I figured out your scoring system.

[qqwref]

Sorry if this is already common knowledge :) I discovered it by playing around with existing scores and watching a few videos which showed the ingame score counter.

Currently, the score of a song is based on two factors other than your perfect/great/etc count: the maximum possible combo for the song, and the number of judgments in the song. The maximum combo is simply the number of arrows in the song, like in FFR; the number of judgments is the number of chords plus the number of separate holds (so, this is the number of OK and Perfect ratings you would get in a AAA run).

The maximum score for any file is 833.333... * maximum combo. All scores, including this, are rounded down to the nearest integer. When you actually play the file, there are two components to your score. 10% is combo: simply 83.333... * your combo in the file. With the ingame scoring, this is added as you go. The remaining 90% is the judgments themselves. 90% of 833.333... works out to 750 points per arrow - but the game aims to make each judgment worth the same amount (so one chord and one hold are worth the same). This means that each judgment gives you a point value of 750 * (max combo / number of judgments).The judgments work like this:
- 100% for a Perfect or OK
- 75% for a Great
- 50% for a Good
- 25% for a Boo
- 0% for a Miss or NG.
The accuracy % that is shown in the game is simply the amount of this 90% block that you earned. So if you get an FC with 80% accuracy, it means you got 10% + 90%*80% = 82% of the AAA score.

I made a program to calculate scores for this game, as well as compare two scores to see which is better. You can find it here.


PS: When calculating scores yourself, you may note that on a difficult song (such as Speedcore 300 hard) it's easy to get the number of judgments simply by passing the song, but a lot harder to get the max combo. Here's a mathematical formula to get the maximum combo out:
(your score - (83.333... * combo)) / (750 * accuracy)
For instance, AlexDest's rank 1 score on Speedcore 300 hard has 1509655 score, 1123 combo, and 0.9798 accuracy (you can get this from my program even without putting in the max combo). This formula then gives 1927 max combo for the file.

[GHEZ119]

I think Krayz just said this is possibly wrong to some extent, but it's a really good formula and it looks great. Thanks for the explanation, even if it's not perfect, it's good to finally see someone take a crack at it.

Is the combo thing why every so often my displayed score at the end of a song will differ from my actual score by ~1k either side at times?

[Krayz]

I think Krayz just said this is possibly wrong to some extent, but it's a really good formula and it looks great. Thanks for the explanation, even if it's not perfect, it's good to finally see someone take a crack at it.

Is the combo thing why every so often my displayed score at the end of a song will differ from my actual score by ~1k either side at times?

I think it's right. I relayed bad info to you...I was looking at a wrong formula :x

[qqwref]

Is the combo thing why every so often my displayed score at the end of a song will differ from my actual score by ~1k either side at times?

Dunno. Keep in mind that you get extra judgment points when a hold finishes, so maybe you're looking at the score before the end of the final hold(s), and then getting extra points when they finish. It shouldn't be possible to lose points, though, unless the ingame point calculator is wrong.

[GHEZ119]

No, it's at the end of the game, when it's displaying my final score. Sometimes it's slightly off. Might just be lag on my end screwing with the calculator and the scores being submitted properly on the server side. It's normally displayed lower than it really is, so no complaints. *shrug*

[X]

Great job on figuring it out =D (1 million perfects, 1 great, BLACKFLAG SCORE)

[GHEZ119]

A small problem with your program is that when I fill every block with 999,999 (maximum of 6 digits), I can get, like, 350%. I know it's un-realistic and it doesn't need fixing, but, just for reference, since max judgements is the sum of all the above values, you should be able to input a value equal to the sum of the maximum values in the other values above it (perfect, great, good, boo, miss, ok, ng)

And combo could be potentially limitless...

But congratulations on working it out, I had no chance. :( Although I did work out that for an AAA, every 3 notes = 2,500, so 1 perfect in an AAA = 833.333333...

I had a few thoughts floating about, but I never could've cracked the whole thing. :B

[qqwref]

I just modified the program to
- allow 7 digits for combo and max combo/judgments
- calculate minimum possible values for max combo/judgments, and use those values if you provide numbers that are impossibly small

So now you can leave the max judgment number blank, and if you passed the chart it should still give you the right score. You'll still need to put in max combo though, there's no way to calculate this from the score alone.

[GHEZ119]

Problem is, there's no way to calculate Max Combo properly unless you get a Full Combo. Otherwise, you'd have to work out where you went wrong and on what chord type. (for example, if you had 1 Boo on the last note that was a quad, adding 1 to the combo would give 3 less max combo.)

Which is why combo scoring kinda sucks. xD But the program is pretty damn sweet, glad it's saving us the basic addition of working out max judgements. xD

[Tasselfoot]

This is in fact the correct formula.

Thoughts on the formula itself (as opposed to on the created calculator and person who cracked it)?

[kisler]

way too many scores end with 333 due to 1/3 being multiplied in. also, i kind of expected it to rely even more on the judgments, but i guess not. there still is the 10% from combo length.

[stargroup]

Back at KBO we did some calculations (I can't remember what they were at the moment but if you want I can dig them up) and determined that 10% was way too high to set for combo score.

In fact, we determined that the optimal percentage for combo was so low that we simply took it out.

[GHEZ119]

I don't think taking combo out is the right way to go about it, it adds a further element of gameplay, in my opinion. I feel it's better than just basing it on judgements because it adds another dynamic and makes the experience different and more unique from other games. (Besides FFR's combo scoring, of course.)

As for the numbers, they could be changed, but I don't see why. They're good as they are. Plus I like the idea of there being a possible theoretical maximum score of 666,666 on a chart. :B

[kisler]

if you really care, #99 medium has 666 notes and another chart has 1337

[stargroup]

Technically, any scoring system you decide to use is fine, and this is arbitrarily defined by whoever designs it.

However, the point of a scoring system should be to measure the skill level of the player (or time spent on the game, a weight of both, or maybe other factors too. once again, defined by who designs it), so when you design a system you need to ask yourself if the scoring is fair enough such that you like it.

If score A > score B, but the skill it takes to get score A is not necessarily higher, then your system might not be very fair. The combo system does add an extra dimension to your score calculation, sure, but it does so at a cost of this equilibrium. However, one could also argue that combo'ing most of the song does take skill, but removing the combo part of the score won't affect the fairness by too much.

So again, sure, there's nothing wrong with this system, I'm just saying that there are ways to change it to make it more fair.

[GHEZ119]

Kisler, I was on about 666,666 AAA score, which I think *isn't* quite possible right now as no charts have the right amount of judgements.

However, 3 charts have 1,337 notes.
Copypasta Hard.
Eating Candies Hard.
Sleet Hard.

[Blazier]

I never really liked the idea of combo going into the scoring. 10% seems low enough to be reasonable... but I think getting rid of it altogether would be even better.

If you could somehow up the scoring in the parts of the song that the steps are "harder"... then you'd have a very good scoring system...

Does Stepmania do some multiplication with time elapsed for each note hit? Seems sensible, since songs generally get harder as they go on...

[kisler]

Kisler, I was on about 666,666 AAA score, which I think *isn't* quite possible right now as no charts have the right amount of judgements.

bound and gagged medium

I never really liked the idea of combo going into the scoring. 10% seems low enough to be reasonable... but I think getting rid of it altogether would be even better.

then there's no point to full combo, since there's no score difference between people who get 1 miss and people who get 4 greats. one matched the max combo of the song...the other only passed it.

If you could somehow up the scoring in the parts of the song that the steps are "harder"... then you'd have a very good scoring system...

that would require calculating notes per second over intervals, and then we'd all argue over what the intervals should be.

[Gartabla234]

Thanks a lot..

[Blazier]

then there's no point to full combo, since there's no score difference between people who get 1 miss and people who get 4 greats. one matched the max combo of the song...the other only passed it.

I think that's very reasonable... The problem with comboing is that someone who misses a single note in the middle of the song will be beaten with someone who misses them at the start or end. It's not nice.