Why not add a reduce-pp-amount by x factor to the DT mod? You could even have this factor scale differently depending on AR. I'm pretty sure you thought about that, just interested in what your thoughts against it are.

February 18, 2015

Just like the last 1000 times this has been asked: You could make a nomod map which is the exact same map as another map with DT. It would be plain stupid if the version with DT would give less pp despite being the exact same map. If you feel like DT is overpowered then what's actually the case is that you feel fast beatmaps with moderately high OD and AR are overpowered, hence those deserve a nerf and not the mod DT which just acts as a "generator" for such a beatmap.

Liked by: kablaze DroppedBass pyon what Tang Cloud Genesis

I was analysing HD and I noticed that the bonus it gives at lower accuracies on 5.7 stars maps at OD7 is huge. https://osu.ppy.sh/s/93398 at Sudden death 90.50% acc at HD is equivalent in pp than a 95% no mod play, but at 99% HD seems to be comparable to the SS no mod which is much more better.

February 18, 2015

The hidden bonus is almost completely independent of accuracy. What you are seeing is simply that each % of accuracy gets worth more as it gets closer to 100%. As a consequence the relative values of accuracy and the hidden bonus also change. Keep in mind that stuff like that also greatly depends on the map in question.

Liked by: DroppedBass

Is there any point in bonuses? Won't players just get good at doing high CS as they did with DTs?

February 18, 2015

If that ends up happening and high CS is generally accepted as too strong then I have no problem with reducing its value again. I'm just adjusting pp to the feedback I get as good as I can.

Liked by: DroppedBass Fenza

Does a score with 10 missed sliderends have more pp than a score with a sliderbreak that results in 10 missed combo?

February 18, 2015

Should be the same pp as long as the overall amount of 300s, 100s, 50s, misses and maximum combo is the same. That's all the numbers the pp algorithm gets to work with after all, so it's impossible to distinguish 2 such scores.

Liked by: DroppedBass

https://osu.ppy.sh/u/652457 hr and dt are now friends forever

February 18, 2015

Best FriendS

Liked by: pattrik VanillaSushi Try Harder Rafacar DroppedBass B1rd what Whale~ Ryan (Rairyu)

so to clarify, you mean that getting a 100 on a sliders will give it the same aim pp as a circle, or a full slider just with lower acc?

February 18, 2015

Yep. And of course with a combo lower by 1. (Which is another very minor pp decrease) And of course assuming you don't break your combo.

Liked by: Michael Warming

I don't see you talking about underweighed sliders much when it's one of the biggest current problems with ppv2. I have an idea, calculate pp of a short slider route, but not the shortest (which is humanly impossible). But also make getting 100's on sliders give as much aim pp as a circle would.

February 18, 2015

If I don't take the shortest slider route than fast repeat sliders suddenly explode into overpoweredness. Also getting 100s on sliders without breaking already gives (almost, maybe 99.5% due to side effects of the lower accuracy) the same aim.

Liked by: DroppedBass

So what's the difference between a back and forth jump and a straight line jump? I don't think one is easier than the other. Make the back and forth 30 degrees, and now it's easier than the straight line 150?

February 18, 2015

I'd agree with that. I'd however also say that chained straight line jumps are (marginally) harder than chained back and forth jumps.

I'm curious if it's not just that very long jumps are overrated? I don't know how pp works for the jumping distance, but the general "accepted" difficulty of distance is logarithmic (with inverse to size of target) (see fitts's law)

February 18, 2015

My original approach pretty much used fitte's law with the logarithmic scale but that gave pretty bad results. (pre-osu!tp times)

You just made me realize something else, though. The scaling of distances based off CS is currently normalizing map scale based on circle _radius_. I never tried doing this with circle _area_. That might explain why small circles were always perceived as "underpowered". I'm afraid such a change might affect CS4 vs CS5.2 too much, though. Would have to experiment. Also then it would completely violate the fact that players can just change the resolution of osu! to scale their circles to arbitrary sizes. :x

Liked by: DroppedBass

Have you ever thought about slightly increasing the amount that scores excelling in one area are increased? I know you talked about the formula, (aim^1/x + speed^1/x + acc^1/x)^x and said that x=1.1 or so. I believe there might be some merit to slightly increasing x?

pooptartsonas

February 18, 2015

Tried it once. The entire top50 became TV size DT people for some reason. My hypothesis is that aim is too prevailant on those maps compared to the other stats, making maps like koigokoro and so on explode even more.

Liked by: Bonsai DroppedBass

Ehm tom, I'm not sure at all, but seems that OD is a way too overrated compared to the amount of tapable elements on a map. Because oftenly 500 combo easy DT patterns at 99's% are giving more pp than no mods 3 times larger with similar jumps, more streams but at OD7.5 or OD8 at equivalent accuracy.

February 18, 2015

You should probably wait until aim, speed and accuracy are actually displayed separately before guessing how much the OD contributes. It's usually lower than what you all seem to think. Also, are the nomod jumps you are talking about just larger but slower? That would partly explain what you describe.

Liked by: DroppedBass

I doubt this will ever happen, but what if the community of osu grew so large to the point of say cod or csgo. do you think this game would still be a good game or would the new community ruin it? Would you still play?

February 18, 2015

Why would I stop to play the game because of other people?

Liked by: DroppedBass 7402 Try Harder VnSoul Tang encryptik Kailyn

If a lot of hit circles on a map increase the acc bonus do sliders increase the combo bonus? (Especially the fast ones)

February 18, 2015

What do you mean by combo bonus? The bonus for map length is determined by the amount of hitobjects, not by the maximum combo you can achieve. The scaling how much pp you get for a play based on which combo _you_ achieved relative to the maximum possible combo takes the combos of sliders into account.

Liked by: DroppedBass

What are the benefits of going open source? (I'm completely ignorant about this kind of stuff)

February 17, 2015

The benefit is that everybody can see how exactly pp is implemented and can contribute changes to the algorithm.

Liked by: DroppedBass Olivarus Bufa Kolamanov Darklord142

on what Mode Diff-Formula are you working right now? and are you planning to revisit some older stuff soon? like CTB&Taiko

February 17, 2015

After my recent changes CTB seems to be in an okay state. At least far better than before. Taiko also seems more or less alright when comparing it to osu!mania which is what needs the most work at the moment in terms of the raw pp algorithm.

I'm currently not doing any work on that, however. I've just refactored the way difficulty attributes such as "Aim" and "Speed" are internally stored to make it easier for future applications such as a display on the website and ingame easier to be done properly. This also involved polishing up the pp processor.

I've began work on making the pp processor ready to go open source, including some refactoring, only using libraries which have appropriate licenses and further things. The pp processor should go open source sooner or later. Keep in mind, that this does not include the difficulty algorithms since those are located in the osu! client.

Liked by: DroppedBass

http://ask.fm/Tom94/answer/125516042190 Back and forth jumps (Airman) are quite hard too

February 17, 2015

Those are back and forth jumps with spacing changes. Without spacing changes back and forth jumps are usually relatively easy. Of course ideally a pattern difficulty algorithm should also include spacing changes in the greater scope instead of just stupidly rating each jump individually based on its spacing (and maybe in the future also by its angle).

You can find many more things that make patterns hard and in the end coming up with a formula that actually captures all these things properly without doing anything completely wrong is no simple task as you maybe see here.

Liked by: DroppedBass

Would that mean maps like daidai genome or koigokoro would be debuffed? that approach would mean that HR gets nerfed as well, and making no progress in the pp system (unless only really pp giving maps like that were nerfed for their patterns, i don't really know what you're planning)

February 17, 2015

Only maps whose difficulty primarily lies on patterns like that would get nerfed. Overall maps like koigokoro and genome would in all likelihood be nerfed more than most other maps resulting in a more balanced set of difficulties. I can't say that for sure without actually testing it properly, though. So far all the approaches I tried had some severely bad side-effects.

Liked by: DroppedBass Bufa Kolamanov

The closer a jump's angle is to 90º the harder it gets imo.

February 17, 2015

For me jumps with an angle larger than 90 actually become even harder until they become very close to a line. Things like pentagons or hexagons are the hardest for me (compared to squares with the same spacing).

Liked by: Bonsai DroppedBass Bufa Kolamanov

http://ask.fm/Tom94/answer/125514296782 he's not saying change to change it alot, pretty sure just like a slight buff. i mean i guess its understandable unless you also slightly buffed 9.8 a bit less than 10, but the 2nd sentence: "more people play dt so no point in buffing it for the 90% that can't

February 17, 2015

Even then it's an arbitrary and inconsistent change (buffing high level HR plays which shouldn't be buffed in this regard) trying to cover up the actual issues of such DT maps, being certain jump patterns just being overrated usually. See my previous answer for a probably way better approach.

Liked by: Yano

What is it that makes acute angles so overpowered?

February 17, 2015

Angles aren't taken into consideration by the algorithm, so with the same spacing squares or any other pattern would give the same amount of aim difficulty as acute angle patterns.

What makes acute angles OP is that they are easier to play in almost every case. Pretty much all the "overpowered" DT maps currently have long streaks of large acute angle jumps. Even talking about the arguably OP united + HR map acute jumps are prevailant.

I will definitely try to include that in the algorithm which should hopefully help reducing the difficulty of such maps.

Liked by: pattrik DroppedBass Wieran CptBlackBird kablaze Darklord142 FunOrange Tang Kynan

http://ask.fm/Tom94/answer/125513871822 buff ONLY od10? so dt doesnt get a buff or a nerf and only HR maps that are od10 will get a slight buff to balance it out at the lower ranks?

February 17, 2015

So that all HR maps which are worth a lot give 600pp instead of 450? Or so that maps with OD9.8 (which are also frequent with HR) give much less than OD10 for absolutely no reason?

I agree that the way the accuracy bonus scales is not optimal, but just giving a bonus for one specific number is a horrible, horrible idea.

I wouldn't even be so sure whether a probability-based approach to OD scaling would help much. It's just that 90% (probably even more) of the players prefer going fast at AR9.6 rather than slower at a for many hard to read AR10 while having to get good accuracy at ridiculously high OD.

Liked by: DroppedBass Bufa Kolamanov

I'm really lazy, and come from a time where 94k DPS (My current DPS) was considered high, what's the easiest and cheapest way to get 200-500k+ DPS easily, preferably spellcaster?

cheezstik

February 17, 2015

Please don't make this page also get PoE questions. That being said, I don't give people individual build help which takes a massive amount of time. Use your own brains or read guides, whatever you prefer.

Liked by: Bufa Kolamanov

glad you can focus on the higher ranks to balance out dt and not give a fuck about all the shit ar8+dt farmers in the 2k-600~ range rofl

February 17, 2015

That was an AR bonus change. How the hell am I supposed to nerf DT compared to HR with an approach based bonus if DT+AR8 has a _lower_ approach rate than HardRock? I'm getting more and more inclined to just respond with insults to questions like that which are clearly just trying to provoke by repeating known issues without giving anything constructive.

Seriously, anybody who doesn't understand peppy's behavior needs to go into some balancing department in a game. And I'm saying that as someone who only gets a fraction of the annoyance that people like peppy do.

Liked by: Bonsai DroppedBass Bufa Kolamanov Yano

Are you the same tom94 from poe that made the crazy builds?

cheezstik

February 17, 2015

Yep.

Liked by: DroppedBass

when will the clientside star-rating agree with the webpage on CS5.3+ maps?

February 16, 2015

On cutting-edge it should agree already. If not, please point me to the map in question. If you're on stable then it will agree as soon as the next stable release happens.

Liked by: DroppedBass Michael Warming

Tom Long Dong

Nothing here yet