forum

Show hit error histogram in the hovering statistic

posted
Total Posts
11
This is a feature request. Feature requests can be voted up by supporters.
Current Priority: +4
Topic Starter
chinmeitou
Instead of just showing the minus to positive average hit error, what about showing the hit error histogram?
abraker
Why stop there? A histogram is not as cool as what stepmania has:

Topic Starter
chinmeitou

abraker wrote:

Why stop there? A histogram is not as cool as what stepmania has:


A histogram would look neater though.
Topic Starter
chinmeitou

abraker wrote:

Why stop there? A histogram is not as cool as what stepmania has:

I mean this could be implemented in map editor for better timing, but for rank score information, histogram would be sufficient enough.
abraker
I have a histogram here on the right. It tells me almost nothing of value but that the replay is not a bot

Topic Starter
chinmeitou

abraker wrote:

I have a histogram here on the right. It tells me almost nothing of value but that the replay is not a bot

The original -xx ms ~ +xx ms also tells nothing, so consider removing that?
Full Tablet
A histogram wouldn't give much more additional information, considering the table of judgment counts the result screen already gives.

The hit error v/s time graph gives more information. But, the linear scale used makes it hard to discern errors near 0ms (too many points clustered there); what about using a scale like Ln(1+Abs(x))*(x/Abs(x))?

Example

Linear Scale

Ln(1+Abs(x))*(x/Abs(x)) Scale
Topic Starter
chinmeitou

Full Tablet wrote:

A histogram wouldn't give much more additional information, considering the table of judgment counts the result screen already gives.

The hit error v/s time graph gives more information. But, the linear scale used makes it hard to discern errors near 0ms (too many points clustered there); what about using a scale like Ln(1+Abs(x))*(x/Abs(x))?

Example
Linear Scale

Ln(1+Abs(x))*(x/Abs(x)) Scale
What counts as information though? I think histogram gives enough information about average hit error and standard deviation, and is much more compact. Here is an example http://osureplay.com/#/replay/f1536ea1f ... a00e8c5440
And as I've said, hit error v/s time graph is useful, but is a little bit too much for a score page. It is suitable for map editor to test timing.
abraker

chinmeitou wrote:

I think histogram gives enough information about average hit error and standard deviation, and is much more compact.
Hit error and std dev is already displayed in results screen.
  • hit error is the +/- values you see when you hover over the HP graph. Those are actually useful for determining the offset you need.

    Std dev is hidden behind unstable rate and is also visible when you hover over the HP graph. At least in taiko and mania, those are useful for determining accuracy to a more precise degree than the max 100% a player can get due the discrete hit ranges
The histogram offers no additional meaningful info to those two values but the nice pretty visualization of a distribution. We know it will have the following properties:
  • The histogram's width can be determined using the std dev, which is a value we already get to see

    The highest point of the histogram should be around the midpoint of the hit error range if it is a normal distribution, and hit error range are values we already get to see

    The histogram itself will never have values more or less than the boundary at which the hit error counts as a miss. The domain for the histogram is confined within known values

    The histogram will likely have a shape of a normal distribution if the play was done by a human player
As you can see the only addition info the histogram offers unneeded visualization and whether the distribution is normal
Radiohead
I'm just gonna go with the person above me.
qwr
I've noticed some cheaters configure their cheats to have their hit error centered around some constant non-zero offset. If there is large spread, this is weak evidence, but with good UR (ex. 70 UR) this is suspicious and I have used it in osureport reports.
Please sign in to reply.

New reply