My score in http://osu.ppy.sh/s/12186 with HR has 2acc with 87.74%acc on OD9.8, don't you think the algorithm subestimates how hard is to get accuracy with high OD when accuracy percentage is low? For example, getting 94%acc on an OD7 map tends to give more accuracy score than 87.74%acc OD9.8.
The amount of sliders / spinners also plays a big role. Apart from that pp/tp is supposed to reward a good performance, which 87.74% surely is not. It does not reward mediocre performance on maps you can barely play in a particular aspect.
Because due to algorithm limitations large stream spacing gets thrown into speed - I know it's not the best solution and I'm having a few ideas how I could tackle this problem. For now it'll stay like this. You can also see a stronger example of this in thelewa's "made of fire" score.
I think that the current system values fluke plays too much. Rank currently is determined mainly by performances that are well above the average performance of the player. What about if instead of weighting the very best plays, it weights the median plays more?
Because most people (I know of) define skill by the peak of what someone is capable of doing. It's not like in the olympics they take average performances either.
In the case there are 2 maps that are exactly the same, but one has an extra circle one measure after the previous note at the very end. If a player gets SS in one of these maps, and in the other gets SS until the last circle, which is a miss. In the second map, what aspects give less points? How?
The 1-miss run would give less pp. While the calculated difficulty of the map with the extra circle would _always_ be higher, the performance difference of an SS vs 1-miss more than makes up for it.
I have some preliminary thoughts on possible improvements to aim that I can make more detailed if you're interested: http://pastebin.com/raw.php?i=bh5vmW99
I have thought about very similar things regarding snap vs flow already - actually I have already tried some variations, not very successfully - and I'd love to experiment around with it more, but currently my time is focused on the other modes. I might get back to this at some point in the future, thanks for the text! :)
Does SS matter in new system (apart from obvious difference between 99.9% and 100% accuracy)? In previous system PP was multiplied by 30%, if I remember correctly.
It was far less than 30%. And no, there is no specific bonus for SS anymore.
http://osutp.net/scores?bid=32610 This map contains no circles, only sliders. But I think there should be at least some acc points. If a player lifted sliders too early, the sliders would give him 100. No acc points at all doesn't make sense.
It would mess up accuracy calculations on maps with mostly hitcircles due to the way accuracy is reverse engineered from mere 300, 100, 50 and miss count. Sadly this is not possibly without making things worse in general. :/
Seems like tp considered it as if it was nomod. That's weird... will probably not look into it very soon, though. I prefer working on the official system to make it surpass tp at some point. Also the daily update might fix this. :P
OMG MY PP DROPD BY TRNNI REPOT U UR SYSTEM IZ SHEET OSU IS SRSBSNS DON'T FUCK WITH MY RANK OMG I SUDOKU
I'm having a hard time not stealing you waifu for shnuggles, she's too cute I:
So you used my idea and you didn't credit me ≧Д≦ (about making scores worth more if they excel in one aspect by (Speed^2+Aim^2+Acc^2)^(1/2), you just changed 2 by another number).
I demand you credit anonymous for this.
Because surely nobody in the world ever thought about this exceedingly complicated formula before. :P
When do you think the recalculation of PP will end? if it hasn't already?
I still haven't seen a change in the 600 pp that I gained so does that mean that it is there to stay?
It will still take some time. If you want to know for sure, then make a new highscore on _any_ beatmap.
Part 2:
Also, as long as I get a full combo I should get the maximum aim score. Wether I get a 50, 100 or 300 solely depends on accuracy. (This doesn't apply when I always hit the jumps too late, which would mean my aim is too slow)
The same argument that I just gave towards streams applies to jumps.
Part 1:
Does the Accuracy of my score affect my aim and speed values? E.g. I get 87% on a streamy 180 BPM map with many jumps. Let's say I always hit the streams too FAST. That shouldn't affect my speed value in a negative way, should it?
You could start the stream too early and end too late. Or you could simply get a low accuracy because you can't stream very stable at that speed. Therefore accuracy does affect speed slightly (a really small amount).
Have you considered this for when high score != high performance? For each score, calculate pp from it, if a new play appears for the map, calculate it's pp. If it is better pp but lower score, store in database the data of the play as an extra entry, if later that entry is not needed, delete it.
That'd conserve as much space as possible but it would slow things down for re-calculations compared to just storing the highest pp score for _all_ maps. This is because with your method for every score you would need 2 lookups.
Does you current rank have any effect on the amount of PP you gain? Or it only considers the difficulty of the map? For example, if someone that was rank ~10,000 had a high acc DT score on an insane, would they gain more PP than someone that did the same but was rank ~100 (because it would be harder
The amount of pp you get depends on the amount of pp you already have, so while your rank is not directly influencing the amount of pp you get it is indirectly.
Does spun out affect PP? I mean, completing a spinner doesn't take any skill at all and it's only 1x300.
Spunout multiplies the pp you get by 0.95. While spinners might be trivial they still are part of the game. Also there are exceptions, just look at augoeides.