Random Play

[tomskillman]

I've noticed something interesting with shows set for random playing of sequences. I've created a show for the 2009 display. It currently contains 9 songs. I have let it play now for 5 nights. Each song has a number associated with it. My Windows Command that runs at the beginning of each song logs the song to a log file. I have analyzed the log files and I find interesting results:

Song # 10/09/2009 10/10/2009 10/11/2009 10/12/2009 10/13/2009 Average 1 20 21 14 20 19 18.8 2 28 22 21 10 15 19.2 3 24 24 16 11 18 18.6 4 17 19 27 20 20 20.6 5 5 1 1 2 1 2.0 6 18 22 19 26 12 19.4 7 16 21 20 17 28 20.4 8 22 14 15 20 13 16.8 9 0 0 0 0 0 0.0 0 20 21 12 16 16 17.0

Hopefully that displays correctly as a table.... As you can see, song number 5 is rarely played. Song 9 is not yet in the show, which is why it hasn't played at all....

Has anyone else actually looked at statistics of song freqency? I find it interesting that on random play, song 5 is rarely played....could it be the way Random Play is done in LOR? Sequence #5 is What's This from the Nightmare before Christmas - is it possible that the apostrphe is confusing LOR? I'm going to try removing the apostrphe tonight and see what happens. Any other theories? The sequence is actually the second one listed in the show playlist.....

[ItsMeBobO]

Hey Tom.
Very interesting find and cool use of command line. Let us know what happens.

I hope you don't mind that I unwrapped your numbers. Every other song plays up to 10 times more often than #5.

[bob]

Can you please send me copies of your show file, sequence files, and media files? My email address is bob@lightorama.com. Thanks.

[tomskillman]

Bob,

I'm gathering that for you now.....should be sending in a little while. I'll also send you my raw data logs showing what played when...

[tomskillman]

I removed the apostrophe from the sequence that is playing so few times (What's This) and it didn't make a difference. Tonight, I moved it in the playlist to see if it's the location in the playlist that's causing the issue.

[jerrymac]

I had nothing to do with the programming in LOR but I do know numbers, and statistics very well. The issue you are seeing is something I would very much expect with "Random Selection". Each time a sequence of numbers is looked at for a random selection the odds of any single number comming up is the same (and calculatable 1 in 9 in your case), the next time the sequence is selected for a random selection the odds are exactly the same as in the first selection (1 in 9). After 100 attempts at the selection the odds are still the same (1 in 9) as to that single number. Long story Short after 1000 selections it would not be unreasonable for any single number not to have been seclected in a random selection process if the slection is in fact random.

Sorry not much help but random is random

[ItsMeBobO]

I agree with your facts but not your conclusion Jerry. As the sample size gets larger, the likelihood of imbalanced outcome should go down. If its random... that is the question posed, not a given. Since each days result shows a dip in the same spot, and additional days repeat the prior imbalanced pattern, it does not appear to be random.

As a programmer who has generated random numbers many times, it appears there is a possible issue with the code that applies the random number. It may be repeatedly reseeding the random number generator with the same value, or incorrectly adjusting the fraction result of RND() into a whole number. Or incorrectly creating a bias when bypassing numbers which have come up recently so a song won't repeat back to back.

Presented with this sample data on a program I wrote, I would conclude it should be reviewed for something that is causing the bias and add it to the to-do list. I would also need to verify the record keeping method is accurately recording the plays.

[jerrymac]

Valid point, In original Microsoft code (early DOS days) when the random number generator was added, would exactly duplicate the same seqeunce of numbers EVERY single time and progamming had to be added for keyboard input to stop the generator (since you could not exactly stop at the exact same point every time what would occur would be a differnet number each time. Later code was modified to add a seed that would kick off the random generator at a different point each time, masking the problem (as computers do not truly understand "random"). All things being equal any number should show up in a reasonably large sampling (but statistically any number might also not show up).

Of no consiqence but food for thought, besides we are here to have fun and entertainment.

[tomskillman]

Just checked back....even after re-arranging the order of the songs in the show list, that same sequence still plays much less than the others (1 time last night in 6.5 hours). I have also noticed a lot of times where the same song is played back to back....maybe I should implement my own random number generator to be used if there is nothing in the queue.......I haven't put the numbers into my spreadsheet yet for the last 2 nights, but with the song only showing up one time last night, I don't expect much else to change either....

[tomskillman]

Bob,

I think I found the problem....Looking at the status screen from the new version of LOR tonight, it showed a problem with the media file for the sequence that doesn't play as often.....I have adjusted that as of the end of the show tonight...we'll see how it plays tomorrow night. I would suggest though, that in the next version LOR check to see that the same song isn't played twice in a row in random mode....