Is random totally random?

[donny]

Hi, I have a little question....

I've heard people talking about this earlier, so
When playing mmc random sometimes it seems that just some tracks are being played....
Does that program select the songs totally random, or does it include stuff like pref and such?
Maybe this is because my collection is pretty large, so the random numbers can't pick all of the songs?

[Spoons]

much of this was discussed recently but we lost the threads that covered it due to the last server hack that took place.

having said that, from memory random was set something like 100 tracks apart? so if you had 300 tracks the first song should not be picked up again until after 100 tracks. well something along those line. spoon could clarify that one a bit more though.

as there was a whole discussion on how random play is defined. my definition of random play was to never have the same track played twice and once all the tracks played it would stop.

i think spoon is putting out an option on how random play works to meet 2 needs. One need to suite most people and the other need to suite me :D hehe just joking :shocked:

also from memory random play still didn't work 100% according to the current setup.

but hey if everyone wants to go over and discuss "random play" and how they would like to see it implemented into dap, i am all for it

[MasterOfPuppets]

I already posted my ideal solution to the randomness question in the old thread, but I wanted to see how many people actually understood it / liked it.

What it was was a way to play the songs that had a high 'Times Played' number more often. Given three tracks in an MMC with a Times Played of x, y, and z, if Random Play was selected, they would be played in a ratio of x:y:z.

[donny]

I don't know is that good... 'cause then songs that we listen more would then be listened even more and others couldn't get any time for themselves... And on the other hand I think that that wouldn't be nice for programming
I read in the help file that songs with biger pref. are more often played, is that true?

[xoas]

As I recall from the last thread on this, Spoon indicated that tracks with a higher preference rating would have a greater preference of being played. Spoon was not specific about how this works out in practice although it sounds as though he has a type of algorithm similar to what Master of Puppets outlined except that X, Y and Z (in his example) are fixed numbers which Spoon wants to keep fixed. I am not sure whether this is because changing these ratios would be technically difficult or if he just finds the idea of making these ratios variable would be foolish for either practical or esthetic reasons.
The practical significance of weighting a random selection process, as Spoon does, may or may not be very noticeable. If there is little distinction on the preference ratings for the tracks in your collection (if you rate most of your tracks about the same), then you would not notice much difference based on the preference rating of a track (although tracks that are rated much lower might take statistically longer to get played than they otherwise would have). The other factor that comes into play is the size of the overall music collection. If you had 100,000 (for example) the chance of any given song being picked to be played next will remain very small regardless of preference. And if you only have 25 tracks in your collection the situation would be reversed.
I definitely like Spoon's current random selection (which is based on giving tracks some degree of preference based on its rating and also-as Razgo points out-does not allow any track to be selected within a hundred times of the time it was last played). I also liked Spoon's proposals for providing some other alternative selection modes (which included random album selection and also-I think-included Razgo's proposal for a random selection process which would not allow any track to be played a second time before every track in the music collection was played). I believe that his suggestions represented some definite improvement. I am not sure about trying to allow more tweaking of the random selection process because it is not clear to me that more tweaking would necessarily be better and because I respect Spoon's opinion on the matter.
In any case, dBPowerAMP also allows you to narrow a random selection process by artist or genre(s) or even by year if you should so choose or it lets you play exactly what you want and in exactly the order you want. So I guess that I am not so worried about a track getting forgotten even though this is a potential for any large collection.
I hope this hasn't caused more confusion although I do like to hear your ideas.
best wishes,
Bill Mikkelsen

[Spoons]

that covers a fair bit xoas.

are you sure the preference rating comes into play with random play?

when i did a test on a random play with a blank previuosly played file with 300 tracks it went through 106 tracks before making it's first repeat. but from that point forward it continually repeated tracks up to 4,5,6 times when there were another 200 tracks not yet played.

so i don't know if thats a bug or just a tweek that's needed.

there are 2 things I would like to see in future versions.

1. the ability to clear the "previously played" list. (for now i just delete the ".Previously played" file.

2. play all tracks randomly once only so it stops playing when done.

Spoon edit... Don't want them to delete any EXES

[Spoons]

i think the random album is a natural progression of usefull features also.

and some how I can't help but think "random play" and "selective play" can work together really well. it has "artist", "genre", "preference", "year". just needs "album" added to it :D

[xoas]

Razgo-
I do not know for a fact that preference ratings affect random selection although I believe I understood Spoon to have said that it did (please let me know if I am wrong, Spoon).
The other thing I wanted to be clear about is the difference between random selection and everything being chosen. If you have a deck of cards and choose one card at a time in random order without returning any card to the deck, then each card remaining in the deck should have an equal chance of being chosen on any given round and all cards will have been chosen at the end of 52 rounds (54 if you count jokers). This is random selection from a group without replacement and this is what you would like to see. This is a perfectly valid use of the term "random".
You could, however, chose a card randomly from a deck and return it to the deck and keep selecting randomly. In this case, each card has an equal chance of being chosen on any given round and over an extremely large number of draws each card should be chosen roughly the same number of times. In this case, however, any one given card could go a very long time without being chosen at all. The larger the deck of cards the more likely it is that a particular card may go a very long time without being chosen. This is an equally valid use of the term "random". My point is that this type of random selection carries no guarantee that every given card will be selected by a certain time.
On the other hand, your experiment could point to a bug in the selection process if a very large number of cards ends up not being chosen, although I understand from our prior thread that there may have been an unanticipated problem in how long a track had to play to get "counted" by the program as having been played. If that was the case then your experiment could be refined to check the amount of randomness in the selection process.
best wishes,
Bill Mikkelsen

[donny]

Just cheking something Razgo...
do you delete prevplayed.EXE or something like Razgo.Previously Played in the dbpower amp dir?

[Spoon]

No deleting of any exes.

The random play works like the card example where each card taken from the deck is returned to the deck for the next choosing.

When choosing a card is taken and if it has a low preference then it has a high probability of going back into the deck without playing.

[Spoons]

sorry i said exe file by mistake. i meant the data file previously play is stored in. although it must get rather large after a while?

[donny]

No I don't think the prev played file gets too big, it just contains the previously played 100 files, which probably won't be played again until they are out of there.
I'm saying probably because I had a few examples when songs that were played about 10 or so files before get to be played again. Of course these files are with high pref.

[Spoons]

yeh your probably right. inside my 64KB file is 324 lines.

for each track:
1. artist
2. track name
3. album name
4. track location/path