Simple buff for low AR

[owbra]

Background

Most would agree that low AR is very underweighted with the current PP algorithm. To address this, I'm proposing a simple buff to aim PP based on AR.

Calculation Proposal

For AR less than 7:
Multiplier = 1.6

For AR between 7 and 10:
Multiplier = 1 + (10 - AR)/5

This factor is multiplied by aim PP to give the buffed aim PP. This translates to a 20% buff for AR9, a 40% buff for AR8, and a 60% buff for AR7 and below. These values may seem a bit extreme, but remember this is only being applied to aim PP. After total PP is calculated, the buff turns out to be very modest.

The rationale for the constant bonus below AR7 is that plays in this range are far more likely to be memorized as opposed to sightread (not including easy/medium difficulties). So, there should not be any additional AR reading bonus.

Some Results

AR7.5
-GN | Ryoushi no Umi no Lindwurm [Death of the Quantum Sea] 96.49% FC
314pp -> 401pp

AR8
chocomint | Yakumo >>JOINT STRUGGLE (2014 ReWorks) [Phantasm] +HD 98.98% FC
354pp -> 392pp
Mismagius | C-TYPE [SS-TYPE] 100% FC
232pp -> 298pp
WubWoofWolf | Cry for Eternity [Legend] 99.33% FC
361pp -> 418pp

AR9
Vaxei | Rainbow Dash Likes Girls (Stay Gay Pony Girl) [Holy Shit! It's Rainbow Dash!!] 99.77% FC
487pp -> 546pp
chocomint | Kami no Kotoba [Voice of God] +HD 99.86% FC
460pp -> 511pp
Rafis | FREEDOM DiVE [FOUR DIMENSIONS] 99.93% FC
586pp -> 627pp

EZ/HT Scores
Riviclia | Koi no Hime Hime Pettanko [Taeyang's Ultra Princess] +EZ,HD,DT 98.13% FC
485pp -> 644pp
-GN | The Promethean Kings [The Merciless] +EZ,HT 98.68% FC
397pp -> 520pp
Cappy | Those Who from the Heavens Came [Fengshen Yanyi] +HT 99.75% FC
395pp -> 429pp

Notes

1. The low AR buff shouldn't apply to FL scores, since AR is practically irrelevant in this case.
2. Some may have concerns about easy/medium maps receiving buffs since they naturally have low AR. This shouldn't be a major issue, since the buffs will be minimal (i.e. 40pp -> 44pp).
3. HD already has a low AR buff in the current algorithm. This buff will be applied on top of the one that already exists, giving HD plays two separate buffs.

[abraker95]

This will buff lower difficulty maps since they typically have lower AR

[stanriders]

And what about maps that are easy to read (like original daidai genome) or maps on which EZ is actually making the map easier (like hidamari)? AR8-AR9 buff also seems uncalled for and will make hard/insane maps overbuffed

[stanriders]

Also i dont see why AR7 is given the same buff as AR1 for example, they're nowhere near in terms of reading

[owbra]

This will buff lower difficulty maps since they typically have lower AR

The buff for these maps is so small that it doesn't really matter. As I mentioned above we're talking about buffs like 40pp -> 44pp. If this isn't acceptable then we could easily add a check requiring base aim PP to be above a certain threshold before applying low AR bonus.

And what about maps that are easy to read (like original daidai genome)

Maps like daidai genome may be easy to read with AR8, but the point is that they would be even easier to read with AR9. This buff addresses that increase in difficulty from AR9 to AR8.

maps on which EZ is actually making the map easier (like hidamari)

If EZ makes a map easier, then the base aim PP value should reflect that. If that's not the case, I don't think it's in the scope of this fix to address.

AR8-AR9 buff also seems uncalled for and will make hard/insane maps overbuffed

I don't really think so. Like I said, the theory is that these maps deserve a buff because they would be even easier with higher AR.

Also i dont see why AR7 is given the same buff as AR1 for example, they're nowhere near in terms of reading

It's basically because AR7 is the cut-off. Below this value, it's assumed that memorizing the map is easier than reading, so no additional reading bonus is given. This value can be changed depending on what people think, but I think AR7 is pretty accurate.

[stanriders]

we're talking about buffs like 40pp -> 44pp

This might not be lot for you, but for lower ranked players thats a pretty big difference.

Maps like daidai genome may be easy to read with AR8, but the point is that they would be even easier to read with AR9. This buff addresses that increase in difficulty from AR9 to AR8.

I meant daidai with EZ (DT or no DT, doesn't matter), which is pretty easy to read even if you're absolutely not used to low AR due to circle placement and lack of overlaps.

If EZ makes a map easier, then the base aim PP value should reflect that. If that's not the case, I don't think it's in the scope of this fix to address.

You're trying to buff reading-heavy maps, right? If reading a map isn't hard it shouldn't get a bonus, so its within the scope of this proposal.

the theory is that these maps deserve a buff because they would be even easier with higher AR.

That's a matter of preference and map layout. Lower ranked players won't call ar9 easier than ar8 and on some maps higher ar makes them more uncomfortable. I personally would always prefer ar9.2 over ar9.8 or ar9 over ar10 because to me it feels more comfortable and easy to read.

Below this value, it's assumed that memorizing the map is easier than reading

This isn't true. Plus, you're basically giving every EZ play a 1.6x bonus no matter the ar.

I don't think adding buffs based just on AR for non-hidden plays is a right idea. If anything, it would make more sense to use object density instead of AR, but then it can be abused by making maps easy to read even with high density and streams would also be overweighted while streams are the easiest pattern to play on low AR.

[abraker95]

@owbra Low AR is not hard because it's low AR. Low AR is hard because of the patterns. I recommend you research what those patterns are and how they affect difficulty at various AR.

[owbra]

This might not be lot for you, but for lower ranked players thats a pretty big difference.

If it's an issue, then there's a simple fix as mentioned above. If raw aim PP is below a certain threshold, assume low AR is not a factor in reading, and discard the bonus.

I meant daidai with EZ (DT or no DT, doesn't matter), which is pretty easy to read even if you're absolutely not used to low AR due to circle placement and lack of overlaps.

The only question that matters here is, "is this map easier or harder to read with lower AR?" All else being equal, AR8 is harder to play than AR9. It doesn't make sense to try to evaluate this buff by comparing EZ versus nomod on Daidai, because EZ alters the difficulty of the map in several ways other than AR. It's not an apples-to-apples comparison. The same argument applies for Hidamari.

Lower ranked players won't call ar9 easier than ar8 and on some maps higher ar makes them more uncomfortable.

Low-rank players won't even be affected by the buff if the threshold correction is used.

I personally would always prefer ar9.2 over ar9.8 or ar9 over ar10 because to me it feels more comfortable and easy to read.

Sure, but self-reported preferences aren't really the most scientific way of going about this. Plenty of people will say they prefer low AR despite performing better on average with high AR lol. The only thing we have to work with is the general trend we see in plays, which is that players perform worse as the AR decreases.

Granted, the AR9-AR10 region may be a bit fuzzy since we're edging on the upper-limit of human reaction times with AR10, meaning visual strain starts to kick in. I'm open to introducing another buff for this end of the spectrum (actually one already exists, but it only applies for AR10.33 and above).

This isn't true. Plus, you're basically giving every EZ play a 1.6x bonus no matter the ar.

It's a simplifying assumption. It's never going to be 'true' but it can be good enough. Either way, we have to incorporate some kind of diminishing return, because memorization is obviously more prevalent in plays with very low AR. The formula can be adjusted so the buff slightly increases from AR7 to AR0, but it can't be too large or we'll be rewarding tons of reading PP for plays that were just memorized.

Low AR is not hard because it's low AR. Low AR is hard because of the patterns.

This really isn't true. There's inherent difficulty in low AR that has nothing to do with patterns. Reading is more difficult when there are more objects on the screen, regardless of how they're arranged. Sure, that difficulty may be amplified by the existence of difficult patterns, such as overlaps, but it's wrong to say that there's no increase in difficulty from high to low AR without those patterns.

[abraker95]

Reading is more difficult when there are more objects on the screen, regardless of how they're arranged.

Reading is not always more difficult when there are more object on the screen. Streams have many more object on screen than jumps yet they are definitely not harder to read than jumps that have an anchored overlap.

Sure, that difficulty may be amplified by the existence of difficult patterns, such as overlaps

Does difficulty increase due to difficulty patterns such as overlaps? Yes. Is it amplified due to difficult patterns such as overlaps? No. There is no base difficulty to amplify because it is derived from how those patterns appear. There can be a visible overlap at lower AR and no visible overlap at higher AR. Or there can be no visible overlap at lower AR. Whether there is an overlap is dependent on how notes are arranged. Simply multiplying the difficulty by a constant number undermines that - you assume there's always going to be an overlap by multiplying like that.

but it's wrong to say that there's no increase in difficulty from high to low AR without those patterns.

AR with approach time of 5 seconds with notes appearing once every 5 seconds has the same visual complexity as AR with approach time of 1 second with notes appearing once every 1 second. The only change is the rate at which you need to process that information. I argue the latter is harder because you need to process notes appearing faster.

[hostilew]

Seems pointless to do an AR buff that only extends down to AR 7. I'd say thousands of players can probably read maps below that. And I disagree with this stuff about low AR always being objectively harder, Just look at any linear map like chocobo or any stream map. You are greatly over valuing how difficult low AR is on those maps.

[owbra]

Reading is not always more difficult when there are more object on the screen. Streams have many more object on screen than jumps yet they are definitely not harder to read than jumps with an anchored overlap.

Very true, that's why the buff is only applied to aim PP.

Does difficulty increase due to difficulty patterns such as overlaps? Yes. Is it amplified due to difficult patterns such as overlaps? No. There is no base difficulty to amplify because it is derived from how those patterns appear. There can be a visible overlap at lower AR and no visible overlap at higher AR. Or there can be no visible overlap at lower AR. Whether there is an overlap is dependent on how notes are arranged. Simply multiplying the difficulty by a constant number undermines that - you assume it's always going to an overlap by multiplying like that.

This is all true, except, my proposal doesn't assume anything about overlaps, and doesn't have anything to do with overlaps. The goal of the proposal is to buff the intrinsic difficulty of low AR which is independent of object placement. Patterns are irrelevant because my proposal buffs an aspect of reading that has nothing to do with patterns. If someone decides to buff overlaps in the future, it would be a buff applied on top of this one.

AR with pre-appear time of 5 seconds with notes appearing once every 5 seconds has the same visual complexity as AR with pre-appear time of 1 second with notes appearing once every 1 second. The only change is the rate at which you need to process that information. I argue the latter is harder because you need to process notes appearing faster.

This isn't an accurate portrayal of the problem because you're essentially comparing two different maps with different AR. A more appropriate comparison would be between a map with pre-appear time of 5 seconds with notes appearing every 5 seconds, and a map with pre-appear time of 1 second with notes appearing every 5 seconds (i.e. same map, different AR). Simply from a timing perspective, the map with the shorter pre-appear time will be easier.

Seems pointless to do an AR buff that only extends down to AR 7. I'd say thousands of players can probably read maps below that.

Well, it doesn't only extend down to AR7. There's a constant 60% buff to aim PP below AR7. If you're arguing that there should be an additional bonus above 60% for maps below AR7, that's probably not a good idea because EZ will start to become ridiculously overweighted.

And I disagree with this stuff about low AR always being objectively harder, Just look at any linear map like chocobo or any stream map.

Using your chocobo example: yes, the map may be 'easy' despite being AR6, but it would be easier with AR7, and even easier with AR8. This buff attempts to capture that increase in difficulty associated with decrease in AR.

Also, stream maps do not receive as much of a buff because the buff only applies to aim PP.

[kornkaobat]

I think we should buff AR using a parabolic y=ax^2+bx+c ( a > 0 ) function and set the vertex point x='most comfortable AR for players'.

[riesha]

I think we should buff AR using a parabolic y=ax^2+bx+c ( a > 0 ) function and set the vertex point x='most comfortable AR for players'.

I'm proud of you for lea r ning this us e ful piece of ma t hematics in your recent zoom cl a ss. However, this idea is clea r ly not viable and i wish you thought about it more before posting on this github issue. Best of luck d ear.

[kornkaobat]

How is it not viable, constructive criticism please not ad hominem.

[riesha]

How is it not viable, constructive criticism please not ad hominem.

Because AR itself shouldn't be used in pp calculations ^-^ But i would love to see a proof of concept of your idea in your favorite programming language!!

[kornkaobat]

Because AR itself shouldn't be used in pp calculations ^-^

Why? Doesn't the current system take into account AR? Do you want to use voodoo instead?

[stanriders]

It doesn't make sense to try to evaluate this buff by comparing EZ versus nomod on Daidai, because EZ alters the difficulty of the map in several ways other than AR. It's not an apples-to-apples comparison. The same argument applies for Hidamari.

You're free to change just AR without changing anything else. Maps won't get any harder

self-reported preferences aren't really the most scientific way of going about this

You're trying to buff a subjective thing that is reading. Unfortunately the only point of reference for it is player feedback. Scientifically less time to react -> harder.

The only thing we have to work with is the general trend we see in plays, which is that players perform worse as the AR decreases.

This is only true because how meta currently is. Couple of years ago standard AR for maps was 8 and I'm sure people from that time wouldn't agree that its harder than 9.

Very true, that's why the buff is only applied to aim PP.

There are maps that don't have much aim pp, but are incredibly hard to read

intrinsic difficulty of low AR which is independent of object placement

This isn't true either. Objectively, if you throw out object placement and patterns then low AR is easier than high AR because you have more time to react. Low AR is only hard because of patter recognition, AR itself isn't difficult.

Doesn't the current system take into account AR?

Its just because there's no proper way to measure reading yet

[hostilew]

I'll be more specific. It doesn't seem wise to instate a buff for low ar but to stop scaling it below AR 7.
You're basically overlooking half the reason to do a low AR buff and it should be addressed here.

Also you should be using note density instead. AR 9 at 240 bpm is as difficult to read as AR 8 at a comparatively lower bpm. This wouldn't affect streams disproportionately because the buff should be aim based aswell like you said.

[paulthebest]

ok this is getting a little wild here but hear me out.

Low AR can be easy if the map is slow. It's only difficult if the map is fast-paced because the map would be very dense

So basically, a map is harder to read if it's dense. So instead of buffing low ar, buff the density

[ghost]

Introduction to my Easy difficulty Adjustments

If playing with lower AR while the map has overlapping circles and complex patterns and fast speeds, i believe peppy shall increase low AR score multiplier for maps like these, but leave the score multiplier as is for the more simple pattern maps such as the usual 4 or 3 star maps. And the easy mod imo should be like this: AR -0, CS -1, OD -2, HP -2, 3 lives (The usual). That way Easy is more balanced for even all star maps. For pro players that play the lower maps, DT will help the AR be more readable for your speed.

Experiments with Adjusted Easy and the Easy Mod itself

This was tested PERSONALLY with the DA (Difficulty Adjust) mod in osu!lazer. I am also a 300k 6 digit globally.
this was tested with
Mods are a hint... the first is the DA, then the ones in "()" are the EZ
1.26 Star Map to 1.19 (It's Been So Long mapped by: Shmiklak)

Difficulty: CS: 1.5(1.25), OD: 0(1), HP: 0(1), AR: 2.5(1.25)
Mods: DA(EZ)
Play Stats: Score: 104,845(104,855), Accuracy: 100.00FC(100.00FC), Misses: 0(0), Grade: SS(SS)
Improved Easy

Easy

1.26 Star Map to 1.54 (It's Been So Long mapped by: Shmiklak)

Difficulty: CS: 1.5(1.25), OD: 0(5.11), HP: 0(1). AR: 2.5(5.67)
Mods: DA(EZ), DT
Play Stats: Score: 115,329(115,341), Accuracy: 100.00FC(100.00FC), Misses: 0(0), Grade: SS(SS)
Improved Easy

Easy

5.50 Star Map to 5.26 (Ai no Sukima mapped by: Log Off Now)

Difficulty: CS: 2.8(1.9), OD: 7(4.5), HP: 3.2(2.6). AR: 9.4(4.7)
Mods: DA(EZ)
Play Stats: Score: 810,394(52,472), Accuracy: 98.83(51.06), Misses: 1(83), Grade: A(D)
Improved Easy

Easy

5.50 Star Map to 5.26 (Ai no Sukima mapped by: Log Off Now)

Difficulty: CS: 2.8(1.9), OD: 9.11(7.44), HP: 3.2(2.6). AR: 10.6(7.51)
Mods: DA(EZ), DT
Play Stats: Score: 45,860(27,908), Accuracy: 45.26(39.22), Misses: 83(95), Grade: D(D)
Improved Easy

Easy

Conclusion

This Experiment here shows that my own changes within the Easy mod gives significant improvement in skill and balance even for beginner players, and higher rank players.
if not even just the more stable decreasements within how the different difficulties decrease using my stated changes within the easy mod, It seems the mod "Easy" divides them by 2, rather they are fine, but keep the AR the same.

This way by decreasing OD will give the timings way less lenient, which beginner players struggle on, decreasing this by 2 or dividing it by 1.5 will make the timing less lenient better for their timing.
Decreasing the CS difficulty will give the centered hit less strict, since beginner players also struggle with aiming, this will help keep the circles larger better for their aiming skills. Decreasing this by 1 or dividing it by 1.5 will provide quite the balance.
Decreasing the HP wont significantly do much, but even when I started as a beginner, this was a big problem for me. Decreasing HP Drain will help them stay alive a bit longer if not a while longer before they fail a life, especially for those maps that have HP 10. Decreasing this by 2 or dividing it by 1.5 will provide quite the balance.
Keeping the AR the same, for beginner players playing lower star maps such as 3 star, they still manage to maintain AR. The biggest problem for beginner players is the difficulties listed above.

I also believe the score decrease multiplier should stay the same (0.50) with the new Easy mod. Otherwise I believe it would balance out increasing the scoring multiplier for higher star maps with complex patterns and overlapping circles for the regular Easy mod.

[Givikap120]

1. This is not right place to do this: main osu repository exist
2. EZ mod is poorly designed (mostly because of the AR change), this is a well-known thing. Nobody is going to change it because it will invalidate all set EZ scores