I was browsing through the West L.A Meet up discussion article and found it really fascinating. It will be about humans generating random number strings and the many applications where this would be useful. It’s too bad I can’t attend. Off the top of my head, I feel like I can only come up with one digit randomly by looking at my watch, not sure how I would get more than that. Does anyone have a decent way to generate random numbers on the spot with out a computer?
Pick a nearby object. What letter does its name begin with? Convert that from a letter (base 26) to a number and truncate.
Probably has some systematic bias from the names of common everyday objects overwhelming it, but a decent start.
EDIT: Oh wait… that also has the problem of being biased because you’re truncating and there are only 26 numbers. Maybe the bias against zxwq will almost cancel it out?
You can cast a lot of dice into a shoebox and shake the box on edge so that they all end up in a line and then read them off as a base-6 number, or other bases if you have other shapes. This is just from the diceware page. I personally can’t think of a more efficient way of consistently generating random numbers.
Use something like 10 biased RNGs, like just trying to think of random-seeming sequence the naive way, then convert them to binary, reverse the order of every other one, and XOR them.
This isn’t an even smearing, but looking at a random piece of text and converting the letters into 1-26 should be sufficient for many purposes. If you want additional randomness, add the letters of the first nontrivial word up mod 26(or mod 10, or whatever).
No, that won’t work due to Benford’s Law. In this case, there will be a lot more 1′s and somewhat more 2′s than the other 8 digits. I.e. 10 letters have numbers beginning with 1 and 7 have numbers beginning with 2, but none have letters beginning with 0. The non-random distribution of letters in English text will probably also skew your results.
Hence why I said “sufficient for many purposes”. If you’re trying to choose between 3 places to eat lunch, for example, “the next letter of text mod 3″ is a perfectly acceptable method for determining it. If you’re trying to encrypt nuclear launch codes, not so much.
Benford’s Law applies to the first digit, whereas Alsadius’s use of modulo means taking the last one, which would be much less biased (the bias would be drowned by the bias from common words and letters).
I was browsing through the West L.A Meet up discussion article and found it really fascinating. It will be about humans generating random number strings and the many applications where this would be useful. It’s too bad I can’t attend. Off the top of my head, I feel like I can only come up with one digit randomly by looking at my watch, not sure how I would get more than that. Does anyone have a decent way to generate random numbers on the spot with out a computer?
Read the serial numbers on the paper money in your wallet?
Pick a nearby object. What letter does its name begin with? Convert that from a letter (base 26) to a number and truncate.
Probably has some systematic bias from the names of common everyday objects overwhelming it, but a decent start.
EDIT: Oh wait… that also has the problem of being biased because you’re truncating and there are only 26 numbers. Maybe the bias against zxwq will almost cancel it out?
Use the random numbers from your watch in groups to get more digits.
A mobile phone with an Internet connection and random.org.
You can cast a lot of dice into a shoebox and shake the box on edge so that they all end up in a line and then read them off as a base-6 number, or other bases if you have other shapes. This is just from the diceware page. I personally can’t think of a more efficient way of consistently generating random numbers.
Use something like 10 biased RNGs, like just trying to think of random-seeming sequence the naive way, then convert them to binary, reverse the order of every other one, and XOR them.
This isn’t an even smearing, but looking at a random piece of text and converting the letters into 1-26 should be sufficient for many purposes. If you want additional randomness, add the letters of the first nontrivial word up mod 26(or mod 10, or whatever).
No, that won’t work due to Benford’s Law. In this case, there will be a lot more 1′s and somewhat more 2′s than the other 8 digits. I.e. 10 letters have numbers beginning with 1 and 7 have numbers beginning with 2, but none have letters beginning with 0. The non-random distribution of letters in English text will probably also skew your results.
Hence why I said “sufficient for many purposes”. If you’re trying to choose between 3 places to eat lunch, for example, “the next letter of text mod 3″ is a perfectly acceptable method for determining it. If you’re trying to encrypt nuclear launch codes, not so much.
Benford’s Law applies to the first digit, whereas Alsadius’s use of modulo means taking the last one, which would be much less biased (the bias would be drowned by the bias from common words and letters).
Use memory techniques to memorise a hundred is what I plan to do.
Your random numbers will be more generally useful if other people can verify the randomness.