forum

Performance Points feedback and suggestions (Standard)

posted
Total Posts
2,750
show more
Dexus
I personally would like to see the weighting system shifted to where the top five or so scores have a higher weighting. The taper effect seems to sudden and then draws out too long. Past the top 25 it should be where it hits that 1% weight. The list view should be shorter as well since I and others don't bother to look very far down in the list. With this hitting your top five would be more rewarding and players would have a better idea of placing scores around their top ten instead of continually replacing their very top performance. It seem useless to score anything less than your best to move anywhere. More consistent players would be rewarded as well since they would have several good scores placed instead of a handful of odd weighted scores. No mod is kind of crippled because of this fact; there's not enough high rating maps no mod that are relative to where I'm ranked. Big black SS 300pp and Lewa can't FC it yet there are scores worth way more that are manageable in comparison.
Drezi

Dexus wrote:

I personally would like to see the weighting system shifted to where the top five or so scores have a higher weighting. The taper effect seems to sudden and then draws out too long. Past the top 25 it should be where it hits that 1% weight. The list view should be shorter as well since I and others don't bother to look very far down in the list. With this hitting your top five would be more rewarding and players would have a better idea of placing scores around their top ten instead of continually replacing their very top performance. It seem useless to score anything less than your best to move anywhere. More consistent players would be rewarded as well since they would have several good scores placed instead of a handful of odd weighted scores. No mod is kind of crippled because of this fact; there's not enough high rating maps no mod that are relative to where I'm ranked. Big black SS 300pp and Lewa can't FC it yet there are scores worth way more that are manageable in comparison.
Yes please, your best scores should be closer to 100%, and more than 20-25 being weighted higher than 1% just rewards you a bit for setting lots of scores at your average level.

I also hate that passive 416 bonus PP, around half of that would be enough I think (achieved by stricter weighting like 0,999 instead of 0,9994 and not cutting the individual per map bonus in half).
Drezi

Dexus wrote:

there's not enough high rating maps no mod that are relative to where I'm ranked. Big black SS 300pp and Lewa can't FC it yet there are scores worth way more that are manageable in comparison.
I think part of the reason that nomod plays aren't worth as much at higher levels is cause you can't gain significant acc PP with OD7 and 8, compared to OD10 HR and DT plays + the issue with pattern difficulty.
sayonara_sekai
I've randomly jumped 500-1000 ranks 3 times today. What could be causing this? I get 20-40~ ranks from doing good on hard maps and then randomly sometimes it jumps up around 700 ranks.
Drezi
Actually I ran some numbers, and made a weighting that yielded roughly the same amount of PP for the typical player, tested with actual PP numbers.

I used =(COS(X/12,5)+1)/2 with this weighting the top25 performances would be weighted higher than they are now, below the top25 they would be weighted lower, and reach 0 around the top40th performance. I might have made mistakes when it comes to the actual numbers, but you get the idea.

This would mean that having lots of average performances wouldn't be as valuable, as having good ones in the top spots, and having an outstanding Top1 performance wouldn't be as important as it is now.

Blue line shows the current weighting, Red line is the one suggested.
uzzi

Drezi wrote:

Actually I ran some numbers, and made a weighting that yielded roughly the same amount of PP for the typical player, tested with actual PP numbers.

I used =(COS(X/12,5)+1)/2 with this weighting the top25 performances would be weighted higher than they are now, below the top25 they would be weighted lower, and reach 0 around the top40th performance. I might have made mistakes when it comes to the actual numbers, but you get the idea.

This would mean that having lots of average performances wouldn't be as valuable, as having good ones in the top spots, and having an outstanding Top1 performance wouldn't be as important as it is now.

Blue line shows the current weighting, Red line is the one suggested.
I see this being very good alternative to the current weighting system. Kind of anti-farm as well in a sense.
Amianki
I support that only because it would apply to me perfectly. :*)
silmarilen
i dont like the part where it hits 0
Zare


¯\_(ツ)_/¯
silmarilen
that's a really good score, but what are you trying to gain from posting it here?
Zare
i'm criticizing how overrated DT is
I feel like that should give like


170~ pp
GoldenWolf
What if I tell you you're underrating DT
silmarilen
i feel like it should give ~215 pp in the current system
Zare
I'm getting these scores within 5-10 tries usually
Whenever I get a Nomod FC with these amounts of tries it's worth like nothing
silmarilen
>od
jesse1412

silmarilen wrote:

>od
>ar

The only issue with DT is it gives people an AR they're more comfortable with imo. It'd be fixed up a bit if AR was rewarded/punished accordingly but it's such a subjective topic it's not really balanceable.
Vuelo Eluko

Zare wrote:



¯\_(ツ)_/¯
Spaced streams are pretty difficult but a bit overrated, that i'll agree with. but that map isnt exactly free pp i havent been able to get higher than a 240 combo B.

also i thought it was already established pony maps are all free pp because even slow players can get fcs on them with dt and collect their ~5 star fcs because they all have low bpm compared to a lot of 3-.3.5 star maps which can sometimes get into the 280-290 bpm range.
GhostFrog
All of the above posts are pretty much correct, but to add to that, pp would be more accurate on a per-score basis if it worked like tp imo with regards to how aim speed and acc are treated. Your pp from a score isn't too much different from the sum of its aim speed and acc values (a power mean is used to weight the highest value a bit more), which means that most scores that don't give a decent amount of each are almost worthless. In tp, your overall aim/speed/acc values were calculated by putting all of your aim/speed/acc values in order and applying the same decreasing weighting method as is now employed for your total pp. That allowed scores like a low-acc DT FC on mendes (or impressive FCs on low OD nomod maps) to be properly rewarded, whereas the current system makes them seem relatively insignificant. In exchange, the current system allows you to get more pp from your stronger aspect(s) and that probably balances out total pp, but tp's system was certainly better about rewarding you for individual good scores.
Drezi

silmarilen wrote:

i dont like the part where it hits 0
The main idea was to reward your top2 to 5-10 performances higher as Dexus suggested, and not have the weighting draw out for so long. If it was feasible I would have made it hit 0 at around the 25-30th performamce, as I don't believe that any more performances are needed to evaluate a player, setting around that many good scores already proves that you're capable of playing at that level. But doing so would have resulted in large PP drops in everyone's PP totals, which is something I thought couldn't work due to people freaking out, even if cutting everyone's pp in half does nothing in reality.

I think pretty much everyone can effortlessly set any number of scores that are as good as their 40-50th ones, so why should those even be considered/rewarded? The faster you progress, or less scores you set, the more it's holding you back, or alternatively it can be filled up with your 2-3 try FC level of play performances, which again doesn't really prove anything.

It's mostly just phychological reasons that you feel it's bad if it hits 0 in my opinion, but even then it could always be adjusted so that around 25-35 it turns into the current weighting formula and it never reaches 0.
Dexus
spit-balled some numbers to give a better idea

Current Weighting

Player A
250pp 100% 250
191pp 95% 181.45
190pp 90% 171
190pp 86% 163.4
190pp 81% 153.9
total: 919.75

Player B
210pp 100% 210
203pp 95% 192.85
200pp 90% 180
200pp 86% 172
198pp 81% 160.38
total: 915.23

After Weighting Shift

Player A
250pp 100% 250
191pp 98% 187.18
190pp 97% 184.3
190pp 95% 180.5
190pp 93% 176.7
978.68pp

Player B
210pp 100% 210pp
203pp 98% 198.94pp
200pp 97% 194pp
200pp 95% 190pp
200pp 93% 186pp
978.94pp

As you can see if a player places a random or fluke score they can easily usurp a player who is clearly playing at a more consistent level. With the change it would shift these players closer in ranking without giving fluke scores such power and make placing similar scores within your top 10 more feasible. The player sporting a high pp scores IS deserving of a pronounced rank, but it really shouldn't overthrow players who are more deserving to be at that rank. This wouldn't allow farming neither because it would taper off still. At first I was against putting 0% worth, but the more I think about it the more it makes sense to not even count so many scores. 1% of a score shouldn't really be what a player is focusing on in my opinion, but constantly replacing your top score to get anywhere will just cause such anxiety. I'm honestly interested in seeing where all players would be sitting if such a weighting method was put in place.
Ziggo

GhostFrog wrote:

All of the above posts are pretty much correct, but to add to that, pp would be more accurate on a per-score basis if it worked like tp imo with regards to how aim speed and acc are treated. Your pp from a score isn't too much different from the sum of its aim speed and acc values (a power mean is used to weight the highest value a bit more), which means that most scores that don't give a decent amount of each are almost worthless. In tp, your overall aim/speed/acc values were calculated by putting all of your aim/speed/acc values in order and applying the same decreasing weighting method as is now employed for your total pp. That allowed scores like a low-acc DT FC on mendes (or impressive FCs on low OD nomod maps) to be properly rewarded, whereas the current system makes them seem relatively insignificant. In exchange, the current system allows you to get more pp from your stronger aspect(s) and that probably balances out total pp, but tp's system was certainly better about rewarding you for individual good scores.
I kinda agree with this, but I also see how a score with decent values in all aspects should be worth something. Maybe the tp system could be used with the players pp composed of aim, speed, acc AND combined pp as a forth value.
FGSky






why EZ is underrated than HT

HT is very easy mod
GoldenWolf
It's not?

the EZ score has 20 misses more and 4% worse accuracy, that's pretty huge, especially the 20 misses

To compare the score they need to be near indentical
FGSky

GoldenWolf wrote:

It's not?

the EZ score has 20 misses more and 4% worse accuracy, that's pretty huge, especially the 20 misses

To compare the score they need to be near indentical
yeah but look at the pp amount
silmarilen
look at the amount of misses
is there any other map you know of where you can get over 200pp with 40+ misses?
GoldenWolf

FGSky wrote:

yeah but look at the pp amount
not "yeah", you haven't read anything of what I said...
FGSky
nvm

silmarilen wrote:

look at the amount of misses
is there any other map you know of where you can get over 200pp with 40+ misses?
uuh i thought the combo is more important than misses sorry :-)

GoldenWolf wrote:

not "yeah", you haven't read anything of what I said...
no i don't know how to read
-GN
yeah the pp amount increases exponentially as your misscount approaches zero. i think the map is easier with HT than on EZ myself, but getting into the 800s on any of them is really fucking hard and i've only done it with HT once... i hope to hit 840 or more with it still though.
Loves
Merge the average player accuracy (and their rank) into the map difficulty. It is very hard to tell how hard a map is purely by using an un-self-aware algorithm that renders half the hard maps less desirable to play because they're so f----ing hard in the first place with little to no reward.

Algorithm (50%) + average player rank (10% does not include HT/NF)+ average map score/accuracy/combo (40%).

Simply too many maps that are really difficult are underrated by the ppv2 system and vice versa. A bunch of 4.5-5 star maps that belong in the 5.5+ range is not my definition of fair ranking and skill level.
Mathsma

Loves wrote:

Merge the average player accuracy (and their rank) into the map difficulty. It is very hard to tell how hard a map is purely by using an un-self-aware algorithm that renders half the hard maps less desirable to play because they're so f----ing hard in the first place with little to no reward.

Algorithm (50%) + average player rank (10% does not include HT/NF)+ average map score/accuracy/combo (40%).

Simply too many maps that are really difficult are underrated by the ppv2 system and vice versa. A bunch of 4.5-5 star maps that belong in the 5.5+ range is not my definition of fair ranking and skill level.

No. Tom made ppv2 not reliant on anything outside of the map itself so that popularity and other issues would not cause problems with weighting.
Nyxa
That is a lot of bold. We can read, you don't need to make your post more readable than others because it's not any more special or correct than any other.

Dexus wrote:

a bunch of stuff i entirely agree with
I had a similar idea in mind but didn't know how to put it to practice or come up with a solid explanation of how it would work, so I never posted about it. Glad someone did, this is a really neat idea.

Dexus wrote:

At first I was against putting 0% worth, but the more I think about it the more it makes sense to not even count so many scores. 1% of a score shouldn't really be what a player is focusing on in my opinion, but constantly replacing your top score to get anywhere will just cause such anxiety. I'm honestly interested in seeing where all players would be sitting if such a weighting method was put in place.
I think that if you're using 1%, you might as well use 0%. If I set 100 100pp scores that are all weighted 1%, that's 100pp extra that i don't even deserve. As Drezi said, players should aim to improve their best performances, not set a bunch of average performances that are easier than opening a can of soda to set. I can consistently set OD7 TV Size SS ranks given that they're not too hard, and I don't become a better player from doing so. It's not a real achievement, thus I don't deserve to be rewarded for it at all.

So, yes, I agree with Drezi. Only the top ~40 or so (I'd personally say the top 25 but whatever) should actually be rewarded, and rewarded more handsomely than it is now. And Dexus' application of it seems like a much more fair alternative than what we have now. I recall a player telling me that I shouldn't be trying to get top 5 performances all the time, and rather gain pp in slow increments instead - and I completely disagree with that, there's nothing impressive about average scores in my opinion.

So, yeah. Fully support this idea.
Loves

Mathsma wrote:

Loves wrote:

Merge the average player accuracy (and their rank) into the map difficulty. It is very hard to tell how hard a map is purely by using an un-self-aware algorithm that renders half the hard maps less desirable to play because they're so f----ing hard in the first place with little to no reward.

Algorithm (50%) + average player rank (10% does not include HT/NF)+ average map score/accuracy/combo (40%).

Simply too many maps that are really difficult are underrated by the ppv2 system and vice versa. A bunch of 4.5-5 star maps that belong in the 5.5+ range is not my definition of fair ranking and skill level.

No. Tom made ppv2 not reliant on anything outside of the map itself so that popularity and other issues would not cause problems with weighting.
So you agree that Scarlet Rose should stay 5.19 stars? This is just one out of hundreds of examples as to why the algorithm cannot weight a map correctly if it's mapped in a way which flies under the radar on the "difficulty" rating. And I don't get how popularity would affect a maps difficulty if it excluded NF/HT scores. PLease explain.
Vuelo Eluko

Loves wrote:

I don't get how popularity would affect a maps difficulty if it excluded NF/HT scores. PLease explain.
really? there's maps that could have 50 hd/hr or DT 98%+ on the scoreboard but since they have like 2-5k scores theres a lot of HD and sometimes even nomods on the board.
Mathsma

Loves wrote:

So you agree that Scarlet Rose should stay 5.19 stars? This is just one out of hundreds of examples as to why the algorithm cannot weight a map correctly if it's mapped in a way which flies under the radar on the "difficulty" rating. And I don't get how popularity would affect a maps difficulty if it excluded NF/HT scores. PLease explain.
The issue with Scarlet Rose is that the system cannot recognize pattern difficulties at the moment. Tom wants to create one that can recognize pattern difficulty and I don't think he would use player averages to do that simply because of the popularity issue. How could a song be affected due to popularity even if it excluded score reducing mods? Simple, low HP value. If it is low enough then anyone could pass the map thereby skewing the average.
Loves

Mathsma wrote:

Loves wrote:

So you agree that Scarlet Rose should stay 5.19 stars? This is just one out of hundreds of examples as to why the algorithm cannot weight a map correctly if it's mapped in a way which flies under the radar on the "difficulty" rating. And I don't get how popularity would affect a maps difficulty if it excluded NF/HT scores. PLease explain.
The issue with Scarlet Rose is that the system cannot recognize pattern difficulties at the moment. Tom wants to create one that can recognize pattern difficulty and I don't think he would use player averages to do that simply because of the popularity issue. How could a song be affected due to popularity even if it excluded score reducing mods? Simple, low HP value. If it is low enough then anyone could pass the map thereby skewing the average.
Hes already said its not possible to recognize awkward map patterns just yet. And with the hp thing, tom could just implement a balanced scale of hp and acc so that itd be more accurate.
Nyxa
You're looking at things that aren't issues and ignoring the things that are.
Dexus
Hp drain would be irrelevant to acc. If you actually played to a high enough level you would know that drain seems almost non existent.
Loves

Tess wrote:

You're looking at things that aren't issues and ignoring the things that are.
I was just saying a suggestion in regards to your HP drain thing. Dexus is right, HP drain is pretty much non existent for good players. I don't see how using a census of actual information is detrimental to ppv2 since that actual information is probably just as accurate as the algorithm on higher level maps (5+).
GhostFrog
ppv1 was based mostly on comparing to other players' scores and that resulted in a ton of problems. ppv2 is generally more accurate because it judges plays on their own merits - whether or not map A is more difficult to FC than map B doesn't change when another player FCs it.

Anyone who's been following ppv2, whether they started following it with this thread or with the tp system it was based on, is well-aware of its obvious issues and reading/pattern difficulty may very well be the biggest/most obvious. Finding a decent solution for that has proven quite difficult for tom, who's also busy working out the pp systems for 3 other game modes (all of which are still relatively young and all of which have more problems than osu!standard) in addition to having a life outside osu!. I think he's done a really good job with map difficulty overall and I'd rather wait and see if he can come up with some magic to fix the outliers than revert back to a system that ppv1 showed wasn't accurate.
manjumochi

GhostFrog wrote:

ppv1 was based mostly on comparing to other players' scores and that resulted in a ton of problems. ppv2 is generally more accurate because it judges plays on their own merits - whether or not map A is more difficult to FC than map B doesn't change when another player FCs it.

Anyone who's been following ppv2, whether they started following it with this thread or with the tp system it was based on, is well-aware of its obvious issues and reading/pattern difficulty may very well be the biggest/most obvious. Finding a decent solution for that has proven quite difficult for tom, who's also busy working out the pp systems for 3 other game modes (all of which are still relatively young and all of which have more problems than osu!standard) in addition to having a life outside osu!. I think he's done a really good job with map difficulty overall and I'd rather wait and see if he can come up with some magic to fix the outliers than revert back to a system that ppv1 showed wasn't accurate.
The bold part of quote is true, pp in Std is way more accurate than other modes (You can easily farm "pp" in other modes, but is kind of hard to farm pp in Std if you don't have the skill).
show more
Please sign in to reply.

New reply