Alright, really long post incoming.
Not going to say much about the mod thing because I don't really have a constructive enough opinion to say anything. LNs seem fine.
I do have reservations with the scoring component however.
This is how osu!mania ScoreV2 currently works:
How does it work currently?
- 20% of the 1m is (combo / 10 + 1) * base_hit_values.
- 80% of the 1m is accuracy ^ 10.
ComboThe combo component is probably easier to talk about because it shows a decent amount of effect overall - there are a few problems with a combo-based proportion, mainly the fact that it does not take into account the frequency of misses (though this is for more extreme) AND the position of misses in a given chart.
In a chart with consistent difficulty (i.e. the chance of missing is the same throughout the entire chart) - which are common, there's a good chance where you can miss right in the middle and if you fully SSS (all 300g) the chart otherwise, you'd only get 25% of the combo score if you SSS'd the entire chart. That is the equivalent of 150K, which is almost certainly a decisive victory for the other team. Of course, this is the most extreme case, but the fact that you can take that much damage from just a single mistake in the chart is very much overkill. This tends to have similar effects even if you don't miss right in the middle, too.
I think a combo-based mechanic is fine if it's done properly, and changing the proportions to make combo even smaller is more of a band-aid, rather than a fix. I think the problem lies in the fact that combo is not scaled in something similar to a logarithmic scale (obviously not to that magnitude, but you get the idea). Dividing combo by 10 doesn't solve the problem, so I think it would be fine to get rid of the "/ 10" thing. I'm not sure what exponent you can use to scale down larger combos and make smaller combos matter relatively more, but I think it would be a good start to use something like that to scale it down.
Having a exponential to scale down larger combos will also help with the location of misses too, it makes the location of misses matter less, and it also punishes players who miss consistently in harder areas (which is the general case) against players who can smash through harder areas but missed one note in a much easier area.
This current combo component is good if you're trying to strongly enforce consistency/FCs, but the metagame has not shifted
enough for FCs to be taken as
absolute absolute importance, unlike osu! standard. Even if you scale larger combos down with an exponential, it would still make FCs important, but at least the round is still salvageable.
I don't have any practical examples on hand, but judging from most people's screenshots, it seems to be a very extreme direction that not many people like. Again though, I think a combo-based mechanic would work fine if done properly, so it just needs some tweaking.
So that's for combo.
AccuracyWhile accuracy^10 didn't seem like much on paper, but the difference was really substantial when I looked at the numbers. This is what the accuracy component would look like - there's also a 1mil equivalent for easier comparison.
For a comparison, a 96% on score v1 is about 850K on average (it's 664K with scorev2, scaled). You can argue that v1 and v2 are not strictly comparable, but it's moreso to show how much of a drop in score scorev2 could potentially bring with just a 4% drop.
The problem with this is that the exponent used is far too strong of an exponential to use. It might be fine for earlier tournament rounds (forces players to be a lot more consistent and accurate with charts that they should be very much comfortable with), but the exponential is most definitely too great for something like semi-finals/finals, where there is a massive variance in performance across multiple teams. A player who can't get 94%+ on any of the maps in the finals/semi-finals mappool (which is very much viable given the diversity of the mappools) would essentially be dead-weight and would more than likely lose the round unless he has very strong teammates to back him up - players who'd get about 98.8% on average against three players with 97%. That difference is massive and it makes for more blowouts and less variance, which goes against what score v2 is mainly implemented for.
So the main problem is the magnitude of the exponential - but you also want to make the gaps between a 95% and 96% noticeable enough to be noticeably larger than a 99% and 100%. I guess a mediatory point would be something like accuracy^(n-accuracy)? This is what it looks like for accuracy^(6-accuracy).
A 96% with that looks more similar to the one in scorev1, and the drop seems more reasonable, too. The main drawback is that the difference is pretty minute compared to accuracy^5, but I think using a base like that would be a good start. You could do something like accuracy^(7-(2 * accuracy)), etc. as well.
Again, this current system would be fine if you really want players to be deadly consistent and all-rounders, but I think encouraging people to do it to that much of a magnitude is far too much and is too much of a shift compared to the current meta. I think a subtler magnitude is more applicable and will create a finesse that creates more variance and excitement than frustration and blowouts.
That should be all. I'm sorry if not many of my thoughts are coherent, but I have a really bad headache as I'm typing this and I admittedly didn't plan on writing this much to begin with. Hope you can put these thoughts into consideration.