forum

Performance Points feedback and suggestions (Standard)

posted
Total Posts
2,750
show more
NazzzF

silmarilen wrote:

handsome wrote:

did something get recalcualted
read the changelog https://osu.ppy.sh/p/changelog
wait, so i just lost 30pp because of this or what? o_o
Rewben2

NazzzF wrote:

wait, so i just lost 30pp because of this or what? o_o
Yes. I'm pretty sure it was a nerf to everyone's pp though, so your rank may not have even gone down. It hasn't updated for everyone yet. I'm yet to see someone who actually gained pp from the update.
silmarilen
i gained 9pp
people who are more focused on acc than on aim gained pp
Rewben2

silmarilen wrote:

i gained 9pp
people who are more focused on acc than on aim gained pp
I know that was the updates intention, I just thought that maybe aim was nerfed more than acc was buffed so no one saw a pp increase. Guess I was wrong.

Your acc is pretty damn high on the other hand, I guess majority of people lost pp though.
Zare
>people knowing their exact amount of pp
TakuMii

Zare wrote:

>people knowing their exact amount of pp
Amianki
My pp didn't even change in that update.
silmarilen
it probably did, play a map
Vuelo Eluko
question when will our top ranks change around a bit to reflect the new pp value?
GhostFrog

handsome wrote:

The reason for playing HR is if you feel that your accuracy is sufficient that you can perform better than you would nomod.
I would like to expand on this because it's an interesting point that tends to get a bit overlooked in discussions of HR pp.

AR10 doesn't give any extra pp, higher hp drain doesn't give any extra pp. When you play HR, you're going for the extra aim pp from smaller circles and the extra accuracy pp from the higher OD and, depending on exactly how pp works, probably a negligible amount of extra speed pp from the higher OD. We can ignore the speed component of that.

Assuming you can read AR10 sufficiently well and have good enough aim to play the HR version of a map, you can FC the map with HR and get the extra aim pp without any problems. There have been suggestions that small circle size may be underrated (which, because of the way pp works, would be equivalent to saying that large spacing is overrated) and if that's true, it would cause you to receive less aim pp than you deserve - but aim for an average player is probably about 1/3 of their pre-HR pp on a map and even if HR on that specific map gives a 30% bonus to aim (I'm just making up a number because HR's effect on aim is calculated on a per-map basis by treating the HR version of the map as an entirely new map), which is pretty damn significant, you'd only notice it on average boosting your pp by 10% on average. If that's the only bonus pp you get from HR, you'll probably feel like it's significantly underweighted.

The biggest thing that makes HR feel underweighted is that most of its pp gain is locked up in accuracy and most players aren't capable of tapping into that. Let's say you're playing an OD8 map, for example. For any possible OD8 accuracy, there will be a corresponding OD10 accuracy (it's possible that it won't be possible to get either or both accuracies *exactly* on that map, but that doesn't matter) that, assuming normal distribution of hit timings and no outliers (the latter isn't a good assumption, but it's one you almost have to make), would likely require the same level of actual player accuracy to be able to achieve. Now, depending on what assumptions you want to make about things like retries (If you check Full Tablet's posts in this thread, he's done a very thorough analysis of options regarding that), you may end up with different corresponding accuracies - but the important thing is that, given reasonable assumptions, you should be able to choose OD8 and OD10 accuracies for a map that would require the same amount of effort to obtain. If you can play the HR version of the map properly and can read AR10 exactly as well as you can read the nomod AR of the map (and assuming you have good enough finger control to not have a significant amount of outlier hits), then, assuming the system is making reasonable assumptions and using reasonable formulas, you should be getting the same amount of accuracy pp on OD10 as you would on OD8 unless you can SS the map on OD8. When you can SS a map on OD8, it means you can get better accuracy on the map than the hit windows for that OD are able to distinguish and increasing the OD will credit you with the extra accuracy pp the lower OD cheated you out of. If you can't SS the map on OD8, you were already getting the accuracy pp you deserved and increasing the OD won't help you. The average player probably isn't much (if at all) more accurate than being able to occasionally SS OD8 maps, so the average player can't get accuracy pp from HR. If you're playing HR and wondering where all your pp is, that's where it is. It's locked up in the accuracy you don't have.

Now, to be fair, tom posted the formula somewhere that pp uses for accuracy and iirc it's too harsh for accuracies farther from 100% and that does penalize most players who use HR, but ultimately, the point remains: if you can't SS a map without HR, don't expect it to give you anything but aim pp.
Rewben2

Bassist Vinyl wrote:

question when will our top ranks change around a bit to reflect the new pp value?
Well it must. There's no way the scores will remain the same but our total pp changing lol, unless there is a change to how total pp/weighting works, which isn't what the update is doing.

Zare wrote:

>people knowing their exact amount of pp
Is it really so hard to take note that there's been a change to how pp works and then playing a song and looking at the difference? No knowing your pp involved.
thelewa

Bassist Vinyl wrote:

[Tom94] Slightly reduce value of aim and slightly increase value of accuracy in osu! standard.

lewa rank 1 on next play hype?...
I lost 11pp

what part of SLIGHTLY do you not understand
Vuelo Eluko

thelewa wrote:

Bassist Vinyl wrote:

[Tom94] Slightly reduce value of aim and slightly increase value of accuracy in osu! standard.

lewa rank 1 on next play hype?...
I lost 11pp

what part of SLIGHTLY do you not understand
dragonhuman losing 100+ pp
Topic Starter
Tom94

GhostFrog wrote:

handsome wrote:

The reason for playing HR is if you feel that your accuracy is sufficient that you can perform better than you would nomod.
[great explanation]
I'd just like to note, that basing the formula on a normal distribution is still something I want to do at some point. :)
silmarilen

GhostFrog wrote:

bunch of text
tl;dr: if you can't SS a map without HR, don't expect it to give you anything but aim pp.
this post should be linked everytime someone says hr is underrated
Drezi

GhostFrog wrote:

assuming the system is making reasonable assumptions and using reasonable formulas, you should be getting the same amount of accuracy pp on OD10 as you would on OD8 unless you can SS the map on OD8.
It explains why things are the way they are very well, but this assumption quoted is not true as far as I know (?? not sure, correct me if I'm wrong), and that would be part of the reason why people have a hard time getting pp from HR unless they can keep the number of 100s they're getting with it to a comparable amount (since in that case they're getting the aim pp, and they're getting more or at least not much less acc pp). If the quoted assumption is not true, that would mean lower accs on high ODs are in fact a bit undervalued.
Vuelo Eluko

Bassist Vinyl wrote:

question when will our top ranks change around a bit to reflect the new pp value?
j/w if/when it will happen, because it seems weird that some of my highest yield scores that also have abysmal acc havent changed at all yet 16 total pp did...
Neroh
Not sure if this has been said already but, I feel like some ar9 maps just aren't giving enough PP. DT seems to be the only way to get good amounts recently.

For example, this map on eXtra, https://osu.ppy.sh/b/231917?m=0 got me 217pp, which is a lot really.
Yet if I was to FC this map on FREEDOM (a lot harder) https://osu.ppy.sh/b/283705?m=0 with similar accuracy, I'd only get 220pp.

It leaves me with no option of rank gains for some of these maps even though they take a lot of skill to full combo.

I have no idea how much work goes into making the PP system work, but I'm grateful for what you have done. Sorry if I'm asking the impossible haha ^^
Full Tablet

Neroh wrote:

Not sure if this has been said already but, I feel like some ar9 maps just aren't giving enough PP. DT seems to be the only way to get good amounts recently.

For example, this map on eXtra, https://osu.ppy.sh/b/231917?m=0 got me 217pp, which is a lot really.
Yet if I was to FC this map on FREEDOM (a lot harder) https://osu.ppy.sh/b/283705?m=0 with similar accuracy, I'd only get 220pp.

It leaves me with no option of rank gains for some of these maps even though they take a lot of skill to full combo.

I have no idea how much work goes into making the PP system work, but I'm grateful for what you have done. Sorry if I'm asking the impossible haha ^^
There is a big difference between certain accuracy on OD9 and the same accuracy on OD7.
Khroto
The problem of this change concerning the weight of acc. is that on very hard song, at a 'pro' level, with acc. near 100%, the global feedback is that accuracy gives to much pp.

On the contrary, that's right, at a 'common' level, fc an easier song in hr with 97.5% can give you less pp than if you SS it with nomod whereas 97.5% in hr can be more difficult to achieve than SS + nomod, all the more it is also harder to aim with hr.

So, accuracy appears to be more 'mis-weighted' than underweighted or overweighted. And so, rather than changing the weight of acc. (maybe) it would be better to decrease to a lower base the exponential function used to compute accuracy while keeping the fucntion in the same mean average (so that the weight, compared with aim and speed, remains the same). Thus acc. would give a fairer amount of pp in the two cases I described above (well at least I suppose, I can be wrong).
Cinagro
Not nearly at a high level, but just dropping this here for anybody that could enlighten me on this a bit.

I nearly full combod https://osu.ppy.sh/b/351513?m=0 yesterday, playing me at around ~#400 for the song. I expected a few pp out of it, but I didn't get anything. Is this because the star difficulty/ppv2 calculated difficulty for the song is incredibly low compared to all my other "top ranks"? If so, what is factored into the star rating system?
Rewben2

Cinagro wrote:

Not nearly at a high level, but just dropping this here for anybody that could enlighten me on this a bit.

I nearly full combod https://osu.ppy.sh/b/351513?m=0 yesterday, playing me at around ~#400 for the song. I expected a few pp out of it, but I didn't get anything. Is this because the star difficulty/ppv2 calculated difficulty for the song is incredibly low compared to all my other "top ranks"? If so, what is factored into the star rating system?
Have a look at the star rating, it's 3.47. I took a look at the plays from near the bottom of your top performances and there's Mizutani Runa - Philosophyz, which has a star rating of 3.58. It was also a higher acc score than your tear rain and an fc and both the songs are od8. Star rating is a good indication of how much pp a map will give.
Dexus
Cinagro play a LOT of 4.5 star maps and get 97.5% to 98.5% and you should see your rank go up. Anything less than that would be a waste of time for you. As you rank up more and see the maps give less (or you're starting 99%/100% them) then play .1 higher stuff, if that's the same result then go up .1 more. You should improve as a player and go up ranks as time passes.

Group the maps by difficulty and sort by difficulty, then select 5 stars (5 stars shows 4.5 to 5.5) start from the top and work your way down.
xasuma
I am not sure how important the length of the maps is in the algorithm. But I am feeling length should be weighted slightly higher. (It shows consistency)
Sometimes I feel that it doesn't matter much if the map is 1:30 minutes long, or 3 minutes long. (I am sure it does matter) Just that long songs are harder just because of the amount of time you need to stay focused at once.

Just a thought.
higurush

xasuma wrote:

I am not sure how important the length of the maps is in the algorithm. But I am feeling length should be weighted slightly higher. (It shows consistency)
Sometimes I feel that it doesn't matter much if the map is 1:30 minutes long, or 3 minutes long. (I am sure it does matter) Just that long songs are harder just because of the amount of time you need to stay focused at once.

Just a thought.
I could give it an evaluation too. I've seen thousands of "less than a 1 minute" replays, which are given around 240-260+ pps. Ofc they are with DT mods. So here's the fact. It is really easy to farm pp with short 140-170 bpm maps. You put DT on them, and whoala 1 minute fun for a ton of pp and I guess the 2 or 3 minute ones are given the same amount of pp as I've seen :/
Vuelo Eluko

xasuma wrote:

I am not sure how important the length of the maps is in the algorithm. But I am feeling length should be weighted slightly higher. (It shows consistency)
Sometimes I feel that it doesn't matter much if the map is 1:30 minutes long, or 3 minutes long. (I am sure it does matter) Just that long songs are harder just because of the amount of time you need to stay focused at once.

Just a thought.
i dont know, if i compare maps of equal star diff but different lengths i get a different feeling. for example comparing this with this, the TV Sized map is just more difficult than any 1:30 "section" you could cut out of the longer one. In fact, if it weren't for the build-up on the last kiai + the kiai itself i would say its star diff was a bit too high.

you need to be focused longer, but the skill required is less intense overall so it's easier to be consistent.

There seems to be a cap on how much length matters though, because marathons aren't OP sources of PP.
Rewben2

higurush wrote:

I could give it an evaluation too. I've seen thousands of "less than a 1 minute" replays, which are given around 240-260+ pps. Ofc they are with DT mods. So here's the fact. It is really easy to farm pp with short 140-170 bpm maps. You put DT on them, and whoala 1 minute fun for a ton of pp and I guess the 2 or 3 minute ones are given the same amount of pp as I've seen :/
So basically, you think longer maps should be worth more pp? A lot of the top plays I see aren't even tv size though, there is 3+ minute songs in the mix. I think map length in relation to pp is fine as it is.

I think longer maps are better for more consistent players and tv-size is more beneficial for people who spam retry maps and maybe aren't as consistent to keep up a good play for a longer time. Some people would be simply better at one style than the other, I think it's quite balanced.
higurush

Rewben2 wrote:

So basically, you think longer maps should be worth more pp? A lot of the top plays I see aren't even tv size though, there is 3+ minute songs in the mix. I think map length in relation to pp is fine as it is.

I think longer maps are better for more consistent players and tv-size is more beneficial for people who spam retry maps and maybe aren't as consistent to keep up a good play for a longer time. Some people would be simply better at one style than the other, I think it's quite balanced.
Maybe they should worth more in some cases.

Let's say there is a TV size map+DT - it will be around 1 minute - and there's a longer version+DT - it will be around 3 minutes. Let's say the two maps contain the same elements (longer version = play tv size 3 times). The pp evaluator will say it will be the same difficulty, because everything is the same and will give the same amount of pp for both of them. What if you can DT the 1 minute map easily, then when you would do the same with the longer version, you would get tired at 2 minutes and fail the map.
<The same case without DT applies.>

Also if you are playing a song compilation with the same difficulty and every setting being the same as another map. They would give the same amount of pp. However, there is higher chance of missing in the longer map than in the shorter one. This way, you will play the marathon map only once or twice and play the shorter one a ton of times. There was a problem back in the days of people not playing the short maps at all, because it wouldn't give them as much score point as the longer ones would, but this case flipped over and less and less people playing the longer maps. It's a never-ending problem with a ton of reasoning behind everything..

That's what I meant.
Akari-

higurush wrote:

Let's say there is a TV size map+DT - it will be around 1 minute - and there's a longer version+DT - it will be around 3 minutes. Let's say the two maps contain the same elements (longer version = play tv size 3 times). The pp evaluator will say it will be the same difficulty, because everything is the same and will give the same amount of pp for both of them.
copypasting the same map 3 times makes the star rating go up, it's just that the amount it rises is weighted

http://i.imgur.com/lV5BKmc.png
GhostFrog
You do get more pp from maps with higher max combo. It's not that significant though in comparison to star rating except with really long maps (The marathon and this diff have almost identical star ratings, but the marathon gives a lot more pp (or at the very least, it gave a lot more tp and pp is usually extremely similar)). If you did have a map that contained 2 copies of another map, it would have a higher star rating. I don't know how much higher, but the max combo scaling for pp is relatively low because it only attempts to account for consistency, while in theory leaving the added difficulty from the extra length to the star rating system.
silmarilen
the extra pp on the marathon is most likely because it has more notes so you get more acc
GhostFrog
Actually, yeah, you're right. I remembered there being a bigger difference between them. The bonus pp you get from higher max combo really flattens out after a certain point, so the 1275 max combo from In the Middle of the Night gives almost the same bonus as the 9289 combo from the marathon diff. I know there was some amount of combo scaling for tp from looking at pp gains of SS plays on a lot of maps and it probably carried over to pp. (I only checked the aim and speed tp values; I used SS just to avoid any possible accuracy or slight-combo-miss scaling on aim and speed)
higurush
What I overall meant with my comment is that playing long maps are not even worth struggling for. You're better off playing random TV size/short maps if you want pp.
Example:
Various Artists - Nightcore Mix Compilation [Party!] +DT (97.57%) >> 178pp
Primastella - Koigokoro [Delis' Insane] +DT (97.89%) >> 203pp
Koigokoro is way easier..

Also since the ?aim? nerf update, it's a bit ridiculous for the differences in jump maps vs etc

Hatsune Miku - Homework Crisis [Let's Jump!!] (99.56%) >> 206pp
Primastella - Koigokoro [Delis' Insane] +DT (97.89%) >> 203pp
Koigokoro is still easier and it's just a slight difference.

What's more I'm around #100 on Nightcore mix compilation and Homework Crisis, and on Koigokoro i'm like around #1000..
Where's the rank bonus? And as for the Homework Crisis, the "Jump" nerf was this harsh?
uzzi
There isn't a rank bonus anymore...
higurush

- [ U z z I ] - wrote:

There isn't a rank bonus anymore...
Then the "do the same as the crowd" method really is there.
Even if there is no rank bonus the mentioned and complained maps should worth more pp considering their "real difficulty"..
xasuma
Well its just a matter of looking at most people's top play's length. I just think it's slightly underrated.
As someone said above, it would be better for people who are more consistent, that is exactly my point. Being consistent should earn you something, it's a skill not just a play style option.

Right? Lol
higurush

xasuma wrote:

Well its just a matter of looking at most people's top play's length. I just think it's slightly underrated.
As someone said above, it would be better for people who are more consistent, that is exactly my point. Being consistent should earn you something, it's a skill not just a play style option.

Right? Lol
So you think it is the same as with Approach Rate? AR9=AR10 in star difficulty. (this AR stuff is okay though)
silmarilen

higurush wrote:

What I overall meant with my comment is that playing long maps are not even worth struggling for. You're better off playing random TV size/short maps if you want pp.
Example:
Various Artists - Nightcore Mix Compilation [Party!] +DT (97.57%) >> 178pp
Primastella - Koigokoro [Delis' Insane] +DT (97.89%) >> 203pp
Koigokoro is way easier..

Also since the ?aim? nerf update, it's a bit ridiculous for the differences in jump maps vs etc

Hatsune Miku - Homework Crisis [Let's Jump!!] (99.56%) >> 206pp
Primastella - Koigokoro [Delis' Insane] +DT (97.89%) >> 203pp
Koigokoro is still easier and it's just a slight difference.

What's more I'm around #100 on Nightcore mix compilation and Homework Crisis, and on Koigokoro i'm like around #1000..
Where's the rank bonus? And as for the Homework Crisis, the "Jump" nerf was this harsh?
koigokoro is not way easier, if it was way easier it would give way less pp. you think it is easier because you are better at the things the map requires of you (high bpm medium size jump patterns with a couple of burts streams)

homework crisis is ar9 compared to koigokoro ar9.6 thus it feels harder to read, while the map itself probably isnt much harder
koigokoro is also od9 (od7+dt) while homework crisis is only od7, the koigokoro score probably got close to the same amount of acc as the homework crisis score got.
higurush

silmarilen wrote:

koigokoro is not way easier, if it was way easier it would give way less pp. you think it is easier because you are better at the things the map requires of you (high bpm medium size jump patterns with a couple of burts streams)

homework crisis is ar9 compared to koigokoro ar9.6 thus it feels harder to read, while the map itself probably isnt much harder
koigokoro is also od9 (od7+dt) while homework crisis is only od7, the koigokoro score probably got close to the same amount of acc as the homework crisis score got.
I feel fine with long jumps and high spacing maps as well (Remote Control HRHD, With A Dance Number+HR (max AR/HP/OD unranked)/Talent Shredder0108 +HR (max AR/HP/OD unranked)/Saten unranked maps), etc, but I accept your reasons.

What about the Nightcore Compilation, which is connected to this thread's new problem: long maps are not worth playing and struggling for?
silmarilen
i dont know about that map because i never played it, but i agree that longer maps are underrated compared to shorter ones.

the problem is how far you are willing to go with boosting them, what if it ends up the other way around? what if the bonus for length becomes so big that clearly easier maps will give more pp.
show more
Please sign in to reply.

New reply