Beau

Okay, Jo, I have to bring something up that's been driving me nuts. My car insurance renewal just came through. And, like, the number they came up with... it feels so incredibly random. How do they even decide that my specific, slightly-dented 2018 sedan is worth exactly *that* much per month?

Jo

It feels random, but it's actually the perfect real-world example of what we've been talking about with expectation values. It's... it's all about managing risk by calculating the average outcome.

Beau

Right, the whole... multiplying the probability of something happening by its value. But my car is one car! It either gets in a wreck or it doesn't. How can there be an 'average' for just me?

Jo

That's the key! They're not just looking at you. They're looking at hundreds of thousands of 'yous.' Drivers with a similar age, car model, driving record, even zip code. They're calculating what they call an 'expected loss' for the entire group.

Beau

Expected loss... so, the average amount of money they expect to pay out for each person like me.

Jo

Exactly. Let's make it super simple. Imagine an insurance company covers 10,000 people. And their data says that in any given year, one percent of those people will have an accident that costs, say, $5,000 to repair.

Beau

Okay, so one percent of 10,000 is... a hundred people. So they expect 100 accidents.

Jo

Yep. And each costs $5,000. So their total expected payout is 100 times 5,000, which is... half a million dollars.

Beau

Got it. And they have 10,000 customers to cover that cost. So they just divide the half a million by 10,000...

Jo

And you get $50. That's the expected loss per customer. That's the absolute bare minimum they have to charge just to break even. Of course, they add more for their own costs and profit, but that fifty bucks is the mathematical core of the premium.

Beau

Okay, that... makes a lot more sense now. It’s not about me, it’s about the 'me' bucket. And my premium is my little contribution to that big half-a-million-dollar pot they expect to pay out.

Jo

Precisely. And this same logic applies all over the place, especially in finance. Think about investing in a startup.

Beau

Oh boy. That feels even more like a gamble than driving my car.

Jo

It is, but investors try to quantify that gamble. Let's say you have an opportunity to invest $10,000 in a new tech company. The venture capitalists who do this for a living, they'll analyze it. They might figure there's, I don't know, a 70% chance the company completely fails and you lose all $10,000.

Beau

Seventy percent chance of losing everything. Yikes. That doesn't sound like a good deal.

Jo

Right, but stay with me. Let's say there's a 20% chance it does okay, and you get your money back, so you break even—a gain of zero. And then... there's a 10% chance it's a massive success, and your investment becomes worth, say, $200,000.

Beau

Okay, so it's mostly likely to fail, but there's a tiny chance of a huge payoff.

Jo

Exactly. And we can calculate the expectation value of this investment. It's the sum of each outcome multiplied by its probability. So, you have a 70% chance of losing ten thousand, so that's 0.7 times negative 10,000...

Beau

Which is negative $7,000.

Jo

Then a 20% chance of a zero gain, which is just zero. And finally, a 10% chance of a $190,000 gain—remember, you invested $10k so the profit is 190.

Beau

Ah, right, the net gain. So 0.1 times 190,000 is $19,000.

Jo

Now add them up. You have the expected loss of $7,000 and the expected gain of $19,000. So negative seven thousand plus nineteen thousand...

Beau

Is... positive $12,000. So even though I'll probably lose all my money... on average, an investment like this is profitable?

Jo

That's the entire model of venture capital. You make a lot of these high-risk bets, knowing most will fail. But the few that succeed are designed to pay for all the failures and then some. The overall portfolio has a positive expectation value.

Beau

So whether it's setting an insurance premium to have a slightly positive expectation for the company, or building an investment portfolio, it's the same core idea. You're not predicting the future, you're just... playing the averages over the long run.

Jo

You've got it. It's a way to make rational decisions in the face of uncertainty. You accept you can't know the outcome of any single event, but you can have a very good idea of the average outcome over many, many events.

Beau

Which I guess is why casinos do so well. The expectation value for every single bet is in their favor, even if it's just by a tiny amount.

Jo

The perfect summary. They play millions of hands, millions of spins. The law of large numbers means that over time, their actual results will get incredibly close to their expected results.

Beau

So my insurance premium isn't random after all. It just feels that way because I'm only one data point in a giant spreadsheet.

Jo

You're a discrete random variable, Beau. And they've calculated your expected value.