There are several possible reasons for that variation.[ Momiji ] wrote:
103,92 UR = 94.45%?
what
is it because the 100s were major lapses?
- The previous table only accounted for accuracy for circles (not sliders nor spinners). Assuming all sliders were a 300, 98.29% accuracy is equivalent to 97.40% accuracy on circles for that specific map. The game takes the error on sliders as well as circles for the calculation, if you tend to hit sliders less accurately than circles, then it is likely that in that play your UR on circles was less than the value shown.
- The previous table assumed a distribution of hit errors close to a Normal Distribution with mean error equal to 0 (Normal Distribution + an Uniform Distribution disturbance caused by the fps error; and rounded to the nearest millisecond). If your hit error distribution is not close to that distribution, then the estimation wouldn't be accurate.
- The table took the mean accuracy on circles. Because of the asymmetry of the Binomial Distribution, the median is different to the mean (unless the amount of circles tends to infinity, in that case they are equal); even though the mean is a certain value, there is a probability higher than 50% of getting an accuracy higher than the mean.
- The table took the timing windows shown in the Wiki, which are slightly wrong (18ms instead of 19.5ms for 300s).
According to a perfect Normal Distribution of hits (and fixed timing values), the probability of getting an accuracy of at least 98.29% (assuming all sliders are 300s) in that map with an UR on circles of 103.84 is 31.69% (so it's a very slight fluke according to the estimation). The median accuracy with that UR is 98.00% (the mean is 97.87%).
Edit: Also I noticed that that play was OD9.8, not OD10, fixed the values.