ppy
Few but High OD circles blatantly overrated.
Actually I've seen that in more recent times mappers are giving very high OD to very short maps that in some cases are only 81 circle count like in the case of Will Stetson - Harumachi Clover's [Oh no!] which is OD9 but not too hard to add DT mod to make it worth as OD10,4.
The point is, that OD is one of the reasons why those very short maps gives an excellent reward even with not so high accuracy at all. You can even try some simulations at those same maps with Doubletime but by making their OD be OD6 , OD7 (without DT) instead of their base OD8 to 9's. You will see a large pp difference just because of decreasing the OD.
The idea I want to propose is to decrease pp effectiveness for 300's (from acc) in maps with OD higher than 9 but less than 300 circle count; affecting specially the more short maps. This focus would make reasonable that Future's Son [Insane] is affected a lot less than it's [OK DAD] difficulty if played with double-time or HR. While not affecting most of newbies that are getting used to osu! which commonly will play shorter maps.
While I've talked with some friends about the previously mentioned nerf to all those maps with less than 500 clickable elements is that they argue that this "could" affect the motivation to newer players (point in which I'm not completely agree or disagree with my friends).
This idea can even be applicable in the already in mind idea of nerfing too short maps. If the pp reward from doing higher accuracy is decreased in a considerable way in too very short maps is done, this will impair at most players that cannot be considered newbies because of being capable to even SS OD10,4 but only in very very very short maps; players that we cannot say that their accuracy is necessarily consistent at longer maps.
The minimum value of OD which are being affected from this suggestion not necessarily has to be OD9 nor the circle count has to be 300 exactly, both values can be adjusted depending in what you consider the best. But this focus unlike the elements count this targets directly only hit circles and not sliders or spinners.
Any on going idea can be posted here gradually, ofc I have a lot more to propose and even to explain, but at the moment I will leave this topic at this point.
nerfing accuracy pp on short maps is barely a bandaid, there's many long pp maps (like sidetracked day) that get a lot of their pp just from the very high circle count, which is also why a lot of mappers overmap triples and stream to try and squeeze as many circles as possible into their map.
there's not easy solution to this, you'd have to figure out some way to compute how hard it actually is to acc regardless of length and that's very subjective
IMO the problem is that LOW OD maps are underrated because the system cannot evaluate your accuracy, like for example if you SS an od7 map with 60 unstable rate, the same exact play would be worth much more pp and still be high acc if you changed the map to od10+
the only way to solve this would be to calculate pp based on some absolute OD value rather than the map's od so accuracy is "normalized". but this would require replay data which wouldn't be available for old plays (and might not be accurate enough in old replays)
Dunno if it's viable implementing the absolute OD value because of requiring replays, I do think that might be better to study some +1250 circle count maps with varies on OD.
Before doing this analysis we are considering the following:
- Circles are actually the only type of elements that actually are relevant to accuracy. For this reason I'm completely ignoring sliders or spinners because they're not counted in global accuracy calculation before each 300/100/50 value is set to each clickable element; or better said, they're not that relevant.
- This is a quick review, so it isn't deep at all, for this is needed an accuracy simulator to play with the amount of circle/sliders/spinners count without considering the placement of each circle and changing OD values to see how much pp varies from accuracy (ignoring completely the star rating).
- This may be a bit incomplete review because of not knowing certain details about of how accuracy works, specially with elements that aren't hit circles. I'm assuming that circles are the only type of relevant pattern that have a appreciable impact on accuracy value.
Case 1: God only knows -Secrets of the Goddess- (OD6 map with 1022 circles) This was not originally in my plans to be evaluated, but difference of pp from 95% to 100% pp is only 33 pp... The only exception of a map with 1200-1350 circle count of the following maps that are going to be exposed.
Case 2: OctagonCollaboration - 8 Sides of Nico Nico Douga (OD7 map with 1300 circles) Difference of pp from 95% to 100% is of about 53 pp. This is easily beatable by much shorter maps with higher OD, reason why 8 sides of nico nico douga might be an underrated marathon if no HR on it.
Case 3: DJ YOSHITAKA - CaptivAte ~compilation~ (OD8 map with 1363 circles) Difference of pp from 95% to 100% is of about 95 pp, almost the double than in 8 sides of nico nico douga; even with lower star rating at 95% has lower pp value than the previous case but beats the other case by 40 pp in SS.
Case 4: SakiZ - osu!memories (OD9 map with 1210 circles) Difference of pp from 95% to 100% is 115 pp; even if it's slightly shorter than the other 2 maps by aroudn 100 cirlce count. It's known that 95% accuracy OD9 is equivalent or better than OD7 SS; a very valuable argument of why OD7 is very underrated.
Case 5: VINXIS - Sidetracked Day (OD10 map with 1293 circles) Difference of pp from 95% to 100% is around 193 pp; almost the double than the previous case even by being slightly longer than the previous case in osu!memories. Basically in this system if you can 95% FC the song it might be much better to wait and see higher accuracy (98% or higher) and doing a fairly high combo play but with higher accuracy might reward you a lot better.
To be honest, this could be studied with a lot of care if we had an accuracy simulator that evaluates circle count and that could let you play with OD values. But seems that higher OD needs to be nerfed slightly while lower OD but at longer maps needs a buff. Making length something that favors more to lower OD and not brute forcing player to play high OD for decent accuracy rewards.
This post is extending a bit too much, going to post for more , but if anyone knows about accuracy simulator or pp simulator that allows you to variate the settings on a map that would be very appreciable.
@ReynBolt you can use my command line tool https://github.com/Francesco149/oppai-ng and override OD by doing oppai some_map.osu 95% OD7 . there's also the ezpp extension but if i remember correctly it doesn't allow overriding stats. there's also a bot in hvick's stream chat that just lets you link a beatmap and provide overrides, acc, mods in the same way
I will take a look to that a while and then I will leave my opinion.
Also, more than VINXIS - Sidetracked Day I'm more worried about the potential 800 PP farming at GYZE - HONESTY map that with only HR (without needing HD) is worth more than freedom dive 4D +HDHR ... being OD9.5 and 1987 circle count gives so much more from accuracy than freedom dive. It's obvious that Karthy's play is a brutal one, but that values 120's pp more than Cookiezi's HDHR at freedom Dive 4D is a bit too much.
Also in that one case, the problem may be (apart of the +1900 circles of OD10) is that CS5.2 itself is giving higher star amount to more aim maps instead of more speed/stamina ones; which could be potentially fixed if CS5.2 boosts more speed part and little less the aim part... which I would like to post as another topic and not on here. But I kinda feel that HR (CS5 and higher better said) needs a similar modification that the one that happened not so long ago with HD changes. Streamming 230 bpm (an example) in smaller circles needs more precision and constancy; this another proposal would only benefit mid-high to high bpm at smaller circles but not the low bpm spaced bpm streams.
My main argument to say why I find freedom Dive 4D as an harder map to HR is because of the much lower sucess FC's done at HR or HDHR than in honesty.
Okey, let's have a look at 3 very difficult maps, with nomod values for two of them and HDHR for 3 cases. Freedom Dive at being OD8 in nomod is underrated at nomod, but HR fixes this OD issue making it to be OD10 as in the other 2 cases.
1.- Xi - Freedom Dive [Four Dimensions] +HDHR (OD10 map with 1646 circles) With a 95% value of 653 pp and on SS with a value of 831 pp at SS is the map that gives lowest PP than the other 2 that are mentioned. Additional info: 7,07 stars if nomod and 7,48 stars if on HR. Seems to me that star rating increasing for HR is a bit too low for the difficulty increase that it leads in this one map.
2.- GYZE - HONESTY [DISHONEST] +HDHR (OD10 map with 1987 circles) This one map at 95% HDHR FC is worth as 782 pp, very similar to Freedom Dive's HDHR at 99% (773 pp at 99%). While on SS is worth as 956 PP. Obvious problem in here is the STAR RATING itself! Honesty 95% is a much lower tier play than Freedom Dive at 99% HDHR. This one map at nomod is 4,2 CS and with HR is 5,46 CS. To impair this map at most of the 3 cases, it's obvious that HR needs less aim boost and a better boost to speed; because of the 3 cases is the slowest at being 185 bpm; while the other 2 cases are 222 bpm and 240 bpm respectively.
This map is 7,11 stars in nomod, and 7,94 stars in HR. Much more contrast from stars than Freedom dive. 95% value for nomod at honesty is 518 pp and 655 pp at SS nomod. This is just mentioned for a fair comparative with the 3rd map that I will mention. Being OD9,5 without mods.
3.- UNDEAD CORPORATION - Everything will freeze [Time Freeze] +HDHR (OD10 map with 1212 circles) Time Freeze's FC at HDHR is worth as 764 pp in 95% value and on SS is worth as 877 pp of HDHR. In both accuracies values less than Honesty map if played on HR or HDHR if doing the same accuracy; but what is the problem on here? Everything will freeze has only 2 Full combos at nomod and NONE in HDHR which leads that Time Freeze is MUCH HARDER and NOBODY has done an HR FC yet.
This map is 7,65 stars in nomod and on HR it converts to 8,15 stars. This map has a base OD9.4, in 95% FC this map is worth as 578 pp and on SS is worth as 707 pp. In nomod is stronger at both accuracies than Honesty, but magic happens when you add HR to Honesty giving it an absurd steroid of pp because of the extra amount of circle count and length and the fact of the slightly smaller circles.
It's blatant that a shorter map with fewer circles than honesty but higher star rating gets outmatched in pp; the legendary Everything will Freeze [Time Freeze] has lower 95-100% pp value ratios than Honesty if both are played in HR just because of number of elements because time freeze has higher star rating.
Quick Comparative Table: UNDEAD CORPORATION - Everything will freeze [Time Freeze] vs GYZE - HONESTY [DISHONEST]
Nomod versus Everything Will Freeze 95 / 98 / 99 / 100% nomod : 578 pp ; 632 pp ; 644 pp ; 707 pp GYZE - HONESTY 95 / 98 / 99 / 100% nomod: 518 pp ; 578 pp ; 612 pp ; 655 pp Coment: pp difference is shorter at higher accuracy value because of the difference of the circle count.
HD versus Everything Will Freeze 95 / 98 / 99 / 100% HD: 624 pp ; 680 pp ; 712 pp ; 756 pp GYZE - HONESTY 95 / 98 / 99 / 100% HD: 565 pp ; 628 pp ; 662 pp ; 706 pp Coment: Wasn't in my plans to include an HD comparation, but after the crazy results I've got when comparing HR with HDHR at both maps decided that this was required.
HR versus Everything Will Freeze 95 / 98 / 99 / 100% HR : 707 pp ; 777 pp ; 818 pp ; 874 pp GYZE - HONESTY 95 / 98 / 99 / 100% HR : 702 pp ; 777 pp ; 818 pp ; 872 pp Coment: Weird, in HR only both maps are very equivalent !! seems that the mod that causes problems is HDHR itself. Nice finding from my part !! Still is very unfair that both maps are equivalent in HR if Everything will freeze +HR is much harder...
HD HR versus Everything Will Freeze 95 / 98 / 99 / 100% HDHR : 764 pp ; 835 pp ; 877 pp ; 934 pp GYZE - HONESTY 95 / 98 / 99 / 100% HDHR : 782 pp ; 852 pp ; 901 pp ; 956 pp Coment: I ignored about the real impact of HD that in some HR can do !! I believed that the problem was HR itself, but seems that if both mods are together things can occur in certain maps. Honestly I don't know why HDHR makes the diference of making Honesty to give more pp than Time Freeze even if in all other 3 cases it's not the case; but I think that it can be because HD scales better at longer maps and with higher elements count.
SR being imperfect is a reccuring problem of all mods i believe. I only know catch for sure but I can think of at least 2 5* ranked maps which only have 15 FCs (any mods) (namely Feel U [5.2*], and Amber's love was like a marble [5.83]). I believe the policy would be to implement something and to discuss it after. I want to try some things myself, but i will start with catch obviously.
