Options Priced With Nothing but Arithmetic: the Binomial Tree

Black and Scholes gave the world a formula for the price of an option in 1973. It was correct, it was beautiful, and it was almost completely opaque. To understand where it came from you needed stochastic calculus, partial differential equations and a tolerance for heat-equation analogies. Traders used the formula. Very few of them could say why it was true.

Six years later, John Cox, Stephen Ross and Mark Rubinstein published a paper that stripped the whole thing back to arithmetic. Their model, the binomial tree, prices options using nothing more complicated than multiplication and a bit of averaging. And when you make the time steps small enough, it converges to the Black-Scholes formula exactly. It is one of the great pieces of exposition in all of finance: it did not just simplify the answer, it revealed the reason.

The problem: the right answer with no visible reasoning

The deep insight behind option pricing is that an option can be replicated. If you hold the right amount of the underlying stock and the right amount of cash, and you keep adjusting that mix as the stock moves, your portfolio will end up worth exactly what the option is worth, no matter what the stock does. If you can build a copy of the option out of stock and cash, then the option must cost what the copy costs. Otherwise you could buy the cheap one, sell the expensive one, and pocket the difference for free.

That argument is the entire foundation of derivatives pricing. But in the Black-Scholes setting it is buried under continuous-time mathematics, where the stock wiggles infinitely often and you rebalance infinitely fast. The idea is simple. The machinery is not. Cox, Ross and Rubinstein wanted to show the idea without the machinery.

The key idea via analogy: a coin flip you can hedge

Imagine the world is absurdly simple. Over the next period, the stock can do exactly two things: go up by some percentage, or go down by some percentage. That is it. Two branches, like a fork in a road.

Now here is the trick. In that two-branch world, you can build a portfolio of stock plus borrowing that pays exactly the same amount as the option in the up case and exactly the same amount in the down case. Two unknowns (how much stock, how much cash), two equations (the up payoff, the down payoff). Solve them. You get one answer. Since your portfolio and the option pay identically in every possible future, they must cost the same today. Done. That is the option price, and you found it with high-school algebra.

Then comes the second trick, and this one is the magic. Take that price you just computed and stare at it. It looks exactly like a weighted average of the two future payoffs, discounted back at the interest rate. The weights are numbers between zero and one that sum to one, so they behave like probabilities. But they are not the real-world odds of the stock going up or down. In fact the real odds never enter the calculation at all. This is the famous risk-neutral probability: an artificial set of odds under which every asset is expected to earn the risk-free rate, and under which the price of anything is just its expected payoff, discounted.

The astonishing part is what is missing. Nowhere did we need to know whether the stock is expected to rise 20 percent a year or fall 5 percent. Two traders who violently disagree about the stock's prospects will still agree on the option's price, because the hedge works either way. That is the single most counterintuitive fact in derivatives, and the binomial model makes it visible in half a page.

From one coin flip to a whole tree

One period is a toy. So chain the periods together. After one step the stock is at one of two nodes; after two steps, one of three; after many steps, a spreading lattice of possible prices. Because an up move followed by a down move lands in the same place as a down followed by an up, the tree recombines, and the number of nodes grows gently rather than exploding.

To price the option you start at the far right edge, where the option's payoff is known with certainty (a call is worth the stock price minus the strike, or nothing). Then you walk backwards, node by node, applying the one-period rule over and over. When you arrive back at today, you have the price.

Slice time finely enough, with each step tiny, and the jagged tree smooths into continuous motion. Cox, Ross and Rubinstein showed that in that limit the binomial price converges to the Black-Scholes formula. So Black-Scholes is not a separate theory. It is the binomial model with the graininess taken out.

Why it mattered

• It made option pricing teachable. Before this paper, understanding option theory required graduate mathematics. After it, a smart undergraduate with a calculator could derive the core result in an afternoon. Virtually every finance textbook now introduces options through the binomial tree, and only later mentions the differential equation.
• It handles American options, which Black-Scholes cannot. The original formula prices only European options, exercisable at expiry. But most listed options can be exercised early. In a tree, checking for early exercise is trivial: at each node, compare the value of holding on with the value of exercising right now, and take the larger. That one extra line of logic solves a problem that has no clean closed-form answer. This is not a footnote, it is why the model is still used in production decades later.
• It generalizes without breaking. Dividends, changing interest rates, exotic payoffs that depend on the path taken: all can be bolted onto a tree with modest effort. The tree is less a formula than a framework, and its flexibility is exactly what practitioners needed.
• It gave risk-neutral pricing a face. The idea that you can price by pretending everyone is indifferent to risk seems like a mathematical sleight of hand until you see it fall out of a two-branch hedge. This paper is where most people first genuinely understand it.

The honest limitations

• The world does not move in neat up-and-down steps. Real prices gap, jump on news, and occasionally move further in one second than the model thinks possible in a month. A tree with fixed up and down sizes assumes away exactly the tail events that blow up hedged books.
• It inherits the constant-volatility assumption. Standard binomial trees use one volatility number for every node. Real options markets price different volatilities into different strikes, the volatility smile, which a plain tree simply cannot reproduce. Later work (implied trees, local volatility) exists precisely to patch this.
• Convergence is slow and wobbly. As you add steps the price converges, but it oscillates on the way rather than settling smoothly, so a tree with too few steps can be noticeably off, and the option's sensitivities can be jumpy. Practitioners use various tweaks to tame this.
• It is computationally clumsy for many underlyings. A tree on one stock is fine. A tree on five correlated assets has an unmanageable number of nodes. High-dimensional problems went to Monte Carlo instead.
• The perfect hedge is a fiction. The replication argument requires trading with no fees, no bid-ask spread and no market impact, at every single node. In the real world that hedge costs money, and the cost is not small.

The one-line takeaway

Cox, Ross and Rubinstein showed that if you shrink the world down to a single up-or-down move, option pricing becomes simple algebra, the stock's expected return vanishes from the answer, and risk-neutral pricing appears on its own, and that if you then chain enough of those tiny worlds together you get Black-Scholes back, plus the ability to handle early exercise that Black-Scholes never had.