Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[zeltak]

hi guys!

I've posted a similar Q a few weeks ago but even though i got some good answers it still dosent seem to work, although the duplicate artist, album etc has been fixed i still notice for example that in a random playlist with 5 or 4 stars and no dups (~) with a library of lets say 20,000 5+4 star tracks i still get tracks that appears really often while other tracks never appear at all. now i could just say its luck of the draw but i get artists/tracks that apperead for example 10 times this past month while other artists/tracks never appear even though they have twenty times the songs!!! how do i slove this really annoying problem? smartlist are one of the most important feautures in MC and it just dosent seem to work really randomly as in taking all the songs and putting it in a hat and drawing randomly    

btw an example of one of my playlist is :
[Media Type]=[Audio] ~nodup=[artist] ~sort=[Random] ~n=100 [Rating]==5,=4 [Last Played]=>=4320 [duration]=<600

thx alot

zeltak  ?

[Matt]

Take out the ~n=100, then count how many of each artist there are.

Add the ~n=100 back and run it a few times and count again.

The ratios should be quite close if you use a large enough sample size.

I'm guessing some of the other criteria are making it so it doesn't seem random since you're getting 20% (or whatever) of the 4 and 5 rated songs -- not 20% of all the songs in your library.

[Sir Alan]

Most people confuse "random" with "even distribution".  Take a lottery, for instance.  It is actually more common for "clumps" of numbers to appear than for a seemingly evenly-spaced selection.  Also, the fact that number 42 appeared in the last four draws does not make it more or less likely that it will appear in the next draw.  It has exactly the same chance as any other number.  What has gone before has no bearing on what will happen next time.  The same goes for a random selection from the music library:  it is perfectly possible for the same track (or artist) to appear regularly in entirely random selections.  Or not at all.

Such repetition could be prevented by building a separate reference list pointing to the library entries, shuffling it, and then working through it sequentially, preserving it between runs.  Even then you would more than likely get clumps of the same artist, or long intervals where an particular artist does not feature, but at least you would not get repetition.  (But I bet if you have two versions of the same song by different artists, they will be placed next to each other!  That's Sod's Law.)

[Matt]

Try the latest build.  There were some enhancements in this area.

If it still doesn't feel random, something else is going on. (and we can keep digging)

[roognation]

Please excuse in advance any ignorance on my part (I have searched a number of posts, and nothing).

I think there are a lot of people on the forums that would like to know what happens inside MC when [Random] is used in a Smartlist.  If MC is built with VB and using the RND function, well I think the community here demands something more elaborate for MC.   There are plenty of options for VB and getting decent pseudorandom numbers.  It looks like this thread also was left hanging: http://yabb.jriver.com/interact/index.php?board=3;action=display;threadid=9866

I also remember talk about a GPL-able pseudo-random function that was developed by some Japanese graduate students: http://www.math.sci.hiroshima-u.ac.jp/%7Em-mat/MT/emt.html.  Also see http://www.agner.org/random/ or perhaps the VB here http://www.codetoad.com/vb_random.asp   Why haven't these been evaluated or implemented?

Someone also mentioned that humans do not recognize random too well, and this may be true, but humans are good at recognizing patterns...and that is what we see (I see from my 25k track library tracks from the same album.  Sequentially!).

Can we get an open and clear/concise discussion from the MC team on what is going on?  The elusiveness is a bit frustrating.

Thanks!

[GHammer]

I have commented on this before.

I've been given the short course in statistics, etc.

The fact of the matter is that the same titles show up WAY too often in random lists.

Do I care if the list is statistically random? Probably. And from the results over about a year, there is NO way the process is random.

Recently played artists/selections seem to have a much better chance of being included in a random playlist.

Anymore, I just do a playlist of 4 star rated files for example and put MC into shuffle mode.

I agree with others, there are solutions available to get random numbers/seeds other than the RND function included in any development platform. If those were even close to being unpredictable, encryption tools would not use other methods of geting a seed for example.

[Matt]

MC's random algorithms are very good.  We've done statistical analysis of millions of shuffles.

Remember that newer builds sacrifice some true "randomness" in the name of maximizing the distribution across a certain field. (normally Album)  This increases the apparent randomness, so has been a welcome change.

If some songs show up more than you expect, there are three possibilities:

1) Some other rule in the smartlist is causing it
2) You notice some songs more than others
3) This is random -- it works that way some times (think rolling a Yatzee)

And if you see groupings, that's normal random behavior.  Math shows that it's more common than most people would expect.