Kuron's Insane +DT gives the same as OK DAD + DT well thanks boys

BRAH wrote:

Kuron's Insane +DT gives the same as OK DAD + DT well thanks boys

Gangsta is kinda weird...

I have a lot of FL plays but still shows "0" at memory , help ?

обновление ужасно. нет множества скоров , куча багов, отсуствие карт опять-же, глупое распределение классов... я надеюсь вы вернете все назад или приведете в порядок, было не приятно увидеть такое. удачи, не люблю альфа версии
update badly. No sets soon, a lot of bugs, lack of maps again the same stupid distribution of classes ... I hope you'll bring it back or will put in order, it was not nice to see that. Good luck, I do not like the alpha version

Endaris wrote:

Ye, I still feel bad because you aren't even getting good acc on the inflated OD and my best acc score is a shitty one-try play on OD7+DT too.
Feels wrong, so wrong :/

My thoughts on this and why this is such a delicate situation:

What I came to realize whole working on this project is that generalizing by skill has its own downsides. An OD 10 map and OD 6 map of the same reaction would have a big difference in difficulty. Still even with the big difference in difficulty, they would still have the same reaction values. When multiple skills come into play, and one dominates over the other, we feel like other skills must be raised because of how hard the map is. It's things like this which makes it quite hard to judge some scores as the player's ability to get them is influenced by all other skills.

There doesn't seem to be an elegant solution to any of this, is there?

Well, I'd think that a complicated map should reward more accuracypoints at lower accuracy compared to an easier map.
The mainproblem here is measuring "complicatedness" which I'd attribute to reading, density, polarity-switches between 1/4 patterns and stressinducing moments.
As "stressinducing moments" I'd take stuff that involves a huge change(!) of spacing without the actual rhythm changing such as jumps within streams or very sudden antijumps after a jumppattern with huge spacing as these make it harder to keep your hand synchronisation perfect.
This kind of complicatedness is what puts map with additional difficulties apart from maps that just have some triples in trivial context. Because right now it can't:

Only reason why X-Encounter is so high because the curve for OD10 starts a lot higher than OD8, the other reasons why it is hard to acc get ignored more or less. Little Princess doesn't belong on one level with Cold Green Eyes either.

I would also involve a small factor that measures how much the hitareas of the circles overlap. The less overlap there is the harder feels the map to acc which especially causes lowbpm+high OD to receive a rating that I often see as undervalued.
Right now the accuracy-rating of both the pp-system and osu!skills feels like:
[(multiplier for amount of circles) * (OD value for SS)]^(achieved accuracy)
with SS making the exponent = 1 so that you get the maximum OD-value with the circle-weightage

Amount of circles (=tenacity+stamina rating) already tends to be lower on lowbpm so I'd say that it would stay kinda balanced if you include a multiplier based on hitarea-overlap for 1/3, 1/4 or 1/6 patterns.
As complicatedness affects the accuracy you get I'd say that you could try adding it as a summand to the exponent. That way a 95% score on a complicated map would have the chance of scoring a similar accuracy value as an SS on a trivial map that only uses 1/2 jump-patterns with some sliders inbetween. And it would also score a lot higher when SSd.
Not sure where to put the hitarea-overlap-component though.

plz create another mode

I've got this open all the time in my browser. Really nice concept, site looks snazzy. The one suggestion I have would be to display the scores that each play earned in the "Recently added" section of the user page.

riffraff11235 wrote:

I've got this open all the time in my browser. Really nice concept, site looks snazzy. The one suggestion I have would be to display the scores that each play earned in the "Recently added" section of the user page.

I think that's a great idea! Would be very nice to have a grid with the recently added scores and their respectives points which were able to get in the top 100 (or just all points, I'm not sure which one would be better), it would help me to track them when making new scores

What are all the reasons that can make your scores go down?

I know that if you overwrite your score with a new one, which can be worse in some aspect (more misses, worse acc, etc.), then you can lose some points, but it's usually something like -1 to -3 for me. But are there any other reasons? Can you just lose some plays, etc?

Or maybe there was some change in how scores' are calculated?

I am asking, because I suddenly lost a helluva of points, in comparison to what I had. Well... maybe it's exaggeration, but losing more that 10% of some scores looks a little disturbing, especially as my recent new plays were mostly better than previous ones, like getting an SS or changing +HD,HR into +HD,HR,DT without really losing acc or so.

http://puu.sh/pKag8/c7d79841b6.png
http://osuskills.tk/user/Scarlet%20Evans

P.S.
I am glad to see that you added Reaction! Good job! =)

2Kao: Fixed the bug. Thanks for the report!

2TankSuperGaming: Probably the maps are too short/too easy to give any points

2Minamoto-kun: разработка это такое дело, да. редко когда всё гладко с первого раза заводится. скоры постепенно возвращаются (я запустил скрипт для этого). это должно было когда-то произойти, т.к. структура базы данных была некорректна до этого
насчёт

куча багов, отсуствие карт опять-же, глупое распределение классов...

распиши, пожалуйста, по-подробнее если хочешь, чтобы мы могли принять какие-то меры
кстати говоря это не альфа версия

2Scarlet Evans: Most likely some of your scores were lost because of the recent update, sorry. Please, use the manual submit feature or wait till they get added back with our script running

2riffraff11235: Thanks for the suggestion. We'll think about it


Also,
The combo influence thing may be changed in future to apply only to certain skills, but I don't guarantee anything

Another thing I wanted to mention, but I always forgot to do so
On default font size, the "versus" page is little bugged, visually. The bottom panel is overlapping with the versus results, like here:

http://puu.sh/pKbKh/702bcd7f67.png // http://puu.sh/pKbYy/c171fc5c59.jpg

It gets solved, if I lower the size of font/elements on the page, either using [Ctlr]+[-] or [Ctrl]+[MouseWheelDown] by one... not sure how it's called, step, level?

http://puu.sh/pKbL0/fb3beeb402.png

For higher resolutions it can work fine, but otherwise, maybe it would be good to add a scroll bar or something, so the lower panel won't automatically overlap with results?

My browser is Firefox, my screen resolution is 1366x768.

EDIT:// Thank you for your response, Kert! I will wait

Fixed!
Clean the cache or ctrl-f5 on that page/main page

Thank you!

Sorry for posting so often, but this one really disturbs me. What exactly "Reaction" measures?

I mean, I would suppose that, after applying +HD,HR,DT, the 6.5*-7* AR 11 maps that are 200-250 BPM, with not necessarily easy patterns, would need much more reaction than 2.4* AR 10.87 map that is only ~110 BPM and mostly 1/2 snap, passed below 85% because my reaction was too slow and I ended up getting 100s' and 50s'.

The first example is Shihori - Magic Girl !!(Frostmourne) [Lunatic] +HDDTHR (97.74%) played by Rafis and gives 889 reaction score,
the second one is Morenatsu Project. - I Will Always Wait For You (KittyAdventure) [Gone] +HDDTHRFL (84.9%) sh*t-passed by me and gives 808 reaction score

I am trying to play highest ARs' from some time and I would say that 20 ms between AR 10.87 and AR 11, which is 320 ms --> 300 ms, is a huge gap. Much bigger than 470 ms --> 450 ms, 620 ms --> 600 ms etc., as we are slowly going down to a limits of what human can do (good that we don't have AR 12 ^^. )

Is there anything more to reaction that almost pure AR now? Look at the
889 reaction points for Shihori - Magic Girl !! played by Rafis, that is 6.68*, 234 BPM and is really hard, then at
883 reaction points for Tom Day & Laura Lethlean - Couldn't Be Alone (Single) (Pjecoo) [Wafu's Loneliness] +HDDTHR (96.38%), played by AdventTechno.

The second one is 3.39*, 128 BPM and notes are placed every 1/2 beat, which makes it almost 4 times lower note density. It's very easy to read and because you are practically reacting to single notes due to low note density, I would say it's much easier to pass/FC this.

This is why my very first impression is: aren't easy maps overweighted in comparison to the harder ones? Or maybe the hard ones are underweighted?

I know that highest AR rates are very hard and needs quick reaction even on easy maps, but getting vaguely the same reaction score like extremely hard and few times faster maps looks a little weird for me? If it's not weird, then maybe it's my intuition that is advising me wrongly and Reaction is measured in a little different manner than I would expect, making the map difficulty and the need of having literally few times less time for reaction negligible, in terms of Reaction scores gained?

Scarlet Evans wrote:

Is there anything more to reaction that almost pure AR now? Look at the
889 reaction points for Shihori - Magic Girl !! played by Rafis, that is 6.68*, 234 BPM and is really hard, then at
883 reaction points for Tom Day & Laura Lethlean - Couldn't Be Alone (Single) (Pjecoo) [Wafu's Loneliness] +HDDTHR (96.38%), played by AdventTechno.

The second one is 3.39*, 128 BPM and notes are placed every 1/2 beat, which makes it almost 4 times lower note density. It's very easy to read and because you are practically reacting to single notes due to low note density, I would say it's much easier to pass/FC this.

I talked about something similiar in the previous post.

abraker wrote:

What I came to realize whole working on this project is that generalizing by skill has its own downsides. An OD 10 map and OD 6 map of the same reaction would have a big difference in difficulty. Still even with the big difference in difficulty, they would still have the same reaction values. When multiple skills come into play, and one dominates over the other, we feel like other skills must be raised because of how hard the map is. It's things like this which makes it quite hard to judge some scores as the player's ability to get them is influenced by all other skills.

Was there a change on the formula for stamina and tenacity?


R.I.P. Mad Machine
I will never forget when you gave me like 700 in tenacity or what
now you've dropped to 400

Kert hasnt responded on how combo isn't skill in a certain area necessarily. Mad machine being 430 pts for a 0 miss is entirely silly - it is nowhere near that low of a level.

Khelly wrote:

Kert hasnt responded on how combo isn't skill in a certain area necessarily. Mad machine being 430 pts for a 0 miss is entirely silly - it is nowhere near that low of a level.

Actually I have 2 misses

but still the same score gave me like 700 before

and tower of heaven 94.84% nomod FC gets 525 tenacity which is way higher than that 89% 270bpm stream
to add on

I don't think tower of heaven requires even slightly lower stamina to clear compare to d.m.c. In fact, I think that d.m.c should give at least 50 lower points on stamina compare to tower of heaven since d.m.c involves some short stream that doesn't exceed 15 notes with a lot of slider to recover your stamina from, while tower of heaven consist of 3 segment of 170bpm deathstream with not much time for stamina recovery in between
and also, considering that minori eggplant difficulty only has 2 short stream of 10+ notes and 1 long stream that consist of around 40 combo, i don't think it should be only 10 points lower than d.m.c either

You guys forgot to account for the CS change in EZ while trying to calculate agility. My play on Bravely You [-Tochi's Insane] +EZDT (1xsliderbreak) gives 753 agility points, while a DT only FC gives around 777.

Hola Josde

BTW is this normal?
Mayba I need to let it load during 20-30 minutes? Lets play a LoL match.

Quick question: What is it about your setup that prevents the site from fetching new scores every time it's refreshed, the way that osu!track does?

Can you replace "Sight-reading" as description for Reaction with "High AR"?
Has nothing to do with sightreading really.

riffraff11235 wrote:

Quick question: What is it about your setup that prevents the site from fetching new scores every time it's refreshed, the way that osu!track does?

Kert set limits on how many times refresh can fetch new scores due to concerns that all of the requests submissions may go over the api limit if spammed or such. That's as much as I know.

do combos influence Accuracy at all? or just only acc.

abraker wrote:

riffraff11235 wrote:

Quick question: What is it about your setup that prevents the site from fetching new scores every time it's refreshed, the way that osu!track does?

Kert set limits on how many times refresh can fetch new scores due to concerns that all of the requests submissions may go over the api limit if spammed or such. That's as much as I know.

I just realized that some scores I thought should have been submitted probably didn't meet the criteria to be counted. The site has definitely improved a lot since the first time I've used it.

This has probably come up before, but i can't just read through 55 pages aaaaaa...



Could i pls get my account and not the person from like 2008 with the same name? ;3;

Did you try searching for the name you originally registered with?

Josde wrote:

You guys forgot to account for the CS change in EZ while trying to calculate agility. My play on Bravely You [-Tochi's Insane] +EZDT (1xsliderbreak) gives 753 agility points, while a DT only FC gives around 777.

CS is calculated in precision, agility is meant to be raw aim... it was discussed the need to account the CS or not account it on agility a very long time ago, I think it is already decided...

snyviper wrote:

CS is calculated in precision, agility is meant to be raw aim... it was discussed the need to account the CS or not account it on agility a very long time ago, I think it is already decided...

Well, remember that if circles are big enough or close enough, you can reduce aim needed for some patterns to almost zero. Especially, if the circles don't overlap on HR/nomod, but do overlap nomod/Eazy. You could just point once, then click more than single circle with the same cursor position.

There can be more extreme examples than two notes clicked per single cursor pointing, instead of one. It doesn't really need to be because CS=2 or so, it can be met on "more normal" CS maps too.
I don't remember exactly what map was it and I still know just a handful out of all maps, but I there was a map with quite fast triangles made from very closely placed notes, that would overlap on +EZ or +EZ,DT.
Then, you would still need very fast triangle-like movements on HR/nomod (+DT), even if they would be small movements, while on +EZ,DT you would just need to point cursor in the middle of the triangle and keep clicking. I think that on that map there were patterns with multiple of such triangles (3 of them?) placed in a manner, where they had a common middle, while they were spinning.

So, instead of making 9 small, but still quite rapid moves to click the triangles, especially on HR that makes them disjoint, you would just need to point once on EZ and keep clicking. If there are few such patterns on the map, the change in agility needed could be really quite visible and easy to feel.

These don't really need to be such triangles, there are many patterns that can be affected greatly by CS. You can not have enough agility to make some stars on HR consistently, while EZ makes them big enough to easily pass them and FC the map. Remember that you don't need to click a middle of the circle, so things like CS5.2-->CS4 or CS 4 --> CS 2 can really make clicking easier and reduce agility needed. Even if these are usually small amounts, on a long run they all add up and the difference is more significant

snyviper wrote:

CS is calculated in precision, agility is meant to be raw aim...

So, that wall of text above being said, and all these pages in this thread that I read in past... I still don't understand... how can CS be completely ignored in Agility? Like Abraker said, "when multiple skills come into play, and one dominates over the other, we feel like other skills must be raised because of how hard the map is." So, even though precision already uses CS, it should be calculated into agility too, as on smaller CS the distances between the sets of pixels inside which you need to click are smaller and these areas are much bigger itself. How is this possibly not affecting raw aim, especially when sometimes it reduces aim needed on some patterns by many times, even to zero? It's not that we just can't aim small circles, the distance between the circles can be physically greatly reduced and with big circles we can sometimes click "wherever we want", if we just moved in the right direction, then try to correct the error with next circles.

abraker wrote:

I talked about something similiar in the previous post.

I know that many questions were already answered, more or less, but the people keep asking the same questions, over and over, as they often don't really understand how things exactly works and they have their own interpretation and perception of things like "Raw Aim" or "Streaming". And going through almost 60 pages is not something easily feasible for everyone

People can think of splitting "Streaming" and "Tapping stamina" apart in different ways and kind of no one, except You guys, knows how it really works. So, they keep asking again and again, giving their own thoughts, suggestions, try to give constructive (or not) feedback, etc. The same with other skills.

I think that it would be a good move to add a little more precise descriptions, that explains what the skills exactly do and, at least briefly, give us some bites of informations about the factors needed in the calculation and their significance.
Laconic 1-2 word descriptions like the current ones can keep causing problems and misunderstandings. What do you think about it?

I agree, completely ignoring CS in "Raw Aim" just doesn't work. Aiming speed and aiming precision are strongly intertwined, and the overall difficulty is the result of the combination of the two, something like how the area of a rectangle is the product of the two sides.

As an example look at Cookiezi's Toumei HDHR play.

It gets him ~900 aim, and 500 precision.

Now I could FC that map at say CS3 or CS2 and get the same 900 agility. I could also FC a CS5.2 version at let's say 160 BPM and get the same 500 precision.

So I made two way easier plays, yet according to osu!skills I demonstrated the same 900 Agility skill and the same 500 Precision skill as Cookezi with his HR play.
Needless to say, someone who can only nail the aforementioned two plays is not even close in actual skill to being able to get that HR FC, but this is not reflected in the skill points in any way, on paper we'd appear to have the same skills.

Also map length (total length of difficult parts) seems to be largely irrelevant. What's the difference in raw aim awarded if you copy paste the same 1 minute map over 10 times? Natsukoi Hanabi DT gets me 830 aim. People can just retry it for 10 minutes and get the FC. If the map was the same map copypasted over 10 times, these same players wouldn't even be close to being able to FC it, you'd need to be at a much higher skill level to pull that off, but looking at the 800 agility awarded for that HD Uta [Himei] FC, I assume the raw agility for that lengthened map wouldn't reflect this much at all.

Basically any CS can be turned into whatever you like by resizing the osu! window, so pp approach of rescaling maps to a standard CS is a perfect basis, where that alone would fall short is the fact that you can't just infinitely increase your tablet area.

So what if we're stuck with having to play the upscaled map without being able to increase our area? You're forced to play with a higher sensitivity, which makes things gradually harder.

Imagine any map shrunk to an 1cm x 1cm area in the middle of the screen. Obviously it's harder than the original map, due to effectively having to use an extreme, much higher than normal sensitivity. This is what happens on a smaller scale as we increase CS, unless you can increase your area (there's a limit here, even with an infinite sized tablet, everyone only trains to use an area of a given size) as much as you increase the osu! window size to get back the original CS.

So I believe standardizing CS and accounting with a function for the "sensitivity" factor as CS increases is the way to go. You can still extract individual Agility and Precision values, by dividing up the total between the two using a ratio based on the current agility and precision numbers for example.

snyviper wrote:

Josde wrote:

You guys forgot to account for the CS change in EZ while trying to calculate agility. My play on Bravely You [-Tochi's Insane] +EZDT (1xsliderbreak) gives 753 agility points, while a DT only FC gives around 777.

CS is calculated in precision, agility is meant to be raw aim... it was discussed the need to account the CS or not account it on agility a very long time ago, I think it is already decided...

Then agility should be "aim speed", not "raw aim". You can't compare the "raw aim" needed to FC a map in CS2 to the one you need to FC that same map on CS4; albeit your aim speed would be the same CS2 is way more lenient on mistakes, making a less skilled player be able to FC more easily. CS is a factor in aim, just because it is calculated elsewhere doesn't mean it shouldn't be taken in count for "raw aim".

CS is considered when calculating Agility, just not to that extent as in Precision
i.e. we calculate distance between circle borders and not their centers

It still isn't representative of actual skills, if you just separate aim speed and aim precision from the get go, as I've pointed out before.

Calculating border to border distance doesn't fix this fundamental flaw, it's only enough to solve edge cases where you wouldn't even need to move your cursor due to huge circles.

Every system that tries to break up a whole performance in pieces will have an error. That's natural so deal with it I guess.

But this one thing isn't difficult to fix, you just normalize CS, and even without additional tweaks it's pretty good. Literally the game itself allows you to do it, if you wanted to. You can still calc the weight of skill components if you got the whole thing right.

Blue and yellow areas (difficulties, skills required) aren't equal, but the current osu!skills considers them so. (Image is just meant to show the principle, not actual ratios/equations).

I'm sorry for having caused confusion, I meant things were already decided, but I don't remember what was decided