for ms based system or judge for standard systems have been changed. So essentially, if you change from j4 to justice, you can see what your scores would have been had you played on justice instead. Obviously this doesn't consider that you'd have probably failed almost anything playing on justice before you finished it, but that's not important. And again, obviously, conversion can't be done for any scores that don't have replaydata saved.

tier 2 (stuff ill probably do before i run out of steam in this project)
- implement and devise alternative systems to the current curve. My major issue with the current curve is it's slightly too harsh at the tail end and not quite harsh enough towards the beginning. Resolving these issues using the current system would require different intervals to have different curves applied to them but then we're kind-of-not-really heading back where we came from. Better alternative is to use something based on normal distributions.
- various small qol implementations regarding commonish situations. Mostly stuff like oneshotmirror mode that will make any mirror selection set in player options automatically turn off after selecting a different song (so you can quit out and move to another song, mirror will turn off, or complete the song, mirror will turn off, but not if you want multiple attempts using mirror and don't select something else). Also an alwaysusemirror option that's specific to the chart key so that the chart in question is always automatically played with mirror enabled without altering the actual notedata or .sm files. This is mostly for me honestly and I could give shitall whether anyone else finds them useful.

tier 3 (da pipe dreamz)
- various miscellaneous features including built-in lightweight difficulty calculator, kind of like osu's star rating, only if they knew what they were doing
- very rough pattern/file type detection that can classify files with one or more tags of stream/jumpstream/jack/handstream etc.
- sorting system that allows you to sort cross sort between file type and estimated difficulty. Basically what everyone has always wanted forever but nobody has the balls or brains to do, and sadly I don't think I'll be able to do anything beyond a cursory implementation of this before getting disgusted or bored.
- Built in bpm collapser for dem ddream files that uses the chart key hashing system to make sure it doesn't make any edits that fundamentally change the file. You guys are all really stupid for not asking me to run my flux fixer on your stuff the performance difference is incredible. But I guess I could possibly reward that stupidity by eventually doing it for you.

08-26-2016, 12:00 AM	#28
MinaciousGrace FFR Player Join Date: Dec 2007 Location: nima Posts: 4,278	Re: hello allow me to introduce to u my Wife yeah the error bar script isnt meant to be turned off at the moment since there are dependencies on it, refer below (or above i guess i still use the backwards ffr forum viewing bs) however you can turn off the bars themselves by going to bganimations/screengameplay overlay/errorbar.lua and doing this this won't actually disable the actors if you're looking to further improve performance but you can do that relatively easily if you're somewhat savvy with lua, otherwise i'd just wait until the next update since it'll be even more performance oriented and customizable anyway also oops the judgecounter is slightly misplaced i think i'll have an update where all of that's resolved within a few days that will be compatible with the existing 5.0.10 builds before making xwidget fork exclusive releases as for the score tracking you have to enable that in the player options, avg will show you your percent score on what you've played and ghost will show you offset from a target not sure about notes/best score stuff missing under general that shouldnt happen keep in mind that I'm very much in the "oh shit i better just rewrite everything myself" phase rn @Arcorann yeah the eval bug screen was a silly oversight on my part, it's been fixed. As for your other concern, the vast majority of the time (90%+ for files without <-significant-> influence from hold dp) your score will look nearly identical to what it would be on MAX DP. The rest of the time it isn't necessarily that it's more strict than percent scoring. In certain situations it's less strict (lots of flubby misses, or, if you just happen to get a TON of greats on the very inner edge of the great window), and in others it's more strict (you mash something to combo it and get a bunch of boos). Even so the variance tends to be rather small for files outside of the consideration of hold influence (1% for anything but extreme outliers). Although this is for the j5 variant, I have -not- tested the j4 variant in any capacity. That being said what exists now was the result of the minimum amount of effort required to come up with something that isn't based on arbitrary groupings at arbitrary intervals. My 2nd tier priorities at the moment are to implement and test alternative methods that have been suggested by others as well as revisit more complex methods that I considered before deciding the superstructure to allow all of this to work is completed first. The end goal is to create a system that players can look at the results of and say "yeah, that pretty much sums up how well I played" to the extent possible with the information provided. That doesn't necessarily stop with a curve function designed and tested in the span of under 6 hours, and indeed we may never get there collectively as a community. However we aren't ever going to cross that finish line if we just sit down and stop moving (aka the last decade). Since the timing data/replay data is being saved to absurd levels of accuracy when I've played enough stuff I can go ahead and make use of my actual programming skills and do some larger scale analysis of many different methods applied to the same set of offsets given varying situations. The important thing is to dissociate how you are scored by the current system with how you should be scored. If you want everything to match what DP would give you the best option is to just stick with DP. The current systems remain in place, unchanged, and are accessible for people who may want to use the other features of Wife without committing to a new scoring system. well rather than editing my previous post I might as well finish what i was going to say in there, here my current priority list goes more or less like this (not fully complete, when I really reach a good stopping place to sit down and consider everything I'll update this and write it out in more detail) tier 1 (stuff im actually gonna do) - Fix song search so it isn't somehow causing tidal waves in alabama or whatever its doing right now. Not because i particularly care about having this feature. Not because it's important to any other development, and indeed it's quite possibly the most independent feature of anything I'm considering. But because I want to show that someone with 0 game design or development experience prior to about a sum 2 weeks of doing applied scripting in lua while learning it can write a fast, efficient, fully integrated while also independent search functionality into a game that hasn't had it in the official build for almost 15 fucking years now. It's honestly just a disgrace at this point. - finish writing new implementations of the existing gameplay screen features. The unfinished ones are currently the nps display/graph and the percent/target tracker. Reorganize them in a way that isn't clearly cobbled together after 25 hours of continuous programming. Further optimize those that are mostly completed, and make sure they're customizable and independent from one another to a degree that's reasonable for the amount of time that i want to spend on that. - Finish the final integral component of the wife system which is to recalculate existing scores in the database depending on whether or not the parameters for ms based system or judge for standard systems have been changed. So essentially, if you change from j4 to justice, you can see what your scores would have been had you played on justice instead. Obviously this doesn't consider that you'd have probably failed almost anything playing on justice before you finished it, but that's not important. And again, obviously, conversion can't be done for any scores that don't have replaydata saved. tier 2 (stuff ill probably do before i run out of steam in this project) - implement and devise alternative systems to the current curve. My major issue with the current curve is it's slightly too harsh at the tail end and not quite harsh enough towards the beginning. Resolving these issues using the current system would require different intervals to have different curves applied to them but then we're kind-of-not-really heading back where we came from. Better alternative is to use something based on normal distributions. - various small qol implementations regarding commonish situations. Mostly stuff like oneshotmirror mode that will make any mirror selection set in player options automatically turn off after selecting a different song (so you can quit out and move to another song, mirror will turn off, or complete the song, mirror will turn off, but not if you want multiple attempts using mirror and don't select something else). Also an alwaysusemirror option that's specific to the chart key so that the chart in question is always automatically played with mirror enabled without altering the actual notedata or .sm files. This is mostly for me honestly and I could give shitall whether anyone else finds them useful. tier 3 (da pipe dreamz) - various miscellaneous features including built-in lightweight difficulty calculator, kind of like osu's star rating, only if they knew what they were doing - very rough pattern/file type detection that can classify files with one or more tags of stream/jumpstream/jack/handstream etc. - sorting system that allows you to sort cross sort between file type and estimated difficulty. Basically what everyone has always wanted forever but nobody has the balls or brains to do, and sadly I don't think I'll be able to do anything beyond a cursory implementation of this before getting disgusted or bored. - Built in bpm collapser for dem ddream files that uses the chart key hashing system to make sure it doesn't make any edits that fundamentally change the file. You guys are all really stupid for not asking me to run my flux fixer on your stuff the performance difference is incredible. But I guess I could possibly reward that stupidity by eventually doing it for you. Last edited by MinaciousGrace; 08-26-2016 at 01:39 AM..