We priced 54,860 box spreads. The guarantee is real. The profit isn't.

Start here: what a box spread actually is

You do not need to know any options maths to follow this. A box spread is four options on the same underlying, all expiring on the same day, at two strike prices. Call the lower strike K1 and the higher one K2. You buy the K1 call, sell the K2 call, buy the K2 put, and sell the K1 put.

The first two together are a bull call spread: it pays nothing if the market finishes below K1, pays the full distance K2 − K1 if it finishes above K2, and slides between the two in the middle. The second two are a bear put spread, which is the exact mirror image: it pays the full distance below K1, nothing above K2, and slides the other way in between.

Add them and the slides cancel. Wherever the market lands, the pair pays K2 − K1. Not approximately — exactly. A 580-point-wide box on the S&P 500 index pays 580 points at expiry if the index finishes at 6,000 and 580 points if it finishes at 9,000.

That makes a box spread a bond wearing an options costume. You hand over some cash today and you are contractually owed a fixed amount on a fixed future date. Buying a box is lending money. Selling one is borrowing money. And the instant you see it that way, the ceiling on what it can possibly pay you becomes obvious: it can pay you an interest rate. Nothing more. A trade whose outcome does not depend on the market cannot be paid for taking market risk.

First, the good news: the guarantee really does hold

We took a live box — S&P 500 index options, strikes 7,285 and 7,865, expiring 18 December 2026, so 580 points wide — and worked out what each of the four legs pays at eleven different settlement prices spanning about ±9% around where the index was trading. Here is what the two halves do, and what their sum does.

So far so good. Now the part nobody quotes: that box cost 573.90 to buy at the prices actually on the screen. You pay 573.90 today and collect 580.00 in 80 days. That is a profit of 6.10 points — $610 on one contract — and it is genuinely risk-free at expiry.

It is also an annualised return of 4.85%. The 3-month Treasury bill that week paid 4.24%. You did not discover free money. You discovered a way to lend at roughly the going rate, using four option legs instead of one government bond, and it was the single best wide box in that entire expiry out of 8,327 we priced.

Why 81.6% of them lose: the four spreads you have to cross

Every option has two prices. There is the price you can sell at (the bid) and the higher price you can buy at (the ask). The gap between them is the market maker’s fee, and a box makes you cross it four times — twice buying, twice selling.

That fee is a fixed number of dollars. The payoff of a box is K2 − K1, which you choose. So the wider you make the box, the smaller the fee is as a share of what you collect. Narrow boxes are nothing but fee.

Feed that cost into the return and you get the chart the guarantee-return videos never show. The same box structure, the same expiry window, the same day — sorted only by how wide it is.

Two things are worth sitting with. First, the SPX line crosses zero and keeps climbing — but it crosses at about 3% width and tops out at 1.54%, which is still less than half the Treasury bill. Second, the SPY line never crosses zero at all. Same index. Same trade. Boxes on the ETF lost money at every width we measured, and the reason is in the column chart above it: SPY’s median spread eats 7.2% of the payoff against SPX’s 2.9%.

Time works the same way width does

The spread is paid once, at the start. The interest accrues every day. So the longer the box runs, the more days there are to amortise that one-time cost over — which is the same arithmetic as widening the box, applied to the other axis.

At a year out the median wide SPX box yields 4.31% against a 1-year bill at 4.50%. That is the closest this trade gets to working, and notice what “working” means: you have converged on the return of the thing you could have bought in one click, and you are still 19 basis points behind it. The options market is not mispricing interest rates. It is pricing them correctly and charging you a toll to access them.

The cleanest result in the survey: every American box lost

S&P 500 index options are European-style and cash-settled — nobody can exercise them before expiry, and nothing is ever delivered. Options on ETFs and single stocks are American-style: the person on the other side can exercise any day they like, and the short put you sold at K1 can be handed to you at the worst possible moment.

That difference is priced. Market makers charge for the assignment risk they are taking on the other side, and dividends make it worse, because a dividend gives a call holder a reason to exercise early. Sort our seven underlyings by how much of a box’s payoff the spread consumes and the ordering writes itself.

The bottom line of that chart is Coca-Cola, which has the highest dividend yield in the set, the worst spreads, and a median box yield of −137.30% annualised. The top line is the one underlying where nobody can be assigned early. Between them: 21,713 American-style boxes, 100.0% of them priced at a guaranteed loss, best case −1.24%. Not most of them. All of them.

The 0.52% that did work — and what happened three minutes later

Across all 54,860 boxes, 285 beat a maturity-matched Treasury bill on gross pricing. That is 0.52%. Charge the four legs at 65 cents a contract — $2.60 a box — and 267 survive, or 0.49%.

But a quote is not a fill, and the only way to tell a real edge from a stale screen is to ask the same question twice. So we ran the survey again later the same afternoon and re-priced every winner three minutes on. Of 424 boxes that beat their bill in the first pass, 409 could be re-priced — and 227 of those, 55%, no longer beat it. The edge was not an edge. It was the screen catching up.

Here is every wide winner that did survive, with the smallest open interest across its four legs — the number that decides whether the trade exists in any size at all.

Read the last two columns together. The biggest surviving edge, 1.34 points, sits on a box whose thinnest leg has an open interest of three contracts — there is no size there, and a fourth leg you cannot fill turns a riskless structure into three naked options. The edges you could actually have put money behind are the September 2027 rows: open interest in the 55 to 107 range, spreads under 1% of the payoff, and an advantage over a Treasury bill of 0.18 to 0.22 percentage points.

The honest ceiling on the guaranteed-return trade, measured across 54,860 real boxes on a real afternoon, is about twenty basis points over a Treasury bill — available on four-leg positions in a 7,660-strike index, held for a year, in a contract where a single missed fill removes the guarantee entirely.

The one thing a box is genuinely good for

This is the part that gets left out of the debunkings, and it is the reason professionals use boxes at all. Everything above is about buying a box, which is lending. Turn it around. Selling a box is borrowing: you receive cash today and owe the fixed width at expiry. And the rate you borrow at is set by the same options market that was charging you a toll on the way in — only now the toll and the rate work in your favour if the alternative is your broker’s margin desk.

On the 29th, the middle of the SPX book implied borrowing at about 5.92% for six months and 6.01% for a year, against a 6-month Treasury bill at 4.33% and a 1-year at 4.50%. So roughly a point and a half over the risk-free rate, on a fully collateralised, fixed-term, no-margin-call loan against an index that cannot be recalled. Individual quotes implied better — as low as 5.43% and 5.60% — but a single best quote on one side of the book is exactly the number that died in our re-quote test, so the medians are the defensible figures.

Whether 5.9% is a good deal depends entirely on what your broker charges to lend you the same money, and that varies by an enormous amount between firms and balance sizes. We are not going to invent a number for it: look up your own margin rate and compare. That comparison is the whole decision, and it has nothing to do with a guaranteed return.

What this means for your money

“Guaranteed payoff” and “guaranteed profit” are different sentences. A box does guarantee what it pays. What you pay for it is set by a live market, and on 81.6% of the boxes we priced that price was higher than the payoff. A guarantee on the receipt side of a trade tells you nothing about the price side.

Any trade with no market risk can only pay an interest rate. This is worth carrying past this article. If a structure genuinely cannot lose regardless of what markets do, then it is a loan, and loans pay the going rate. Anything advertised as riskless and paying far more than Treasuries either has a risk that has not been named or has a cost that has not been counted.

Mid prices are not prices. Every figure in this piece uses the bid you could sell at and the ask you could buy at, and that single choice moved the median yield by 16.40 percentage points. This applies to every multi-leg options screenshot you will ever be shown, not just boxes — the more legs, the larger the lie.

On American-style options, the guarantee has a hole in it. The payoff is fixed at expiry. Between now and then, the short leg can be exercised against you, at which point you own stock and a set of options that no longer cancel. Every one of the 21,713 American-style boxes we priced was already a loss before that risk even arrived.

Four legs means four chances not to get filled. A box is riskless only as a completed set. Legging in — filling three and chasing the fourth — is how a structure sold as having no risk becomes a naked position in a moving market. This is the mechanism behind the well-documented retail blowups in short boxes: an account sells boxes for what looks like cheap financing, the position is assigned or partially closed early, and the “riskless” label stops applying while the market is still open.

None of this is advice, and nothing here is a trade to place. We priced a large number of boxes on one afternoon and reported what the screen said. If you want the reasoning behind how we price and test things generally, it is on the methodology page, and why a high win rate is not an edge is the closest lesson in spirit to this one.

What would change this read

• A different day. This is one afternoon. Spreads widen and narrow with volatility, and a calmer or more volatile session would move every number here. The direction of the finding is structural — four spreads, one interest rate — but the size of the loss is a snapshot.
• Commissions unlike ours. We charged $0.65 a contract, $2.60 a box. A zero-commission account improves 18 boxes out of 54,860 from losing to winning. An account paying more than that makes even the September 2027 survivors marginal.
• Execution better than the screen. Every figure assumes you pay the ask and receive the bid on all four legs. A trader who routinely gets filled inside the spread on a four-leg combination order would see better numbers than these, and that is a genuine limitation of this method rather than a rounding detail.
• The four-day gap in the benchmarks. The Treasury yields are FRED’s observations for 25 September; the option quotes are live on the 29th. Bill yields moved very little over those four days, but the comparison is not simultaneous and the appendix says so.
• Strikes outside our window. We priced every pair within ±6% of spot. Boxes built from far out-of-the-money strikes have wider percentage spreads and were not included, so the survey is, if anything, generous.

Data appendix

Every figure quoted above, with the period it belongs to and where it came from. Boxes were priced from full option chains at executable quotes — buy at the ask, sell at the bid — for all strike pairs within ±6% of spot, on four expiries per underlying, across SPX, SPY, QQQ, IWM, AAPL, KO and TSLA. Yields are annualised simple returns: (width ÷ debit − 1) × 365 ÷ days.