There’s a stock market scam that goes something like this: make a

list of 1024 people, and send them an “investment newsletter”. The

copy sent to the first 512 people says that a particular stock will go

up; the other 512 get a copy that says that that stock will go down.

Let’s say it goes down. You throw away the list of people whom you

told the stock would go up, divide the remaining 512 in two, and send

them another “investment newsletter”. You tell the first 256 that some

other stock will go up, and tell the other 256 that that stock will go

down. Eliminate those to whom you gave a false prediction, divide the

remaining ones in two, and send them another tip, as before. Do this

ten times, and you’ll wind up with one person to whom, by sheer

numbers, you’ve given ten good predictions in a row. You then tell

that person that he’ll have to pay you to receive further stock

tips.

One problem with this scam (from the scammer’s point of view) is

that there’s a lot of waste: you have to start with over a thousand

names and whittle them down to just one sucker. But what if you

lowered your standards a bit? After all, if someone gets nine good

predictions and one bad one, you can still say you have a 90% success

rate, and that should help sell your nonexistent Wall Street wisdom.

What about 80%? Or 70%? If you start with 1024 names, how many

potential suckers will you have if you consider the ones to whom you

sent seven or more correct predictions, and not just the one where you

got all ten right?

But first, let’s look at a different problem. Let’s say you’re

standing at the corner of 1st Ave. and A St., and you want to get to

2nd and D. You’re on foot, so you want to take the shortest possible

path, but there are several of these. How many? You could walk east,

then turn north on D St.; or you could walk east for two blocks, north

one block on C St., then east another block on 2nd Ave.; and so forth.

As you can quickly check, there are four paths from 1st and A to 2nd

and D.

But is there an easy way to calculate the number of paths to a

given intersection, without having to enumerate them? As it turns out,

there is:

If you’re going someplace on 1st Ave., there’s only one way to get

there: walk east until you get where you’re going. And if you want to

go someplace on A St., the only thing to do is to walk north until you

get there.

If you’re going to, say, 4th and C, there are two main ways to

arrive: either you can make your way (somehow) to 3d and C, then walk

north; or you can make your way to 4th and B, and walk east. So the

number of ways to get to 4th and C is the number of ways to get to 3d

and C, plus the number of ways to get to 4th and B.

So we can fill in the number of paths to get to each insersection:

each number is the sum of the one south of it and the one west of

it.

Another question we can ask is, which intersections can you get to

by walking four blocks? You’re going to walk *n* blocks east and

*m* blocks north, so you can get to any intersection where

*n+m = 4*.

We can now return to our original question: you’re going to make

ten stock market predictions for a bunch of people. Let’s say that

making a correct prediction corresponds to walking one block east, and

making an incorrect prediction corresponds to walking one block north.

So we see that there’s only one way of getting ten correct

predictions: get all ten of them right. There are also ten ways of

getting nine correct predictions: get the first nine correct and miss

the tenth; or get the first eight right, miss the ninth, and get the

last one correct; and so forth. This is the same as asking how many

ways there are of getting to Eye St. and 2nd Ave., which we now know

how to calculate. And the number of ways of getting eight correct

predictions is the number of ways of getting to H and 3d, and so

forth.

1 1 1 1 2 1 1 3 3 1 1 4 6 4 1 1 5 10 10 5 1 1 6 15 20 15 6 1 1 7 21 35 35 21 7 1 1 8 28 56 70 56 28 8 1 1 9 36 84 126 126 84 36 9 1 1 10 45 120 210 252 210 120 45 10 1 |

If we turn our table of numbers so that 1st and A is at the top,

and the line of places we can get to in 10 blocks is at the bottom, we

get Pascal’s triangle. This is one of my favorite mathematical

constructs, because it pops up everywhere: statistics, polynomial

multiplication, and even discussions on how to defraud investors.

(Let’s also note in passing that Pascal’s triangle is symmetrical, and

that each row adds up to a power of 2.)

In row *n*, element *m* is

(*n*+*m*)! / *n*! × *m*!.

It’s easy to see why: let’s say you have a bag of cards that say

things like “IBM will go up (incorrect)”, “Alcoa will go down

(correct)” and so forth. Now let’s say that you’ve reached into the

bag 10 times and pulled out 7 correct predictions (and 3 incorrect

ones). There are 10! ways you could have pulled those particular ones

out of the bag, so that gives us the (*n*+*m*)! in the

numerator.

There are also 7! possible orders in which you could have pulled

out the 7 correct predictions, but since we don’t care about the

order, they’re all equivalent, so we divide by 7!. Likewise, the 3!

orders in which you could have pulled out the 3 incorrect predictions

are equivalent, so we divide by 3!. This gives us

(7+3)! / 7! × 3! = 3628800 / (5040 × 6) = 120

ways to have 7 correct predictions out of 10.

So what does this all tell us? If you’re only going to try to

sucker those to whom you’ve given a perfect 10 correct predictions in

your scam, you’re going to have to start with 1024 names just to

whittle them down to one lonely sucker. But if you want to scam those

who got nine out of ten, you get an additional 10 suckers. Lowering

your standards to 8 out of 10 gets you an additional 45 suckers

(1+10+45 = 56), and reaching futher to those who got 7 out of 10 gives

you 1+10+45+120 = 176 suckers out of your original 1024. A vast

improvement over the original. Until, of course, the FCC sends you to

FPMITA prison.

Nice idea. Of course, none of the messages could be eaten by spam filters, and you’d have to have some sort of way to get the people to realize that you’d ended up sending them so many predictions; most people wouldn’t pay attention to these kinds of emails.

Also, I think the chance of giving a correct and useful prediction a bit less than one half. If you just say “the stock will go up” or “the stock will go down”, chances are it will end up doing one or the other, but only to a moderate degree, to the point where the prediction isn’t really all that mind-blowing. “Ohhh, you said the stock would go up, and a month later, it’s up by 3%. Yaaaaaaaaay, go you.” They’d have to be some pretty damn good predictions to be noticed, e.g. predicting a stock will hit a certain price point and then being correct, or predicting significant movement in the price and being correct about that.

And finally, you’d need to base your predictions on something that sounds reasonable. If you just say up or down, but without any sort of justification or reasoning, you will just be ignored, and very few will pay money for it – after all, it really would just sound like you were guessing. Coming up with convincing rationale for your stock picks would be difficult.

LikeLike

It’s still a nice mathematical model.

LikeLike

Actually, I think that Cyde is giving the suckers too much credit. Remember that there are people out there that fall for Nigerian oil scams, and those are about as transparent as it comes. People also presumably fall for some pretty lame phishing scams, to judge by some of the junk that ends up in my mail spool.

I think the hard part (which Cyde does raise) would be holding their attention over the 10 or so e-mails and have them remember that you’ve been so brilliant. Making up bogus but vaguely plausible justifications would probably just be a matter of stringing together a bunch of market buzz words, and making your “successes” sound significant is presumably just a matter of spin. I’m sure that over 10 predications you could probably make claim like “If you’d invested $X according to my suggestions over the past two months, you’d have a return of Y%!!!”. (Actually, that might be a way to continue to hold their attention. On all but the first “predication” e-mail you could include a claim of that form.)

This does, however, indicate a problem with the original system, since you absolutely need the one sucker to stay with you to the end. If you assume a certain failure rate due to spam filters, short attention spans, basic understanding of the market, generally suspicious natures, etc., then you really need the higher return approaches (like going for the 7 out of 10 folks) for any sort of reasonable return. You might even do better with the 7 or 8 out of 10 folks, since I’m sure at least

somepeople will be very doubtful of aperfecttrack record.Very cool bit of math – thanks for sharing!

LikeLike

Pingback: I am ... unhindered by talent

Pingback: We are also … Unhindered by Talent » Blog Archive » Math is for people who don’t want to be suckers

I’m a bit disappointed that people seem to be missing the point, concentrating on the scam and not the math. This is an ancient scam, and I guarantee you that the FCC knows all about it. I was just amused to find Pascal’s triangle in such an incongruous setting.

LikeLike

I recently wrote an article on this scam. In my case I cited a drop from 80 to 40 to 20–to names and phone numbers taken from the Phone Book and Directories. For a thinly traded stock off the Pink Sheets, for example, that’s enough.

It only takes a few suckers running to their brokers to buy to jack the price to a point of profits for the con man maniupulator.

–Jack Payne

LikeLike