Current issues with osu!lazer mod multipliers

[osuKingling]

With the recent addition of the daily challenge along with personally completely transitioning to lazer, I believe there are a quite a few flaws with the current mod multipliers. After speaking with other players I don't seem to be alone in this opinion so i'm writing this up to hopefully bring more attention to the current issues, along with a few proposals to fix them.

Issue 1: DT Multiplier

With the addition of rate change, it has become clear the multiplier between the intervals is currently too small. This is causing higher rate plays to require too high of a relative accuracy to beat a lower rate. An example to show this is the leaderboard of the daily challenge from July 26th.

In the image above you can see the 99.33% 1.80x play is above both the 2.0x and 1.9x plays. It seems be common consensus that the 2.0x score shown is much harder to achieve than the 1.80x play however the multipliers don't reflect this.

Another issue with the current DT multiplier is it's interaction with HR. Currently a HRDT(1.1x) score is weighted higher than a DT(1.4x) score. This one is slightly more map dependent, however from experience playing the daily challenges speeding up HR is often much easier than the DT only equivalent speed. Below is an example from the daily challenge from August 5th.

As you can see, my HDHRDT(1.1x) score with 99.27% beats the HDDT(1.5x) score, despite HDHRDT(1.1x) being only 5.82* yet regular HDDT is 7.5*. Of course we all know star rating is not 100% accurate, however this is a huge discrepancy.

Proposal to fix: increase the current multiplier from +0.02x per 0.1 speed, to either +0.04x or +0.05x per 0.1

Issue 2: HD/HR Multipliers

It's been asked for a long time in the community, but due to the new changes in score system weighting accuracy more it has become even more important to give HR a higher multiplier than HD. On almost all maps, HR is significantly harder than HD due to the fact it increases the mechanical difficulty of a map whereas HD only changes the reading difficulty. HR also increases the OD making it harder to get high accuracy, which is why in the new score system it is even harder to beat HD scores with HR.

There is a number of ways to fix this, however I personally believe the best solution is to both nerf the HD multiplier, and slightly buff the HR multiplier. This is because in tandem with the DT changes above, HR would need a slight increase in order to keep it's weighting relative to that in check.

Proposal to fix: Decrease the HD multiplier from 1.06x to 1.02x, increase the HR multiplier from 1.06x to 1.10x

Issue 3: Classic mod multiplier

This is probably the most controversial issue, however it seems to be universally agreed that the current classic mod is too low. In most cases the difficulty introduced by slider accuracy is not enough to outweigh the current 0.96x multiplier, so I believe an increase is required.

Further more, currently mod combinations like HDHR (1.13x) outweigh CLHDDT (1.12x) which is obviously unfair.

Quite a few people would like it to be 1.0x in order to preserve the integrity of old scores however I personally disagree with that as playing with slider acc is just objectively harder, and also the correct direction for the game. This debate is for another post though, so I will assume we are keeping it less than 1.0x.

Proposal to fix: increase classic mod multiplier from 0.96x to 0.98x

Footnote

Obviously these changes need a formal discussion, so I encourage people to give their input in the comments on my proposals, aswell as bringing up any other issues I may have missed.

As a sidenote: a lot of these issues are also worsened by the new score system, and whilst I believe the new score system is a huge improvement to old scorev1 I do think a few tweaks there would also help with these (primarily slightly decreasing the weight of accuracy). That is out of scope of this post though so may write up another one of these for that.

[Exploity]

Another easily noticeable multiplier issue currently is HDHRFL being worth more than HDDTHR even tho the latter is way harder on most maps.

[Givikap120]

1.06 for HD is fine
It's only too big when HD is together with DT
1.02 will make NM almost always superior to HD

HD should be 1.05
HR should be 1.1
and DT should be like 1.3
HR is similar difficulty to DT 1.2x for average 5-6 star map (and this star range should be main focus as most active players are playing here)

DT multipliuer = 1.3 means that:
DT 1.1x -> 1.06
DT 1.2x -> 1.12
DT 1.3x -> 1.18
DT 1.4x -> 1.24
DT 1.5x -> 1.3
DT 1.6x -> 1.36
DT 1.7x -> 1.42
DT 1.8x -> 1.48
DT 1.9x -> 1.54
DT 2.0x -> 1.6

FL and BL should probably be around 1.2

Those multipliers should be tested tho

[Givikap120]

givikap I tried to write a reply to this but it became less about our differing opinions about HD's affect on score and more just picking apart why argumentatively I think what you're saying is flawed. I don't think it makes sense to treat this as an attempt to derive HD's multiplier from some kind of objective measure in the first place -- it's all up to our preferences and experience playing HD vs. NM, and how we feel that should factor into score. I'm just much closer to the "1x" side of this argument because I personally feel little to no difference in difficulty between HD and NM. also, I shouldn't have to clarify this, but I'm not your convenient strawman that somehow got worse at NM by maining HD... again I think it's not helpful to the discussion to be including stuff like that in your posts

thanks for the scoring examples though. 1.05x has more of an effect in your example than I'd personally like to see but it's not as bad as I thought since lazer's acc scaling effectively takes some power away from every mod multiplier at the top end. I guess I don't have a big issue with 1.05x if the more popular opinion ends up being to "keep HD around where it already is" or something

If there was no difference in difficulty in HD compared to NM - we would've seen everyone playing with HD
What is exactly what happens on HDDTHR, because on AR11 there's indeed little to no difference. But moving to AR10 - and the difference is becoming very clear and noticeable

[cl8n]

I can't tell if it's by malice or ignorance but your first words there ignore my entire point and attempt to refute something I didn't even say. (that's probably my last reply here because I'm just getting annoyed now and don't want to bring that vibe to github )

[Givikap120]

something I didn't even say

I'm just much closer to the "1x" side of this argument because I personally feel little to no difference in difficulty between HD and NM

The fact that top players play with HD almost always when it comes to AR11 and much more rarely when it comes to AR10 - is enough evidence to understand that calling HD a "preferential mod" is fundamentally wrong. It's not about the preference - it's about the HD adding more and more difficulty as AR goes lower and lower.

It's the same situation as some people calling low AR "a preference" when in reality it's just as difficulty increasing as every other way to increase the difficulty such as CS increase or Rate increase. The only difference is that it works in much less predictable and simple way.

If some people think that 200bpm stream is easier than same stream but slowed down to 170bpm - it's not a preference - it's a skill issue on 170bpm.

You can call it "malice" or "ignorance" as much as you want, but was you who ignored the fact of top players playing HD 24/7 when it comes to AR10.6 and not playing HD when it comes to AR10 and especially AR9, and then based your "preference" point on this fact not existing.

[cl8n]

you keep trying to tell me evidence for why you think HD has an objective affect on difficulty and I'm saying that I don't believe this approach is appropriate for deciding a score multiplier for HD. my opinion on HD's score multiplier comes from my personal experience and my desire to see it not affect leaderboards so much as a result. it's not that deep. I "based [my] 'preference' point" on the fact that I have preferences... to which your whole top player behavioral study thing is obviously unrelated; I haven't ignored that at all, I directly addressed how I felt about "[deriving] HD's multiplier from some kind of objective measure"

(okay I actually unsub this time because I forgot to do that 😢 (please stop making threads miserable) (also I clearly failed at not bringing annoyedperson energy to github but u dont win em all i guess))

[osuKingling]

This will also be my final reply but clayton summed things up well

Whether or not you want to go study every top player ever or try to put objective measures on things, the reality is there is a not so small subset of players who will tell you directly that they prefer HD to nomod and find it easier or the same.

Look at maliszewski, cookiezi, xootynator, woey, etc. All these guys clearly only play with HD, set all their best scores with it and show no signs in their plays that nomod is sizably easier for them on lower ARs. What do you say about those players? Do you think if they took HD off they would be able to easily set better scores?

tl;dr you keep trying to put this objective analysis on everything whilst also dismissing any player who says they prefer HD as 'skill issued on nomod'. I just think your argument is biased and disingenuous because you are quite literally dismissing a player (clayton) who is telling you they prefer HD and calling them skill issued.

[felipemarins]

I would like to add that the Traceable mod have a 1x multiplier currently, which doesn't make much sense IMO. I would suggest it to change to 1.06x like Hidden or something close.

[osuKingling]

yup i agree, whenever the skin exploit with that mod is fixed I think it should be the same as HD

[Moltanicaa]

Imo - HR should be obsoleted when MR will ranked

[cihe13375]

I think we can make a simple comparison between scorev1 multipliers and lazer multipliers

Let's say a NM SS play has the same score with an HR FC with X% accuracy.
In scorev1 it roughly means X%*1.06=1, so X=94.3%
Assuming that we want to keep this value X same on lazer. On lazer an FC with 94.3% accuracy has a score of (if I recall the formulas correctly...)
500k + 500k * (94.3%)^5 = 873k
it should be same as a NM SS (1000k) after the score multiplier is applied, so the multiplier is 1000k/873k = 1.145

The same calculation can be done for DT, giving a multiplier of 1.275 (~784k before the multiplier)

The calculations above did not consider non-FC plays, because in scorev1 it's very difficult to beat FC plays with non-FC unless it misses at the very beginning or end. In lazer this becomes possible, but I think it is an improvement. Also from my observation of LGA one need to perform quite well (reasonably close to FC) to get a ~800k score, so the combo part should not have a drastic effect.

The difficulty decreasing mods are different stories though, because in these cases we shall compare e.g. an EZ SS with a NM of X misses and Y% accuracy. That's probably not important if we only consider the top of leaderboards, though.

[osuKingling]

I don't think we want to convert scorev1 multipliers to lazer, in most cases they felt far too low imo. I think something like 96.5-97% is what we want for HR and ~99% for HD as opposed to what it is in stable. Thats my opinion at least, it would need more input for final values

[cihe13375]

maybe we need a community survey (probably within lazer players, since lazer acc is not very comparable with stable) with questions like "NM SS should be same as _____% FC HR" and derive the multiplier from the results (e.g. 97%~1.076x and 99%~1.025x). I recall that the lazer scoring formula itself was built in a similar way (asking players many questions like this and adjust combo/acc portion and exponent to fit the results).

[ZenoG2]

This post only looks at competitive osu!. It ignores lower star ratings. For casual play, I can safely say that HD is harder than HR. This change would completely nuke casuals. So we now have a casual issue
Keep things like how they are and pros complain
Change things and casuals complain.
The whole score multiplier system is flawed.
We should have a dynamic system where score is variably affected based on beatmap but that will complicate things even more, and we all love to know that HDHRFLDT(x2) has a multiplier of 1.52.

[osuKingling]

Sure? If you open the lazer multiplayer lobby tab right now the average star rating is around 4-4.5ish, HR starts to become harder than HD (on average) usually at around 3.5* in my experience so this is another argument in favour of increasing the HR multiplier.

(The reason its about 3-3.5* is because this is where AR starts hitting 10 and OD is usually above 9)

[osuKingling]

Also one last thing - the average player progresses from 1-2* to above 3* within the span of a few days to weeks, there is not many players who regularly/exclusively play maps where HD is harder than HR.

[ZenoG2]

Heavily varies tho. Sometimes it feels like all lobbies are 2* and other times it feels like all is 7*
Also I don't really agree with your 3.5* figure, but that's a controversial claim so I won't argue.

[osuKingling]

Worst case scenario the breaking point definitely isn't above 4*, and I don't think theres any particular weighting above or below this for the average multiplayer lobbies difficulty so I don't think your point above adds any additional weight to your idea

[Givikap120]

This post only looks at competitive osu!. It ignores lower star ratings. For casual play, I can safely say that HD is harder than HR. This change would completely nuke casuals. So we now have a casual issue Keep things like how they are and pros complain Change things and casuals complain. The whole score multiplier system is flawed. We should have a dynamic system where score is variably affected based on beatmap but that will complicate things even more, and we all love to know that HDHRFLDT(x2) has a multiplier of 1.52.

having mrekk DT FC as map #1, surpassing HDHR FCs, is infinitely more important than 7 digit local scores that don't even show up on the leaderboards
and new player usually don't even play HD in the first place

[MrAntonym]

Another mod multiplier issue which I've discovered by playing the daily challenge is to do with half time (HT) vs. wind down (WD)/wind up (WU). The mod multiplier of HT scales based on how it is customized (from 0.1 to 0.5 as you go from 0.5x speed to 0.99x speed) which I think makes sense since there may be maps where even small slow-downs make the map much easier (such that 0.9 score for 0.9x speed or something similar might be viable for getting the top score on a leaderboard).

The mod multipliers for WD and WU work very differently. Regardless of what the initial and final speeds are, the mod multiplier is always 0.5. This means that in terms of your score, it is always strictly better to play WD/WU going from speed (a)x to speed (a-0.01)x (or vice versa) instead of playing HT at (a)x speed. Also, you can play at whatever speed from 0.5x to 0.99x is easiest for you while maintaining exactly the same mod multiplier.

My suggested change would be to make the WD/WU mod multiplier from (a)x speed to (b)x speed just be equal to the HT multiplier for (min(a,b))x speed, but I could plausibly see some bonus weight toward having a large value of |a-b|.

The example below is from today's (Oct 28-29) daily challenge where both myself and someone else have found ourselves in a sea of better players on the leaderboard thanks to being able to use WD in this way. Using WD/WU like this is still unranked, so it doesn't affect the regular leaderboards, but it seems like something that should be fixed even if only for the sake of the daily challenge.