http://ask.fm/Tom94/answer/110921844174 This question made me wonder if getting a bit lower acc is worth it. Obv, 93% is not a good score, but what if I go from a high 99.5%+ HD score to a 97.5-98.5% HDHR score?
http://ask.fm/Tom94/answer/110921844174
K,I understand that. But this score is all about aim and hitting the jumps too early shouldn't affect that (+I got HD). Seeing that you have no information about the 'type' of 100, does the system interpret my aim as too sloppy/bad? If yes, will this be fixed?
It interprets your accuracy as being bad, which results in the low pp. Your aim component and speed component of the score was barely affected. It was simply not considered a good performance, because your accuracy was bad. There's a difference of light-years between 93.5% and 99%, even on lower ODs.
http://i.imgur.com/EEVGCF4.png
My lower accuracy comes from me hitting the notes faster. How/Why did you get idea that a faster score should be treated worse? Please help me understand.
Because speed isn't everything? Try to stop hitting things too early. :P Also even if I wanted to change this I couldn't. There only is information about 100 count, not about how and where they were made.
Is there no chance to also display the aim/speed/accuracy pp of each score, too (and maybe the total pp for each category in the profiles, like in tp)? I know those values aren't as relevant as they are in tp, because the total pp isn't just their sum, but it'd be nice to know.
They are not stored currently, so they also can't be displayed. Storing things per-score is very expensive since there are billions of scores, literally. Might change at some point in the future.
tom has it ever pass your mind to put a fixed value on hidden bonus like 1% from no mods hidden is the most easy mod there is but it makes a little hard for a song so 1 % it like almost nothing what do you think of this? even if you dont like it i would like you to clarify me on why not or yes thank
Hidden already gives an 18% bonus to aim. That's definitely enough. If you're not talking about standard, then ask again and specify that.
Those really were fun times. But hey, don't make it sound like I ruined the servers. I provided free hack-protection software to private servers and even the official ones at times. :p
have you ever used the logitech g400s before? what do you think about mouse players now that you've experienced it for yourself?
I've used a logitech mouse very similar to the g400s and it was nice to play with. Prefering the DeathAdder and the Rival, though. And well, mouse is fun, and it's possible to play it on a very high level. I'd say my skill in general is about the same with tablet and mouse, even though I'm a lot more consistent (less random misses) with the tablet. The lack of drift also helps a lot there.
Have you ever thought of giving map-specific bonuses to account for algorithm imperfections? Like giving Scarlet Rose a TP bonus? Not suggesting it, just curious of what you think.
I did think of it and I came to the conclusion, that it is a bad idea. It introduces another layer of subjectivity which is not always well founded. In addition to that in increases maintenance cost (in terms of time) by a lot, since you will always have to tune specific maps in addition to the algorithm.
osu!mania and taiko don't have hidden factored in yet, so HD+DT pretty much equals DT in these modes. In other modes HD+DT is always superior given all other parameters of the score are the same.
Have you considered holding a contest for the community to try and come up with a decent calculation for the reading difficulty of a map? Whatever you have done for aim/speed/accuracy have been great but maybe something decent can come from a lot of people trying to find a solution.
That would encourage the people who have no idea what they're talking about to post their idea just for the sake of the contest / potential prizes. I prefer it how it is right now, since I'm getting a lot of valuable feedback. More than I can properly evaluate and test in fact, due to time limitations.
Looking at the bottom of my performance list, I was given ~8pp for some Easy/Normal FCs, but only like 3pp for a pass on a map like Airman. Is this really fair? At the very least, the aim and speed required for a pass on Airman should be much more than some Normal FC, right?
Passing with a huge amount of misses and a horrible accuracy is _not_ a good performance which the system tries to reward. The base requirement for gaining pp is doing well, no matter how hard the map is, so you should only expect to get pp from maps you can perform decently on. :P
Theoretically speaking, could a total noob use the weighting system to become rank 1? They make their own pp calculator and only achieve scores on a new account that will give EXACTLY 100 pp. They get 81 of these scores, and since all their scores are tied #1 it is all weighted 100% -> 8100 pp?!
Even if the scores are tied they'd still get the same decreasing weights applied, just in a random order (the order doesn't change the outcome, so it doesn't matter). Secondly, since internally pp scores are floating point numbers you most likely wouldn't be able to achieve even a single tie. :P
D: This update is leading to farmers again, what pushed this idea of showing the PP per score?
The previously shown best performances were already all that "farmers" needed. The system aims to be more open so there can be a better understanding by the community and thus better feedback, resulting in a better and more "farm"-proof algorithm. As shown by tp this is not as trivial to "farm" as one might think.
I'm the same person who posted a pastebin containing aim suggestions a few weeks back and I stopped thinking about it for a while. I'm back with a slightly simpler approach that combines snapping and flow movement into a single framework that's easier to use: http://pastebin.com/raw.php?i=aj04iGU3
I believe F and v3 can't be decoupled in different functions, because treating them individually would flaw the system. Reason are (completely subjectively, but empirically correct) for instance squares (90° angles inbetween) VS line jumps (180° angles inbetween). What mostly makes patterns hard is for one part the reading aspect (which is a complete mystery to me) and for the other part the flow vs. snap part you treat in your description. From my personal experience the hardest patterns to execute are the ones that encourage moving the cursor in circles. Starting from squares (being relatively easy still) over pentagons to hexagons and further. At some point the more circlish a pattern becomes the easier it gets again. Ellipse-like patterns make things even harder and the hardest possible combination is having angular snappable patterns heavily combined with circlish / ellipsish patterns. Forgive me for "circlish" and "ellipsish". :POn a side note your stuff is very interesting to read, but it's very late here and I can't think about it too much atm. It's definitely a better approximation to movement than the straitforward velocity based one I'm currently using. I'd merge your functions f and h into one function taking both F and v3 as arguments and also evaluating angles between the movements according to what I described above.