AutoRate Discussion

Okay, all finished.  I've released 1.5.0-beta1 today, and it can be downloaded here: http://autorate.googlecode.com/files/AutoRate-1.5.0beta1.dmg

Thanks for your patience!

I will be a'testing it to besting of my testing abilities!

Beta 3 is up. Fixes an issue with old parameter values that are no longer compatible persisting when running the new version. Caused ratings to be crazy.

Beta 3 is working as I hoped it would, great job!

So how is your distribution of ratings? From an unbiased point of view, are the controls intuitive enough?

One thing I've nothings and will get fixing is that the sliders for the min and max ratings show stars incrementing up and down but are somewhat misleading. It seems to round up to the next highest full star.

the min rating slider goes 0-0.5-1-1.5-2 despite what the stars indicate.

In actuality you can't specify a min ratings lower than 1 if you elect to reseve 1/2 star ratings for songs with more weighted skips than plays (or you don;t use half star ratings) and 1/2 otherwise.

btw the following command in terminal allows you to edit half-star ratings in itunes:

defaults write com.apple.iTunes allow-half-stars -bool TRUE

(please see post in the distribution thread)

In the next release you should include:

1 -  a read me file explaining in lay terms what the controls are all about, and the kind of effect that they have on the outcome. 

2 - Maybe you might just include those little yellow dialogues that show up when you leave the cursor over an item - you know what I mean.

http://tzisoftware.com/forum/index.php/topic,88.0.html

Just tried out the new version.  I will write a more detail reply tomorrow.  4.5 stars bin seems much more even, but I have 0 songs in the 1.5 bin.  I will give more detailed report in the morning.

There is a definite and explainable reason why you have no stars in the 1.5 rating catagory:

If you calc the percentage of tracks in your library with a play count/frequency low enough required to meet the "1" rating category, and that percentage is more than 11% , you will get zero 1.5 ratings. This applies similarly to all the cumulative bin percentages. The fact of the matter is that the variation in your library is not detailed enough at the lower end. These are probably the same tracks that were previously rated 1.

The ratings derived using this new method result in a mean rating of 3, not counting 1/2 star ratings if they are reserved and assuming a relatively large (1000+ songs) library with ample play count/frequency variation.

For example if you played all your songs completely randomly with no bias against previously played tracks or towards tracks you particularly like, and all the tracks were in your library for the same period of time, then there will be insufficient variation in the library for this method to work and you will get meaningless results. The same would be true of the manual method as well.

The moral of the story is that this behavior is expected and logically correct. There is no discernible difference between the play counts/frequencies of all  the tracks rated "1" (I am guessing with respect to your library?), so how do we decide which ones should be 1.5?  We can't so they all get "1" and as soon as any get played enough to be worthy of "1.5" , it shall be so.

The songs that, in previous versions, used to be rated 1.5 are probably higher now as 2, not lower as 1.

In my library of ~3300 songs I have only 16 rated 1.5 for exactly the same reason. The frequency method is less sensitive to this phenomenon because there tends to be more variation in play frequency than in play count.

That sounds perfectly reasonable to me. (bangs head against wall for writing reply just after getting home from work).  In all other regards the ratings are slightly more evenly distributed.  Also the interface for the various preferences is just that much clearer.


Now if only you could add a feature to tell iTunes when to count changing a track as a skip.  Sometimes 2-20 seconds just isn't enough time for me to race over to the ipod/laptop.  Sigh.  further ranting on that (which is totally an iTunes issue and not at all an autorate issue), the skip should be counted based on a percentage of the song played.  That is, if you skip the song before it is x% finished, then it gets a skip rating, where x is definable by the end user.

the skip should be counted based on a percentage of the song played.  That is, if you skip the song before it is x% finished, then it gets a skip rating, where x is definable by the end user.

100% Agreed. Unfortunately this is hard coded into iTunes. The only way to get around it is to create a script that when run, set the track time to say, 3 seconds, and then skips to the next track. In this way it would always increment a skip. You could add it to the iTunes script menu and/or assign a hot key to it somehow.

• Added a stipulation that all (and only) tracks that have a play count of 0 AND a skip count of 0 get a rating of zero regardless of any scaling or the library min and max values as determined form the analysis.
  

Shouldn't it just be for songs with a play count of 0? If a track has been skipped 40 times and never played, its pretty obvious that song sucks.

a rating of zero does not mean the song sucks, it means it is unrated. So if its never been played, nor skipped then you have never listened to it (unless you skipped it after 20 seconds in of course)

Thats what the option to reserve 0.5 star ratigns is for. anything that has a higher weighted skip count than the play count gets a 0.5 rating. Which can be interpreted as your "sucks" catagory.

The lowest rating that anything that has a play count greater than its weighted skip count can have is 1 with this option ticked, 0.5 otherwise.