CD Player's "RAND" not really random

The CD changer in my Nissan Altima, when on Random All, would randomly select from across all six discs.

The Random on the Prius changer seems to only select from within one folder, on one disc.

 

Never tried it, since I use my iPhone for audio. But this has been a common complaint to Apple about iPods and iTunes for many, many years.

The truth is, these systems really ARE random, it's just that us humans are good at picking out patterns. For a while, Apple even introduced a slider in the iTunes preferences to make the shuffle seem more random by reducing the probability of playing a second song from the same artist and album, even though doing so actually makes the sequence of songs LESS random. But it feels more random to you, because you're hearing a wider variety.

It does depend on the algorithm used; for example, my iPod plays songs randomly without replacement: that means that if I set it on random and let it play forever, it will play through every song once before it plays any song a second time. If I reboot the iPod, or explicitly choose a song (rather than just skipping forward/back), that resets the list so it will choose from all songs again. On the other hand, iTunes DJ (previously known as Party Shuffle) is random with replacement: that means that each song played is chosen from all available songs. If I have five songs to choose from, I'm just as likely to get song A five times in a row as I am to get A, then B, then C, then D, then E. (Percentage-wise, either of those will happen 0.032% of the time, or one in every 3125 times.) On the other hand, it's 4 times more likely that I'll get A, then B, then C, then D, then anything but E (0.128%). Of course, to restrict it a lot less, if I say you play 5 random songs, the chances are that you will repeat at least one of them is 96.16%; the other 3.84% of the time, you'll get one of the 120 different ways that you can play each song exactly once. You have to play at least 8 songs to get a 50% chance of playing every one at least once (if I did my math right). And if you play 15 songs, there's still a 10% chance that you haven't played at least one of them. And that's with only 5 songs to choose from!

Sadly, I can't remember the math to calculate the probability of your situation; but it doesn't sound non-random to me at all.

That said, there's two points to be made:
1. This doesn't preclude that there could still be a bug. That would require extensive testing and mathematics, however.
2. Your point is noted; it doesn't FEEL random, even if it mathematically is. You would undoubtedly rather have it feel random than have some nerd like me try to prove to you that it is random. As is often the case in humanity, emotions trump logic.

 

Just to quibble a bit, but no consumer music player has truly random play. Being computer based, they are pseudo-random. Random computer routines are algorithmic, which means they are actually long repeating patterns. If designed correctly they do a good job mimicking a true random pattern, but they aren't really random.

Note that none of this invalidates the rest of this posters comments. As I said before, I am simply making a technical quibble. I suppose it could come into play if the pseudo-random generator were poorly written, but even a rookie programmer would do better than what Tideland describes.

Tom

 

Finally someone else who notices the same thing. RAND in any input (CD/US ) is not a true random...i'm thinking of starting to use the ipod via bluetooth so i can do the random/shuffle through the ipod, but I like the song info on the OEM screen. choices choices choices...

 

My Civic hybrid, when on random using a CF card, would actually play them in random-order, per se. It didn't play song 1,2,3,4,etc but would play them in the same non-sequential order. It took me a couple weeks to realize it (hey, I only drive about 10-15 mins at a time for most of my trips). It got to the point where I knew how many times I needed to FF past songs to get to the one I wanted to hear at that moment.

I just let my iPod Touch shuffle or listen to the radio now.

 

It is obvious that your Prius and you have differnt tasts in music. It will play the songs it doesn't like if you ask nicely, but if you let it pick the songs, it will stick to songs it likes best.

 

It still blows my mind that my engineering-school _Handbook of Chemistry and Physics_ had *tables* of random numbers! I used to keep that old "Rubber Bible" shelved right next to Arthur C. Clarkes _The Nine Billion Names of God_."

 

Yes it does. It is a very low possibility that they have never played over the course of a year, but that is the nature of "random". If there is no replacement, then eventually it will play. Leave your car on for a month, and Im sure the song will come back to the top (assuming your music collection can be fully played in a months time). If there is replacement, then there is no guarentee any song will ever play, but there is a guarantee that a song will play. It could very well pick the same song every time and never advance.

Also, I have that same technical quibble. There is no such thing as random in any computer system. Generally it is a number that counts in some direction starting at some base number and some other increment/decrement and every clock cycle changes. This is then modulo'd over the number of songs (easiest way, but lots of others too), and gives a basic random generator.

 

The notion of random is pretty funny. Couple that with the human mind's capacity for finding patterns where none exist and you get some very strange observations. Keep in mind, in a truly random setup, it's just as likely to play every song IN ORDER as it is to play any other combination of songs. Still technically random, but not to our brain it's not. I remember reading some years ago about how Apple actually has some "un-randomization" code in iTunes and iPod to take some of the randomness out so as to make *us* feel like it's MORE random. Go figure.

 

And to quibble just a bit more, the odds of having a true random generator in any general appliance is very slim.

There are a few good random number generators in the oepn source world - but they require a random input (to start) - and getting a true random start is near impossible without a special purpose chip

 

ok you guys totally overkilled the "random coding" topic. Does anyone want to re-write the sequence or random logic?

 

There are two random options.
1. Play songs on one disc "randomly."
2. Play all songs on all discs "randomly."

 

With systems started by human users it's pretty easy to get a good seed based on the truly random time that the user hits the start button, unless there is some sort of timing constraint limiting when the start button is pushed.

One of the big problems with good pseudo random generators are the very large number of permutations that can be generated for a small number of items. For example, none of the computer card games give a truly random deal of cards. There are too many permutations for any reasonable variable precision.

Getting back to the original topic, what most users want for random music play is really randomized non-repeating patterns. Typically people want their music player to step through all of the songs in the list without repetition, but do so in some randomized way.

Tom

 

*if* we are talking random numbers, and hence *if* we are talking a true random seed, there is no way a human - starting a car will create a true random seed. It must be designed that way. Pressing "the button" applies power to the system - which is always done in a similar way.

A random system must start with a truly random seed. This is very simple *if* you have a NIST certified RNG (random number generator) chip to seed the "pseudo" RNG.

Now - we could play with the light sensor (for headlights located on the dash) - and implemented correctly (using its output), we may come close, or may pass as a RNG - but I think that was not part of the radio design. Using what's on the car, there is {IMHO} no way to produce a truly random event - without some tinkering with some input like a light sensor.

I do not think Toyota engineers thought about a proper seed - or even a NIST verified/certified algorithm - in their design.....

Just my $.02

 

You are assuming that the entire system is powered off, which it is not. A simple counter incrementing at a very high rate serves as a good seed generator when it is sampled by the power-on keystroke. This is true because the human operator has no knowledge of the state of the counter, and the period of the counter is very small compared to the use of the vehicle. As the counter cycles through its range it is equally probable that any counter value will be selected when the car is powered up.

This sort of system is very commonly used to seed random number generators in systems that involve a random input.

Tom

 

Well put - but still way off base.

Any system that requires a 'touch' is relying on that 'touch' to be random. In fact, I could study the system - watch its workings, and get a good idea of how to replicate an event in the past (working on the knowledge of the system and how it powers up).

With good study, I could recreate that original event. This would mean the system was not random.

It may take me 1000 years, but my great great ... great grand child would prove your point invalid. Likely with the correct measuring devices, this could be cut down a bit.

To be fair, the 1000 year estimate is just that - and has no relation to anything scientific - a guess at best assuming how long it takes to power on the device - and knowing more about the 'system'.

EDIT: I would add that the system we are talking about is 90+% likely to not be 'running' - as in system (CPU) clock - this drains battery and is pointless. More likely is some SRAM based clock chip which runs low power, and keeps the system correct in our time frame.