agreed that FL is seriously underrated, but we have to take in account fake FL players using two screens
CookChefSteak wrote:
The new system feels really accurate overall, although it does have several flaws as people mentioned here. I will try to list some of them:
1. Accuracy is overrated - I've noticed a lot of times that clearing really hard maps with accuracy lower than ~95% but with a pretty good score will not give you any performance points. For example, over 15m points on Adult's Toy(which is not easy to get a good score on) but with 88% accuracy will be counted as a bad play, and the same goes for getting over 20m on Dragonforce maps with less than 90% accuracy and similar other situations. It's like the algorithm automatically ignores every play under 90% which is just silly, imho.
2. Penalties - "There are no penalties for a bad performance" yeah right.. If you improve your score on a map(even if by a lot) but have like 2% less in terms of accuracy you WILL lose pp if you got pp for the previous score(accuracy overrated, anyone?). How stupid is that? I can recall a few times I lost between 5 to 10 pp after improving a score(from 3-4 mistakes to fc) but having 2-4% less. PP should not be reducible, especially if those said plays happen within 2 hours from each other. Not to mention this system heavily discourages trying to improve scores and FCing on maps that gave you pp because of the fear of losing your points if you don't manage to get the same accuracy or higher. If the new system has no problems giving you a second chance when you randomly miss and break your FC, why does it penalize you so hard if you randomly mess up a stream(happens to everyone) and get a few 50s?
3. FL is severely underrated - Just look at BluOxy and worst fl player. Need I say more?
Suggestions:
1. Give more weight to FL especially when it's FLDT or FLHD, it's a really hard mod
2. Weigh performance on maps using other things than accuracy. What I mean by that is don't automatically toss a play just because it's not a 95% accuracy play. I don't know if you're working on it already or something like that, but some maps are really hard to keep a combo on, and in those maps you should give a lot more weight to combo and less to accuracy. Combo:Accuracy rating ratio should NOT be global as it feels right now. It should change with each map.
3. Disable pp penalties. There is really no sense in reducing someone's pp because his accuracy got slightly lower, since if he managed to get his accuracy to make such a difference, there is no doubt he'll be able to do it again. And if you call on the "lucky play" card, well I can just as validly pull the "unlucky play" card, so it's really meaningless.
1. If a certain map requires X amount of skills to get a decent score at and that X correlates to a certain pp range, then should pp gain not be expected when you are below that pp range and acquire such a score?Tom94 wrote:
CookChefSteak wrote:
The new system feels really accurate overall, although it does have several flaws as people mentioned here. I will try to list some of them:
1. Accuracy is overrated - I've noticed a lot of times that clearing really hard maps with accuracy lower than ~95% but with a pretty good score will not give you any performance points. For example, over 15m points on Adult's Toy(which is not easy to get a good score on) but with 88% accuracy will be counted as a bad play, and the same goes for getting over 20m on Dragonforce maps with less than 90% accuracy and similar other situations. It's like the algorithm automatically ignores every play under 90% which is just silly, imho.
2. Penalties - "There are no penalties for a bad performance" yeah right.. If you improve your score on a map(even if by a lot) but have like 2% less in terms of accuracy you WILL lose pp if you got pp for the previous score(accuracy overrated, anyone?). How stupid is that? I can recall a few times I lost between 5 to 10 pp after improving a score(from 3-4 mistakes to fc) but having 2-4% less. PP should not be reducible, especially if those said plays happen within 2 hours from each other. Not to mention this system heavily discourages trying to improve scores and FCing on maps that gave you pp because of the fear of losing your points if you don't manage to get the same accuracy or higher. If the new system has no problems giving you a second chance when you randomly miss and break your FC, why does it penalize you so hard if you randomly mess up a stream(happens to everyone) and get a few 50s?
3. FL is severely underrated - Just look at BluOxy and worst fl player. Need I say more?
Suggestions:
1. Give more weight to FL especially when it's FLDT or FLHD, it's a really hard mod
2. Weigh performance on maps using other things than accuracy. What I mean by that is don't automatically toss a play just because it's not a 95% accuracy play. I don't know if you're working on it already or something like that, but some maps are really hard to keep a combo on, and in those maps you should give a lot more weight to combo and less to accuracy. Combo:Accuracy rating ratio should NOT be global as it feels right now. It should change with each map.
3. Disable pp penalties. There is really no sense in reducing someone's pp because his accuracy got slightly lower, since if he managed to get his accuracy to make such a difference, there is no doubt he'll be able to do it again. And if you call on the "lucky play" card, well I can just as validly pull the "unlucky play" card, so it's really meaningless.
1. See point 2 below. I also want to remark, that the score itself, that you get on a particular play is not relevant. Your combo and your miss/50/100/300 are what is important. I also want to note, that having far-below FC is heavily punished by the system. You shouldn't expect to gain a lot of pp from maps that you can't get decent scores on just by them being hard (see adult's toy, dragonforce and such).
2. see below
3. see below
Your suggestions:
1. Is planned, will likely happen.
2. Accuracy is worth less than a third of your play in the general case. Could you provide specific cases where you believe it is overrated? Otherwise I'll have a hard time debugging. By the way, accuracy doesn't simply give you a factor. You can theoretically get 50% accuracy and if you FC a hard map you can still get a very huge amount of points. I'm not quite sure how to feel about this, since many other people claim accuracy to be worth too little. Combo:Accuracy never was global and it does scale (a lot) with maps.
3. This is not something that can be just "disabled". It's also not specific to pp. The scoring system simply works by only storing your highest score on a particular map. I'd love to see this limitation get lifted at some point, but even only having the highest scores is already an extremely huge amount of data to work with.
1. Of course. This is not exactly how the algorithm works but should yield a sufficient approximation. That being said, correctly judging X is the problem here.CookChefSteak wrote:
1. If a certain map requires X amount of skills to get a decent score at and that X correlates to a certain pp range, then should pp gain not be expected when you are below that pp range and acquire such a score?
2. As far as I've noticed, accuracy is weighed too heavily on any map below 1.5 minutes(or >1 minute with DT). Try looking at maps like Eggman's Theme on DT. I lost around 6 pp(out of overall 8) because I improved a score but had 2% less. Note the score I improved had 3-4 mistakes and the improvement was an FC, so it really makes no sense to me. That map shouldn't give that much pp anyways imho, it's too easy to fc on DT.
3. Wouldn't checking whether or not the newly calculated pp would be less or more than the old one be enough? I mean, the servers already store everybody's pp so you have all the data you need available for comparison. But then again, I don't know how the scoring system is programmed so I don't want to assume things.
this is the mapSaberBB wrote:
I want to ask that why I played a hard level map
with HD,HR, full combo
but no pp is added?
I played the map 32mins ago
OK....Tom94 wrote:
If you don't get pp then the system doesn't consider your score good enough. You can also get less than 1pp and not directly see it because of rounding. I won't be answering any "Why didn't get pp for X?" questions anymore for now.
Tom94 wrote:
1. Of course. This is not exactly how the algorithm works but should yield a sufficient approximation. That being said, correctly judging X is the problem here.CookChefSteak wrote:
1. If a certain map requires X amount of skills to get a decent score at and that X correlates to a certain pp range, then should pp gain not be expected when you are below that pp range and acquire such a score?
2. As far as I've noticed, accuracy is weighed too heavily on any map below 1.5 minutes(or >1 minute with DT). Try looking at maps like Eggman's Theme on DT. I lost around 6 pp(out of overall 8) because I improved a score but had 2% less. Note the score I improved had 3-4 mistakes and the improvement was an FC, so it really makes no sense to me. That map shouldn't give that much pp anyways imho, it's too easy to fc on DT.
3. Wouldn't checking whether or not the newly calculated pp would be less or more than the old one be enough? I mean, the servers already store everybody's pp so you have all the data you need available for comparison. But then again, I don't know how the scoring system is programmed so I don't want to assume things.
2. "2%" accuracy is usually a huge margin. Most people see "accuracy goes from 0% to 100%, therefore 2% is not a big difference" which is a completely wrong assumption. First of all, it greatly depends on where you've been previously. 96% and 98% are a 2% difference, but you have twice as many 100s in a 96% score than in a 98% score. That's a 100% relative difference! In your case you went down from around 95% to 92%. While not doubling your 100 count it still increased by a lot. Judging by how short the map is, you shouldn't expect to get too much of a bump for full-comboing it. Imho the decrease of pp is justified on that map.
3. No. The system needs to sort all your current scores to determine your user-pp, so it needs to be able to access your highest pp on a per-map basis for what you suggest to work. That'd force us to store 2 high-scores per map per player. Another problem is, that the pp algorithm will be tweaked frequently, so plays that previously were not deemed good enough and discarded might end up deserving the top spot in the future.
That works only for the first new score coming in. Imagine a new (higher) pp value gets chosen and updated while the score which caused the pp increase doesn't get saved, since there is a higher score lying around already.CookChefSteak wrote:
But since every player's current pp is stored as an integer on the server, why is it not possible or plausible to store newly calculated pp as a temporary variable before calling on the server's pp-update method, and compare it with the stored[old] pp value? It looks like a simple if statement in my head. It just seems like something like that should be very possible in a well-modulated algorithm. I'll say it again though, I don't know how you programmed it so it might not be that simple. Maybe if you could shed some light on the technical aspects of the algorithm I'll understand better.
Correct, the old score is lost, but the old PP value(the raw floating point number) doesn't have to be. If it's stored temporarily as just a number, you could regularly calculate a player's pp after he sets a highscore then compare it to that temporarily stored raw value. That is, store it before you update a player's pp, compare it, update with the appropriate value, and delete the temporary memory segment. There is no need to recalculate the old pp value again using 2 highscores per map as you said. It's a really quick and simple algorithm, and I honestly do not see the problem in implementing that. You must know something regarding the algorithm that I don't if you're still not convinced.Tom94 wrote:
That works only for the first new score coming in. Imagine a new (higher) pp value gets chosen and updated while the score which caused the pp increase doesn't get saved, since there is a higher score lying around already.CookChefSteak wrote:
But since every player's current pp is stored as an integer on the server, why is it not possible or plausible to store newly calculated pp as a temporary variable before calling on the server's pp-update method, and compare it with the stored[old] pp value? It looks like a simple if statement in my head. It just seems like something like that should be very possible in a well-modulated algorithm. I'll say it again though, I don't know how you programmed it so it might not be that simple. Maybe if you could shed some light on the technical aspects of the algorithm I'll understand better.
Now the same scenario happens again: The newly calculated pp value only features the one new score + all old ones in the database. The score used previously to increase the pp is completely lost!
Also, just to clarify: pp is a floating point number. Even if the number you see doesn't change with a score you get, you still might have gotten a fraction of a pp.
You're wrong, this doesn't work. The old floating point value that's still lying around is useless for computing what pp the user would have with the new score. The previously discarded score would be necessary for that computation.CookChefSteak wrote:
Correct, the old score is lost, but the old PP value(the raw floating point number) doesn't have to be. If it's stored temporarily as just a number, you could regularly calculate a player's pp after he sets a highscore then compare it to that temporarily stored raw value. That is, store it before you update a player's pp, compare it, update with the appropriate value, and delete the temporary memory segment. There is no need to recalculate the old pp value again using 2 highscores per map as you said. It's a really quick and simple algorithm, and I honestly do not see the problem in implementing that. You must know something regarding the algorithm that I don't if you're still not convinced.
anyways an example:
1. a highscore is achieved
2. current player pp is 2280
3. 2280 stored as float = current_pp
3. pp method runs with the highscore as an argument
4. algorithm yields 2270 as the appropriate pp considering the new score(float newpp = ppcalculate(args);)
5. newpp > current_pp ? server.query(newpp) : server.query(current_pp);
that's it...
In most cases, when you barely pass a map you aren't really demonstrating that you can play it competently. What you're typically showing is that you're able to bullshit your way through it. Here's an example: http://osu.ppy.sh/s/34544[ Zetka ] wrote:
Kind of enjoying a nice consistent rise in ranks but what I'm not so sure about is how I can spend ages working on a really hard beatmap and when I finally complete it it doesn't count anything towards my pp at all. But if I go and find the easiest beatmap I have, wack on a few mods and get a good score really easily, that gives me loads of pp. Don't quite understand that at all :/
usually when a score gives you nothing when it should you just have to set a new best score on any map for it to update.-Soba- wrote:
Why did remote control hdhr give me 0 pp but is really high on my best performances ?
Did your pp go down? Or just your rank while pp stayed the same?asior wrote:
I just dont get it. Always the same..
I play song 1st time, i failed 2-3 times, im.. lets say #2000 (B)
And my ranking increased for 50
I play song 2nd time on FC, im lets say #700 (C)
And my ranking DECREASED for 20..
Why does that happened ?
So pp can go down ? If yes; when ?scottyyy wrote:
Did your pp go down? Or just your rank while pp stayed the same?asior wrote:
I just dont get it. Always the same..
I play song 1st time, i failed 2-3 times, im.. lets say #2000 (B)
And my ranking increased for 50
I play song 2nd time on FC, im lets say #700 (C)
And my ranking DECREASED for 20..
Why does that happened ?
It's a limitation of the current system but it's not bad really. The system is still rewarding better scores over worse ones. Just because you value crappy acc FC over good acc non-FC doesn't mean the system is wrong. :pasior wrote:
Wow..
So play on easy, get better acc (but worst score) and gain more pp than play without mods but with fc and tragic acc. seems legit...
Can't tell if serious on the first part, but the reason why maps like Scarlet Rose and Adult's Toy are weighted so low is because the ranking system cannot value patterns as of right now. Tom said that he would try to implement it later if he can.dennischan wrote:
People said that Scarlet Rose was the hardest map in existence (according to forums)
Why is it ranked so low in tp?
Is this map underrated?
Adult's toy is also seriously underrated. You can ask almost everyone and find that the consider Adult's Toy harder than say, some random high tp diff such as recommended spell (Spell)
I suggest maybe high bpm songs should rated higher.
The easy map won't give you anything because it's... easy? pp is completely based on map difficulty.asior wrote:
Wow..
So play on easy, get better acc (but worst score) and gain more pp than play without mods but with fc and tragic acc. seems legit...
Go try it.asior wrote:
Wow..
So play on easy, get better acc (but worst score) and gain more pp than play without mods but with fc and tragic acc. seems legit...
Tom94 wrote:
The easy map won't give you anything because it's... easy? pp is completely based on map difficulty.asior wrote:
Wow..
So play on easy, get better acc (but worst score) and gain more pp than play without mods but with fc and tragic acc. seems legit...
Ekaru wrote:
Go try it.asior wrote:
Wow..
So play on easy, get better acc (but worst score) and gain more pp than play without mods but with fc and tragic acc. seems legit...
"Wow, I just wasted my time" in 3... 2...
the scarlet rose being the hardest map is a joke on the forums lol, it's not really that hard (only hard to fc, not to get high acc)dennischan wrote:
People said that Scarlet Rose was the hardest map in existence (according to forums)
Why is it ranked so low in tp?
Is this map underrated?
Adult's toy is also seriously underrated. You can ask almost everyone and find that the consider Adult's Toy harder than say, some random high tp diff such as recommended spell (Spell)
I suggest maybe high bpm songs should rated higher.
That mod makes most Insanes a lot harder to FC while killing your score modifier. >_>asior wrote:
I mean mod, not map type.
EZ mod gives you huge circles, which makes aim trivial and lowers the score you get from aim. It halves the OD, which makes accuracy trivial and lowers the score you get from accuracy. If you have a short stream, the lower OD will significantly reduce the speed required and lower the score you get from speed.asior wrote:
I mean mod, not map type.
Thanks for pointing that out. I'm pretty damn sure, that I already fixed this at some point - the website even shows the correct multiplier on the info page. Well, gotta fix it again.Ziggo wrote:
I found a mistake in the calculations in osutp and I'm wondering if it's the same for the current pp system. When adding the Half Time mod the speed gets multiplied by 0.75, so AR and OD times need to be multiplied by 4/3. The beatmap difficulty on osutp shows different values, though. E.g. ar10 with Half Time becomes ar8.5 instead of ar9 (for correct values check GhostFrog's post in https://osu.ppy.sh/forum/p/2858736). Same issue with OD. I don't know if even the aim and speed values are affected by this, but it might be a good idea to check it out.