Page 1 of 2
Ratings Script Idea
Posted: Fri Jun 08, 2007 3:18 pm
I have found a way to accurately rate songs in MM based on principles of natural selection. It is a manual method requiring much patience, and a lot of clicking...A LOT. In fact my index finger is getting a little sore...
I was hoping a script could be written simply to reduce the number of clicks required, if anyone else is interested in the idea. I am in 1st grade when it comes to scripting anything.
My rating method works as follows:
1. I set every single song's rating to 2.5 stars. Any newly added songs once this process has begun are also set to 2.5 stars
2. I use the Auto Dj feature to show in the Now PLaying List 1 previously played song, the current song, and 2 upcoming songs selected from the entire library at random.
3. I examine the first two songs in the now playing list and choose which one I like better. The song that wins this Competition gains 1/2 star and the one that looses drops 1/2 star. If i am unsure, i just skip to the next step.
(if a song of 5 stars wins over a song of say 3.5 stars, you obviously cannot increase its rating above 5, so you leave it as is. The 3.5 star song that lost still drops to 3 stars however. The same priciple applies to songs with a 0 (bomb) rating)
4. I then double click the 4th song on the list which pulls it and the 3rd song up to 2nd on the list, and compare them. 2 new songs are added at random to be judged next below.
I repeat this process again and again, perhaps 10-100 times a day, if i feel like it.
Since the songs are in competition with one another, they end up rating themselves RELATIVE to eachother. It is a statistical process requiring perhaps several thousands of trials and plenty of patience.
What eventually happens is the songs distribute themselves into a normal curve (or Bell-curve). Eventually, the proportions of songs should reach as follows:
0.0 stars = 1%
0.5 stars = 3%
1.0 stars = 6%
1.5 stars = 12% low std Deviation
2.0 stars = 18%
2.5 stars = 20% avg
3.0 stars = 18%
3.5 stars = 12% high std Deviation
4.0 stars = 6%
4.5 stars = 3%
5.0 stars = 1%
After rating my songs like this for a few months on and off, the ratings in my library are about 2/3 of the way to that distribution above, they still, however, take a narrower bell curve shape and the ratings are far more accurate then any automatic rating script has been able to provide, since they go off played count, skipped count, and not your personal preferences. There are a few oddities at first, like one of your favorite songs dropping to a 1.5 star or something, but after many trials, it works its way up to where it "belongs" relative to the other songs.
Setting ratings to new songs at 2.5 works well, but you may be able to rate new songs by guessing where they should fall, and they will eventually get rated into their true place on the bell curve much sooner once they show up in a few trials.
The scripts functionality, I imagine, could provide a dialouge box perhaps, listing two songs at random (title, album, artist, year, current rating, whatever...), where you are given the choice to select the better song with one click. On the click it would inrcease the rating, lower the loosing song's rating, and load two new songs at random for you to choose. It would also be helpful if you could preview each song with a click if you were unsure from the title. There should also be a skip Button if you cannot decide.
Again, Having a script would reduce the number of clicks from 10 or so to 1-2 for each trial, thus greatly incresing the speed, and saving your finger from aches!
Posted: Fri Jun 08, 2007 3:33 pm
Hmm... I don't know if it is really useful because if you have to chose the better song manually you also can just set your prefered rating, can't you? I mean you need much more clicks with you idea to rate the whole library.
BTW, if somebody really wants to create that script (s)he could use my AutoRateAccurate as base. Because with the autocalibration I use in my script you easily can create normal curve (you already can with AutorateAccurate).
After rating my songs like this for a few months on and off, the ratings in my library are about 2/3 of the way to that distribution above, they still, however, take a narrower bell curve shape and the ratings are far more accurate then any automatic rating script has been able to provide, since they go off played count, skipped count, and not your personal preferences.
As I said you already can create such a curve with AutoRateAccurate. And normally your 'personal preference' correlates with factors like playcounter (and playedperday), skipcounter, etc. But anyway such a script you suggest would be useful for some people. Don't know.
Posted: Fri Jun 08, 2007 3:34 pm
That is quite an interesting approach.
I don't know if anyone is/will be interested in developing such script (if its even possible), but you should be aware that there are 2 different scripts with differing approaches to 'auto-rating' your tracks as/after you play them.
Diddledoo's Auto Rate & Announcer script
Big Bernies AutoRateAccurate script
Both can accomplish something similar to your desires, but not quite the same, based upon various factors such as play counts, skips, etc.
Posted: Fri Jun 08, 2007 5:52 pm
I agree that those two automatic rating scripts are by far
easier, and do allow for a disribution, however they still are largely based on Playcount, which in my opinion has nothing to do with a rating of a song.
The skip-tracking of these do make them much better, but they are still vauge estimates of a songs rating.
Many of my high rated songs are songs which I have loved for years, and played to death long ago outside of MM. Since I play them rarely now, they get inaccurate low ratings under such automated scripts. These scripts are excellent for new libraries of songs, but not for pre media monkey library songs starting with fresh playcounts.
That is why I choose to manualy rate songs, and im sure there are other people with similar issues. However, direct manual rating --- guessing that one song should get a 4 or a 4.5 or perhaps a 3.5 isn't easy to do consistently....pulling rabbits out of hats...ratings out of thin air! My method keeps things consistent by rating songs, not based on playcount or guessing, but rating them relative to eachother.
And it does work beautifully!
Albeit an arduous task...
Having an accurate rating, completley independent of its playcount, allows for more refined and accurate auto generated playlists and what have you...and you can keep those garbage songs that you just can't come to delete at bay, even if they've played 32 times while you were out one day and accidentally left your music playing
Posted: Fri Jun 08, 2007 7:17 pm
diddeleedoo's script uses the play count that you can not change. so it is really a good way of keeping them correct
the playcount you see in the list view is not the one that it uses. because that one can be changed to check how many times a day or week or other schecdule the song is played.
i am not sure but i think big bernys also uses the same playcount that diddeleedo used the one for the core count
(or the one in the database only)
you got to remember though mm 3 wont use the same scripts as mm 2 so both these scripters are going to need to make a new updated one for mm 3, diddeleedoo said he is waiting for the finaly release of mm 3 to even start with redoing his.
Posted: Sat Jun 09, 2007 1:11 am
Of all the strange script requests I have seen, this one must be near the top.
I can not think that anyone would spend all this time rating the same songs over and over again.... And the results are far from accurate, since the songs they meet in the "competition" are chosen randomly.
There is also little reason to expect a normal-shaped distribution for ratings, since you've already made a selection in choosing to have them in your collection. (I have far more 4 and 5-star songs than 0 and 1 since I don't collect crap... And if I have some, I certainly won't have them come up again and again randomly!)
I suggest you rate your songs once and spend some of the time you save to learn how to script. With your patience I am sure you can be of much help to this community
Sorry, Teknojnky, this was NOT an interesting approach!
Posted: Sat Jun 09, 2007 2:15 am
I really like the idea of the script, allthough i wouldn't like doing half of the clicking to rate my library - i think it would take me years (literally) to rate all the songs.
But i think it's a neat idea for not more than about 2000 tracks.
Posted: Sat Jun 09, 2007 6:06 am
i argee with nojac it is kind of really a waste of time.
since the songs are already choosen to be put on the computer
you must have liked them enough for a 4 or 5 star rating, but if you let others put songs on the computer they like then you would need some type of system to rate what you liked.
i just use the rating of the songs to tell me which is played the most, i know that is just a basic way of doing it and i can read the database to see the true playcount. but if the stars are already there to use for the rating. then why not use them, and i am not going to rate every one by my self. so i use the autorate script that works for me.
Posted: Sat Jun 09, 2007 11:21 am
I buy all my music legit and have every song I own ripped to the computer. As such I don't like all of my songs enough for 4 to 5 stars because most cds have their good and bad songs.
Posted: Sun Jun 10, 2007 2:38 pm
I didn't think there'd be too much interest in the idea, it is certianly a highly personalised and rather obsessive way to arrive at ratings. Its also good to do when you're really bored!
im no technojunky, but yes, many moons ago several of my friends decided my music needed some of their music, so i have a bunch of theirs which has become familiar enough that i am hesitant to ever delete.
Even aside from that, if every song i had was a 4 or 5, all my favorites, every 'filler' song weeded out, there will always be some songs better then others. Rating them in competition distributes those high rated songs across the range from 0-5, and just by rules of probability they end up in a normal distribution. The ratings are specific and relative to only what is in the library, so in this example, a one star could still be one of my all time favorites, and a 5 star would be a song almost worthy of worship.
Heres a final thought to ponder, would good songs still be that good if there were no bad songs to compare?
Posted: Sun Jun 10, 2007 3:01 pm
Actually, I think the best usage of this type of rating, would be when you add a new album.
Mass rating and clicking for each song, would be uh just a little too much for most people.
However, one album at a time would be a more digestable chunk.
Of course then that brings up the issue, if you just bought a new album and haven't really listened to it much, your only gonna have as much rating value as you can make for the first few listen.
But anyway, I think its a good idea, for the right time and situation. To me, doing a whole library in that manner would be insane, but to each his own.
Posted: Sun Jun 10, 2007 10:43 pm
the more said the more i think you might beable to use a custom version of either of the two already out there big berny's or diddeleedoo's.
diddeleedoo likes to use the 2.5 star rating for starting out also.
and last i remember his script did not change your manual rating.
i don't know if big berny's was like that or even if he uses the 2.5 stars for the start of the songs.?
but i use an old version of dideleedoo's that he customized for me. it is close to what is online now, but he did a few things to it i wanted it to do.
so maybe when he gets back in to the scripting he will do it for you too, he might make one for you if you email him so might big berny.
Posted: Wed Jun 13, 2007 9:05 am
There seem to be a few people saying this script is a daft idea, when it obviously isn't to the original poster. Why bother trying to persuade him it's a bad idea? If you don't want to or can't create the script, that's fine, but why spend so much effort trying to persuade him it's ab bad idea and pointing him to the track rating scripts when he has already said they don't work for him?
I've only responded because I think this is a great idea for a script, one that would work for me too. I liked the ideas in the rating scripts, but ultimately they didn't work for me because they rely too much on playcount.
While my top favourite songs do correspond to their playcounts, but most of the songs I like a lot like some or like a little don't really get ratings that correspond to those, when using playcount. Also, I have a lot of songs that became favourites or well-liked before I found mediamonkey, plus much of my current listening is new stuff - a lot of which isn't going to be rated very high, but is going to get a high playcount (over a short time frame, too). Those are some of the reasons that basing it on playcounts doesn't work for me.
Posted: Wed Jun 13, 2007 9:15 am
i don't think it is a total waste of time MeMeMe. but i think it would be if used for only one person.
i think it might be more suited for use with two or more people wanting to rate the same songs.. if it can be spilt like that.
but for one person i believe there are ways to customize big bernys script and diddeleedo's scripts to achive what the orginal poster was after. which is why i brought up those scripts. i can't say why the others brought them up.
unless they believe the same about them as i do.
i know with my custom script it raise's and lowers the rating
upon the amout of time in the computer and the number of plays. and i can adjust the number used for the amounts of time in the computer. which is not just the same as diddeleedo added on line but close to it.
and what i think seems to be close to what the orginal poster was trying to get at.
Posted: Wed Jun 13, 2007 11:54 am
rovingcowboy wrote:i don't think it is a total waste of time MeMeMe. but i think it would be if used for only one person.
Surely, whether it's a waste of time or not is for the person actually spending the time to decide?
but for one person i believe there are ways to customize big bernys script and diddeleedo's scripts to achive what the orginal poster was after. which is why i brought up those scripts.
I'd need to better understand the algorithm yours uses, to be sure, but for my purposes and experience with these two scripts, they just don't work for me.
Your script might give a quick start to Billy's process by making an initial guess for the songs initial ratings (rather than starting all at 2.5). But beyond that, it is very innacurate for a large proportion of my music.
I have a lot of tracks that I rate highly, but which don't get played often, and a lot of tracks that I rate somewhere in the upper middle, but which nevertheless get played a lot - because they are parts of sets I use for mood or background music. Plus there are songs in my library that I play more often outside of mediamonkey, like in car CD players, rather than when i'm sitting at my computer. So no system of measuring my playing habits will generate accurate ratings (where I rate the songs according to taste/like).