31
May 10

## Betting on Pi

I was reading over at math-blog.com about a concept called numeri ritardatari. This sounds a lot like “retarded numbers” in Italian, but apparently “retarded” here is used in the sense of “late” or “behind” and not in the short bus sense. I barely scanned the page, but I think I got the gist of it: You can make money by betting on numbers which are late, ie numbers that haven’t shown up in a while. After all, if the numbers in a lottery or casino are really random, that means it’s highly unlikely that any one number would go a long time without appearing. The “later” the number, the more likely it is to appear. Makes sense, right?

Before plunking down my hard(ly) earned cash at a casino, I decided to test out the theory first with the prototypical random number: Pi. Legend has it that casinos once used digits from Pi to generate their winning numbers. Legend also has it that the digits of Pi are so random that they each appear with almost exactly 1 in 10 frequency. So, given this prior knowledge that people believe Pi to be random, with uniform distribution of digits and no discernible pattern, I can conclude that no one digit should go too long without appearing.

I pulled down the first 10 million digits from here (warning, if you really want this data, right click the link and “save as”). Then I coded up a program in the computer language R to scan through the digits of Pi, one by one, making a series of “fair” bets (1:9 odds) that the next number to appear in the sequence would be the one that had gone longest without appearing. My code is shown below. I had to truncate the data to 1 million digits, and even then this code will take your Cray a good while to process, most likely because I have yet to master the use of R’s faster “apply” functions with complicated loops.

# I think this is how I managed to get Pi into a vector, it wasn't easy.
piVector = unlist(strsplit(myPi,""))
piVector = unlist(lapply(piVector,as.numeric))

# In honor of Goofy Programming Day, I will
# track how long since the last time each digit appeared
# by how many repetitions of that digit are in a vector
ages = c()

# Start us off with nothing in the bank
potHistory = c()

# R just loves long loops like this. Hope you have a fast computer
for(i in 1:1000000) {
# How did our bet do last round?
# Skip the first 100 just to build up some data
if(i &gt; 100) {
if(betOn == piVector[i]) {
potHistory = c(potHistory, 9)
} else {
potHistory = c(potHistory, -1)
}
}

# Increase all ages by 1 by adding to vector, then set the one we found to 0
ages = c(ages, 0:9)
ages = ages[!ages == piVector[i]]

# Count occurences of each digit, find the top digits by occurence to bet on
# And you thought Perl was beautiful?
betOn = as.numeric(names(sort(-table(ages)))[1])
}

# Plot the cumulative sum at 1000 point intervals.
plot.ts(cumsum(potHistory)[seq(0,1000000,500)],pch=20,col="blue",xlab="step/500",ylab="cumulative earnings")

So what was the result? How good was my strategy? After an initial 100 digits to build up data about which digits were latest, I placed a total of 999,900 bets at \$1 each. Final earnings: \$180. That’s so close to breaking even that it’s almost inconceivable. I had 100,008 wins and 899,892 losses. My winning percentage was 10.0018% percent.

On the face of it, this result seemed almost a little too good, dare I say even suspiciously good, if you know what I mean. How rare is it to get this close (or closer) to the exact perfect proportions after so many trials? Assuming that the number of wins followed a binomial distribution with $p=0.1$, my total wins should follow a Normal distribution with mean 99,990 and variance $n*p*(1-p) = 89,991$ (for an “n” of almost a million and non-minuscule “p”, the Normal approximation to the Binomial should be essentially perfect). Taking the square root of the result, and we get almost exactly 300 as our standard deviation. That’s much larger than the 18 extra wins I had. In fact, the chances that you will land within $18/300 = 0.06$ standard deviations on either side of the Normal’s mean are less than 5%. Before getting too worked up over this result, I decided to take a look at the graph. Using the code:

plot.ts(cumsum(potHistory)[seq(0,1000000,500)],pch=20,col="blue",xlab="step/500",ylab="cumulative earnings")

I got this:

The graph looks pretty much like any random walk, doesn’t it? So the fact that I ended up breaking almost exactly even had to do with the stopping point, not any “unusual” regularity. Just to see if I might salvage any mystery, I tested the very lowest point, -\$2,453, which occurred after 202,133 trails. Even that falls within 2 standard deviations of the expected mean for that number of trials, and of course cherry picking the most extreme point to stop at isn’t a fair way to go about this. Any last hope that the graph might be unusual? I plotted a couple random walks using numbers generated in R. Most of them looked like this:

This looks to have the same level of “jaggedness” as  the results of my bet on Pi. Unfortunately, I am forced to conclude that the promising strategy of “late number” gambling turned out to be fairly retarded after all, at least so far as it applies to the digits of Pi.

29
May 10

## Weekend art in R (part 1?)

As usual click on the image for a full-size version. Code:

par(bg="black")
par(mar=c(0,0,0,0))
plot(c(0,1),c(0,1),col="white",pch=".",xlim=c(0,1),ylim=c(0,1))
iters = 500
for(i in 1:iters) {
center = runif(2)
size = rbeta(2,1,50)

# Let's create random HTML-style colors
color = sample(c(0:9,"A","B","C","D","E","F"),12,replace=T)
fill = paste("#", paste(color[1:6],collapse=""),sep="")
brdr = paste("#", paste(color[7:12],collapse=""),sep="")

rect(center[1]-size[1], center[2]-size[2], center[1]+size[1], center[2]+size[2], col=fill, border=brdr, density=NA, lwd=1.5)
}

28
May 10

## R: More plotting fun with Poission

Coded as follows:

x = seq(.001,50,.001)
par(bg="black")
par(mar=c(0,0,0,0))
plot(x,sin(1/x)*rpois(length(x),x),pch=20,col="blue")

28
May 10

## The guessing game in R (with a twist, of course)

Maybe you remember playing this one as a kid. If you are about my age, you may have even created a version of this game as one of your first computer programs. You guess a number, the computer tells you if you if you are too low or high. I’ve limited the number of maximum guesses, and randomized the computer’s choice based on the Poisson distribution (more on that later).

Here’s the code. This was part of my attempt to understand how R reads input from the command line. One of the things I learned: you may need to save this to a file and run it with “source()”, instead of running it directly from the console, line by line.

# Classic guessing game with twist
x = 0
gotRight = 0
failed = 0

# Initial lambda for our random var
correct = 2000
initial = correct

# How many guesses should we allow per number
maxGuesses = 7

while(x != Inf) {
# The +1 part makes sure we never get zero, which would trigger 0's forever
correct = rpois(1,correct) + 1

# The advantage of using "cat" instead of "print" is that you remove those pesky quotation marks
cat("I am thinking of a number between 1 and infinity. What is it? (Type Inf to quit)\n")

# Solicit input from the user
x = scan(n=1) # Just one item in this vector

# Be nice and let the user quit.
if(x == Inf) {
cat("The correct answer was", correct, "\n")
cat("You got", gotRight, "right and failed", failed, "times. Maximum allowed guesses was", maxGuesses, "and initial lambda was", initial, ". Goodbye.\n")
break
}

for(i in 1:maxGuesses) {
if(x == correct) {
print("You rock!")
gotRight = gotRight + 1
break
} else {
if(i == maxGuesses) {
cat("You ran out of guesses. I will pick a new random number based on the last one.\n")
failed = failed + 1
} else {
if(x < correct) {
cat("You are too low. Guess again.\n")
} else {
cat("You are too high. Guess again.\n")
}

x = scan(n=1)
}
}
}
}

Note 1: My code makes a couple uses of the aparently controversial “break” function. I can still recall a heated debate I had with a CS professor who believed that calling “break” (in Python) was as bad as crossing the streams of your Proton Pack. That said, I have sucessfully used it on several occasions now without any appearance by Stay Puft Marshmallow Man or changing the natural order between dogs and cats. In R, the biggest problem with using constructs like “break” and “while” is that, for reasons clear only to readers of this blog but not myself, if you ask R for help about either of these tokens using

?term

you get an sent an error or to purgatory, respectively.

Hint: Because the random guesses are Poisson based, using a “half the distance” strategy for guessing may not be the best way to go. The hardcore amongst yourselves might want to calculate the median of the expected value conditional on having guessed too low or high.

Note 2: The Poisson isn’t a very good distribution for for this. Maybe you can find a better one, or at least jack up the dispersion like an overzealous offroader tweaking the suspension of his 4Runner.