Nicolas Liebaert Français

A 101-step tree beats a 1,010-step one, and the market charges extra for falls

This page answers three questions about vanilla options (standard calls and puts). How does the price react when the market moves? Which binomial tree is more accurate? What does the market charge for protection against a fall?

15.8× Less error at 101 Leisen-Reimer steps Against 1,010 Cox-Ross-Rubinstein steps. 14× to 46× from 1,005 to 1,015 steps
4.0× Less error each time the steps double Leisen-Reimer: the error shrinks like 1/n²
13.7% SPY implied volatility at the money 57-day expiry, quotes of 4 November 2026
+3.8 pts Extra implied volatility 5% below the forward 17.5% against 13.7%, same expiry

European option, Black-Scholes-Merton

Saved result, default settings

The live model will load here

Option

This call is worth 8.53, all of it time value.

Option value and intrinsic value (what the option pays if exercised now). The gap between them is the time value: the price of the chance that the option gains value before expiry.
Table of this chart
SpotValueIntrinsic value
Gamma and vega are largest near the strike and fade far from it. Each is scaled to its own maximum, so that the shapes can be compared.
Table of this chart
SpotDeltaGammaVega
Price
8.5267
Delta
0.4939
per point of spot
Gamma
0.01564
per point of spot
Vega
0.3910
per volatility point
Theta
-0.0152
per day
Rho
0.4087
per percentage point of rate

Two binomial trees compared

At these settings, 21 Leisen-Reimer steps beat 1,281 Cox-Ross-Rubinstein steps.

Error against the exact Black-Scholes price, with both axes on a log scale (n is the number of steps). A line that falls with slope -1 means the error shrinks like 1/n. Slope -2 means it shrinks like 1/n².
Table of this chart
StepsCox-Ross-Rubinstein errorLeisen-Reimer error
Black-Scholes price
8.526740
Error shrink factor
3.97
Leisen-Reimer: how many times smaller the error gets when the steps double. 4 means the error shrinks like 1/n²
Early exercise premium
0.3351
American put minus European put

The Cox-Ross-Rubinstein tree gets closer to the exact price as you add steps, but not smoothly. Adding steps can increase the error. The position of the tree’s end points around the strike changes with the number of steps, and this causes its jagged line. At the default settings it lands almost on the exact price at 1,044 steps, by chance.

The Leisen-Reimer tree is built so that the strike always falls midway between the two central end points. Its line is straight and twice as steep.

Delta is the change in price for a one-point move in the spot (today’s price of the underlying). Gamma is the change in delta. Vega, theta and rho are the changes for one volatility point, one day and one percentage point of interest rate. The greeks come from exact formulas. The tests in the repository check delta, gamma and vega against finite differences (the change in price after a small move in the input). Python, NumPy and SciPy run in your browser, in a background thread, so you can keep using the page. The first download is about 31 MB, from this site. Your browser then keeps the files.

The volatility smile of SPY options

At 57 days, a put struck 5% below the forward trades at 17.5% of implied volatility, against 13.7% at the money.

SPY at 766.81, quotes of 4 November 2026. Saved result: the same Python code solved the implied volatility of 608 out-of-the-money strikes.

Option chain of 4 November 2026. Days are counted from that date.

ExpiryDaysStrikesAt-the-money volVol 5% below the forward
6 Nov 2 142 15.5% 27.4%
20 Nov 16 167 14.1% 19.4%
31 Dec 57 299 13.7% 17.5%
The horizontal axis of the chart is ln(strike / forward): -0.05 is a strike about 5% below the forward. Volatilities above 60% stay off the chart: 22 points of the 2-day expiry, on options that cost one or two cents. The prices are a saved snapshot of the option chain (all listed strikes and expiries), from Yahoo Finance’s free feed, because a browser cannot call a market data feed directly. At the money means a strike equal to the forward. The chart opens on a saved result. Once the model is loaded, your browser solves each volatility again with the Newton-Raphson method. A backup method (Brent) handles the far ends of the chain, where vega is close to zero.

Reading the smile

Implied volatility falls as the strike rises, from the lowest puts to a strike 2% to 7% above the forward, so puts with low strikes cost more. A common explanation is that many investors want protection and few want to sell it. Beyond that point the curve turns up, most of all on the 2-day expiry, but those calls cost one or two cents, so their implied volatility is mostly tick size. The slope is steeper for the shortest expiry and flatter for the later ones.

I use only out-of-the-money options (options that would pay nothing if exercised today): puts below the forward and calls above it. I compute the forward from the spot, the interest rate and the dividend yield. An option that is deep in the money hardly reacts to volatility, so a one-tick error (the smallest price step) in its price moves the implied volatility by several points. With the whole chain, implied volatility on the 57-day expiry jumps by up to 40 points from one strike to the next, for no economic reason.

The smile is also the weak point of the autocall page. That model uses one flat volatility, while its capital barrier sits far down the expensive left side of this curve. So the model makes the note look better than it is.

What I could not show

The SPY volatilities are approximations. SPY options can be exercised early, and the fund pays dividends on set dates. I treat the options as European, with a fixed interest rate of 4.2% and a fixed dividend yield of 1.2%. The forward that separates puts from calls comes from those two numbers. I did not measure it from the market. The effect is largest for the shortest expiries.

The chain is a snapshot from 4 November 2026. It has three expiries: 6 November, 20 November and 31 December. Days are counted from the date of the chain, so an expiry may have passed by the time you read this. The page shows the shape of the smile, and these are not prices you could trade. Of the 696 quotes, 569 are the middle of the bid-offer spread (the gap between the buying and selling price) and 127 are last trades up to four days old. The chart and table use the 608 within 30% of the forward.

No arbitrage checks and no fitted surface. I did not test the chain for arbitrage (prices that would allow a riskless profit), and I did not fit a smooth curve through it.

The full source and 32 tests are at github.com/Nicolas8330/option-pricer.