1. osu! forums
  2. osu!
  3. Gameplay & Rankings
show more
posted

abraker wrote:

Edit: I guess food for thought - Is a map with a largely uniform difficulty easier or harder than a map with a large difficulty spike?
That largely depends. There are 2 ways ways I can think of to rate such maps correctly:
  • 1) Separate ratings by how difficult it is to get the letter ranking
    2) Average the maps' difficulty peaks based on a weighting system pp uses.


I have already tried doing it the 2nd way, and yielded results similar to this:
Letter rankings are irrelevant. You can get an A with non-trivial misses on most maps if you're an accurate player. Evaluation would likely have to be done piecewise with a nonlinear weighting for difficulty if we are to work under the assumption that uniform difficulty and easier difficulty with spikes are actually of 2 different difficulties despite mapping to the same average difficulty neighborhood.
posted
I don't know how players do it. The hardest map on there that I can fc is only lagomorphic v_v
posted

chainpullz wrote:

Letter rankings are irrelevant.
Score based then?
posted

abraker wrote:

chainpullz wrote:

Letter rankings are irrelevant.
Score based then?
Score isn't uniform across all maps.
posted

chainpullz wrote:

Score isn't uniform across all maps.
Then it will have to be measured in terms of how hard it is to get to point t, where t is a point in time on the map. That means that my second approach by averaging weights is more reasonable.
posted

abraker wrote:

chainpullz wrote:

Score isn't uniform across all maps.
Then it will have to be measured in terms of how hard it is to get to point t, where t is a point in time on the map. That means that my second approach by averaging weights is more reasonable.
Let A be a long easy to aim section, and B a short difficulty spike. Is it just as easy to "get to the end of" AB as it is BA? Personally I think there is a subtle difference even in this simple unrealistic example. Better yet, consider ABC vs BAD where C and D are just arbitrary song endings and we are only talking about "getting to the end of" the A/B combinations. There are a lot of psychological triggers that are hard to really account for since they vary from person to person.
posted

chainpullz wrote:

There are a lot of psychological triggers that are hard to really account for since they vary from person to person.
Quite right, in fact difficulty only accounts a perfect play. That is, it doesn't account random chokes. While I agree that method doesn't provide the complete picture, it's still better than what we can come up with, and be all means please come up with something better. The difficulty of a map is just an index, and that index can be interpreted as a probability to do poorly on a map. However, that conversion from difficulty rating to chance of doing poorly is very subjective and up to the player to decide. Difficulty makes a lot more sense when viewed from a collective of players instead of the individual. When viewed from a collective, we can universally agree on stuff such as which map is harder (this thread's point).
posted
This is redundant. You can't measure the 'pure aim' value of maps because terms like 'aim' and 'reading' and 'speed' are all vague concepts, interconnected and without any clear definitions. If you want to measure the aiming difficult of maps minus 'reading' difficult and difficulty from hard patterns then all you need to do is look at the star rating.
posted

B1rd wrote:

This is redundant. You can't measure the 'pure aim' value of maps because terms like 'aim' and 'reading' and 'speed' are all vague concepts, interconnected and without any clear definitions. If you want to measure the aiming difficult of maps minus 'reading' difficult and difficulty from hard patterns then all you need to do is look at the star rating.
Not quite. Star rating also takes streams and everything tap related too.
Correct me in case I didn't understand the point of your post right
posted

Kert wrote:

Not quite. Star rating also takes streams and everything tap related too.
Correct me in case I didn't understand the point of your post right
Then just delete the streams. It doesn't take a genius to figure out if the star rating comes from jumps or streams.
posted

Kert wrote:

B1rd wrote:

This is redundant. You can't measure the 'pure aim' value of maps because terms like 'aim' and 'reading' and 'speed' are all vague concepts, interconnected and without any clear definitions. If you want to measure the aiming difficult of maps minus 'reading' difficult and difficulty from hard patterns then all you need to do is look at the star rating.
Not quite. Star rating also takes streams and everything tap related too.
Correct me in case I didn't understand the point of your post right
That and it does not take angles into account either (compare 400bpm square jumps with the circles touching each other to a 200bpm stream)
posted

B1rd wrote:

Then just delete the streams. It doesn't take a genius to figure out if the star rating comes from jumps or streams.
You can't tell if a heavy only-aim map is harder at aim than mediocre-aim-stream one by looking at star ratings

Momiji wrote:

That and it does not take angles into account either (compare 400bpm square jumps with the circles touching each other to a 200bpm stream)
The angle influence is considered to some extent in the last calculation I made btw.
posted
Honestly, if this is an attempt to improve PP, I don't think it's gonna work. The best approach would likely be a machine learning based approach. We'd need to learn a function that calculates PP by beatmap.
posted
felicitousname: It's not about improving pp algoritm. Kert is making an alternative way of ranking players skills. Go there: osuskills.tk

In my opinion, to mark aim the algoritm would need to check:
-basic things, like CS, BPM;
-in map placement of elements;
-in map timings.
-and it would need to know player reading skills.
Rest of things is unneeded.

I'm not a guy who can make complicated formulas, so I won't try come up with one. But for what is needed to make one... Well, it's here, right?
posted
This thread is old and done for and you should feel bad
Please sign in to reply.