Several changes regarding the balancing of the Flashlight mod in osu!standard

[MBmasher]

Hi, this pull request addresses several changes regarding the balancing of the Flashlight mod in osu!standard. I will include a TL;DR at the start so you can get a quick understanding of what this accomplishes.

TL;DR

• ORIGINALLY, the FL bonus was just based on object count. More objects led to a bigger bonus.
• This change applies a nerf on plays that aren't a full combo. Having a lower percentage of the max combo will increase the effect of the nerf. The nerf is more powerful on maps under 200 combo due to the size of the FL radius.
• This bonus is further nerfed for each miss.
• This bonus is further slightly nerfed for any map above 500 objects.
• FL plays which ALSO have HD enabled are buffed. There is a small buff at AR11 and this buff further increases the lower the AR.
• Out of the above points, the length nerf and HDFL buff are MINOR. These changes mainly target non-FC FL plays.

VALUES
This spreadsheet contains values of some plays.

ISSUE: NON-FCS
Most FL plays are judged by the amount of effort and how difficult it was to memorise the map. For most maps, this difficulty is increased due to the shrinking of the FL radius. More specifically, the FL radius decreases at 100 and 200 combo. Missing a note will bring you back to the 0 combo radius. Overall, missing notes need to be punished harder with FL plays as they demonstrate lack of memory, as well as increasing the FL radius which makes the play significantly easier. (see: UNION!! +DTFL 6 misses 1258pp, fieryrage's difficulty on Harumachi Clover which is easily exploited by enabling FL)

Min(Combo, Combo^2 / 200)

Firstly, this formula was used to calculate a new combo. It scales combo below 200 differently such that non-fcs under 200 combo are nerfed harder.

baseFLMultiplier *= (Combo / MaxCombo)^2

The FL bonus is then multiplied by the ratio of combo and max combo, squared.

baseFLMultiplier *= 0.97^Misses

The FL bonus is then given a nerf of 3% for every miss.

ISSUE: LENGTH
For many FL players, the FL length bonus on long maps is seen to be too overweighted. (see: The Unforgiving +FL 1011pp). Thus, a small nerf was applied to FL plays on maps above 500 objects.

Originally, the FL bonus would scale with a factor of 1/1200 per object above 500 objects. It now scales at 1/1500.

ISSUE: HDFL
In the current pp system, HDFL does not get any special treatment - the only bonus added from FL is the HD bonus, which is already applied to all HD plays. However this does not reflect the real difficulty of HDFL - adding HD removes approach circles which removes the ability for players to spot big jumps, and makes some notes' location completely impossible to locate, since new combos remove follow points. It also makes objects fade away, which affects visibility, especially on lower AR. Overall, HDFL requires a buff due to these reasons, which scales higher as AR is decreased.

baseFLMultiplier *= 1.45 - 0.01 * AR^1.4

The FL bonus is scaled through this formula, which increases as AR decreases.

PURPOSE AND OTHER ISSUES
I am certain that a proper flashlight rewrite will take lots of time and effort and I understand that these issues are simply bandaid fixes. However, small and quick changes are still beneficial to the system and especially with issues this glaring, I thought to address them as soon as possible.

Yes, there are other glaring issues. This includes:

• FL only applying a bonus to aim pp, which leaves stream aim FL plays severely underweighted
• No difference between flow aim maps (wide angles) and snap aim maps (sharp angles)
• No consideration for gimmick maps, which includes symmetrical maps, grid maps and other maps which utilise small amount of the playfield
• No consideration for how different circle sizes may affect visibility

However, none of these can be fixed quickly without a completely new system, the creation, implementation and balancing of which could take a very long time.

Overall, this change incentivizes players to go for FL plays that are high/full combos to fully showcase their memory of maps, rather than using it to exploit short, high SR maps. It also rewards HDFL plays more to better reflect their true difficulty and may also incentivize players to achieve those sort of plays.

[smoogipoo]

I haven't gone through the whole post in depth, but on the issue of length am I right to understand that you've taken a single data point on a 1hr long map, a score that no one has done before, and called it overweighted?

I don't know that we should be balancing an entire system around one score - do you have any more data to back this up? Perhaps a map that has more FL scores at higher combo values? The small one-sentence reasoning you've given here is inadequate to me...

[MBmasher]

You can't actually see at the moment because the list of values is not there, but the effect of the length nerf is minor. It only makes a significant difference to the play which I mentioned.

[smoogipoo]

Is the effect minor because you're balancing around that one score, or is the effect minor because there's another map where minor changes have been shown to have good effects compared to other mods? And what's the criteria for having determined whether such effects are good? Is it an arbitrary nerf across the board, or is it compared against HD/HR/DT or AR values?

I'm not disagreeing it is overweighted, but it needs to have a reasoning other than providing a fringe score that it has the most effects on. I can't say that score should be worth 2000pp or 500pp when looking at it in isolation.

[MBmasher]

I'd like to argue that any FL change is arbitrary simply because of the nature of the mod - it can't be compared to any other mod or AR or such. This being a small addition to the current pp framework only adds to that.

Most of the FL community simply agrees that the length bonus for FL is a bit too overweighted, and this change addresses that. The balancing was done purely through estimating what values would be best for each map. I do not currently know of a better way to balance FL.

If this serves too much of an issue, I can just remove it as it is only a minor change.

[smoogipoo]

Do you have sources for that, or a discussion having happened somewhere? Reading comments on that play in particular leads me to believe that this is, in-fact, one of the best scores if not the best score ever set in the game, contrary to the claim that it is overweighted.

[MBmasher]

I have had discussions with -GN himself as well as had a discussion in a Discord server which contains many top FL players, and most of them have agreed.

[isakvik]

I have had discussions with -GN himself as well as had a discussion in a Discord server which contains many top FL players, and most of them have agreed.

Came here to confirm this (-GN himself). I agreed to the point of the length bonus being too generous based on that Unforgiving gets its pp from being a pretty diffspiky map coupled with the length bonus, as well as a concern about HDFL not being worth enough (although I can't really be sure it's in a good spot now either; who would even attempt that?). Maybe it merits some more discussion though, it might be too conservative seeing as it's not a property that could make long maps farmable, and given the community's faith in the pp system it might not be in the devs' best interest to nerf it.

The other changes are very good for integrity IMO.

[stockingz]

ISSUE: LENGTH
For many FL players, the FL length bonus on long maps is seen to be too overweighted. (see: The Unforgiving +FL 1011pp). Thus, a small nerf was applied to FL plays on maps above 500 objects.

If we're nerfing the length bonus (which I totally agree with, don't get me wrong.) we have to get rid of the bonus for maps under 300 combo. It is far easier to full combo a 30 second farm map with flashlight rather than something like Can't Defeat Airman with FL, and in our pp system's current state there are scores that are worth WAY more than that:

Why under 300 combo? Most farm maps that are abused with flashlight are under this combo range, this would effectively make scores on those maps with flashlight worth nothing, but the amount of memorization and skill needed to fc these maps with flashlight is little to none.

Also, the max flashlight dim is at 200 combo, meaning if a play is under that it is EXTREMELY easy to abuse.

I propose that we completely remove the flashlight pp bonus on maps under 300 combo, as it is completely unfair to people who dedicate actual effort into memorizing a map.

[MBmasher]

It's arguable whether or not the FL length bonus is overweighted for short maps. For a case like Harumachi, the map itself is already overweighted and it's only easy with FL because of its style of mapping. Either way, we can't just completely remove the FL bonus for maps under 300 combo like you suggested, because it takes an amount of effort to set those plays and thus they have to be rewarded in some way.

[bdach]

Opinions without any substantive backing arguments that support those opinions are irrelevant. If you disagree with this change, please provide reasoning.

[Purplegaze]

In your PR message, you cited Union +DTFL as an example of overweightedness due to misses + the dim resetting:

Overall, missing notes need to be punished harder with FL plays as they demonstrate lack of memory, as well as increasing the FL radius which makes the play significantly easier. (see: UNION!! +DTFL 6 misses 1258pp)

In my opinion this is not a good example to give, because in this case, the end was a net decrease on the pp of the play. If you evaluate the pp of an FC of the map with everything after 1500 combo cut out, you will find the value of an FC with the acc he had to be actually higher than the final value of the play (approximately 1315pp).

I agree with punishing chokes on very short maps (<250 combo) but I don't believe jot_bab's play is overweighted because it's a choke, and don't think that devaluing plays with any misses at all on longer maps is the right way to fix this.

Also, in reply to stockingz's comment, FL should never give no bonus at all. Even once a map has been completely memorized, the dim makes it much harder to aim, and FL aim is its own skill that is developed along with memory. I'd agree that Harumachi being 147 combo is overweighted because aiming with FL isn't that hard before you get to 200 combo, but maps in the 250-300 combo range already start to be incredibly difficult to FC, particularly common farm maps which tend to have their diff spikes at the end.

Adding to this:

Overall, this change incentivizes players to go for FL plays that are high/full combos to fully showcase their memory of maps, rather than using it to exploit short, high SR maps.

I'm not sure if this means that FCs on short maps are planned to be nerfed drastically too, but just in case it does:
For the reasons stated above I believe that chokes on short maps should be nerfed drastically, but only that. Chokes on short maps, not FCs on short maps (except <200 combo), not chokes on long maps. FL on TV size maps (like 300-500 combo usually) is more commonly used than on long maps, yes, but getting to 200 combo and keeping the FC till the end of a TV size map with high SR jumps is a different skill of its own from long memory, and shouldn't be de-incentivized just because it's more common (which is only probably the case because it's more appealing than playing long maps).

Aside from that, I agree with the HDFL buff and don't have a comment on the length nerf as I'm not personally acquainted with playing FL on long maps.

[MBmasher]

closing in favor of #14217