Down-rating time - Githubissues

This is nota  defect - it is intended functionality. It all depends on how you 
weight the frequency method versus the count method.

If you use 100% weight on the frequency method, then tracks will get a very 
high rating if they are played when new, and if the rate of play (i.e. plays 
per unit of time) decreases subsequently, then the track may get rated lower. I 
say may because it all depends where it fits in the histogram. 

If you use 100% weight on the count method, then tracks will start with a low 
rating and their rating will increase over time as they get played more. In 
this case, the tracks with the most plays get the highest rating, regardless of 
how long they have been in your library.

The defualt is a 50/50 average of the two. Personally, I prefer about a 2:1 
weight on the frequency method versus the count method. In this scenario, 
tracks that are new and played get a mid-high rating immediately, so and that 
rating will either go up, stay the same, or go down as the track ages int he 
library, depending on how often I play it.

In summary... jsut experiment with the Frequency vs count slider and consider 
how it rates tracks that you know you really like - really new stuff, medium, 
and old stuff. See if it makes sense, and adjust otherwise.  Perhaps try the 
two extremes, and then find a balance that works for you.

The other thing is that the lngth of time your library has existed in iTunes 
plays a significant factor. For example, if you imported 10000 tracks just 3 
months ago, the the number of times a track has been played per day, on 
average, may not be overly useful information. I would suggest putting more 
weight on the count method in this case.

In either case, if there is not enough variation amoungst the play counts or 
play rates, then the information may not fit a normal distribution, and the 
results maynot be overly meaningful. This would occur in relatively new iTunes 
libraries (i.e. where the play counts are low). There could potentially not be 
enough information to actually fit the data to a 10 bin histogram. In that 
case, use only whole stars and the resultant ratings will make a little more 
sense becasue it will use only 5 bins when fitting the data to a histogram.

Hopefully this helps some.
Original comment by brandon.mol@gmail.com on 3 Nov 2010 at 6:50
Changed state: WontFix
Added labels: Type-Other
Removed labels: Type-Defect
debeat / autorate

Down-rating time #9
