Broken Wing Butterfly
A broken-wing butterfly buys one option, sells two further out and buys a third beyond them, with the outer wing wider than the inner one: an ordinary butterfly with one wing moved away. The move makes it cheaper, here almost free, and limits the loss on the near side to the small debit. What it gives up is the far side: beyond the moved wing it loses a fixed amount, the extra width times £10 on the FTSE 100 contract, plus the debit. It is designed for a drift to the sold strike that stops short of the far wing, and it saves most on the side where skew makes the far wing cheap.
The symmetric version is covered on the long butterfly page, and the ratio page's family table sets a put version beside the ratio and backspread on one chain. This page works the call side, where the discount is largest. The FTSE 100 is a model input here; the page takes no view on the index.
A butterfly with one wing moved out: 10,800 / 2 × 11,000 / 11,300 calls
Read as two spreads, the risk is easy to place. The lower half, long 10,800 and short 11,000, can make at most 200 points. The upper half, short 11,000 and long 11,300, can lose at most 300. Above 11,300 both are at their limits, so the structure loses the 100-point difference, £1,000, plus what it cost. The maximum loss is therefore (wide wing − narrow wing) × £10 + debit + commission, £1,061.80 here: 17.2 times the £61.80 paid. Leaving the debit out of that formula understates the risk of every debit butterfly.
Which wing to break: calls against puts, and a credit version, priced
Moving a wing out always adds the same geometric risk, £1,000 for every extra 100 points. What it saves depends on the volatility surface, because the saving is the price of the 100 points of protection given up: the spread between the old wing and the new one. On the FTSE 100 surface each further call is cheaper in volatility terms and each further put dearer, so the same move saves more on the call side. The table prices both sides against their symmetric versions, and adds the credit version that a still wider call wing produces.
Probabilities are model probabilities (risk-neutral, lognormal) from the skew surface, not forecasts. Fills: 11,200 call 66.5 (IV 12.36%), 11,400 call 28.0 (11.65%); puts 10,700 at 215.0 (14.19%), 10,500 at 146.0 (14.94%), 10,300 at 97.0 (15.71%), 10,200 at 78.5 (16.10%).
On the call side the 100-point move saves £220, 22.0% of the £1,000 of tail it adds. The mirror-image put structure, 10,700 / 2 × 10,500 / 10,200, saves £185, 18.5%: less, because on the put side the new wing is the dearer strike. The 10,200 put carries 16.10% against 15.71% on the 10,300 put it replaces, while the 11,300 call carries 12.00% against 12.36% on the 11,200 call. The put version is the one that looks like protection, since it sits on the side where falls happen; on this surface it is also the side where the tail is sold more cheaply. Pushing the call wing out to 11,400 turns the structure into a 11.0-point credit: there is then no loss at all below 10,800, where £103.20 of the credit is kept after commission, but the tail doubles to £2,000 and the maximum loss is £1,896.80 beyond 11,400. A credit is available only by widening the wing, and the saving per pound of tail falls as it widens (19.3%). How skew arises is on the implied volatility page.
The worked example: a 5.5-point debit on Monday 17 August 2026
On the skew surface the model probabilities (risk-neutral, lognormal) are 47.2% that the index settles below 10,800 and only the debit is lost, 26.9% that it settles between the breakevens, 6.5% that it settles between the upper breakeven and 11,300, and 19.1% that it settles above 11,300, where the full loss applies. That last figure is taken from the slope of the call price across strikes, which carries the skew. A single flat volatility understates it: N(d2) at the 11,300 strike's own 12.00% gives 15.2%, and even the at-the-money 14% gives 18.7%. The far wing sits only 5.1% above the index, inside the one-standard-deviation range, whose top for 16 October is 11,377.8.
Open this worked example in the strategy builder (the builder solves each leg's volatility from its fill).
Settlement: a tent at 11,000 and a £1,000 step beyond 11,300
The settlement line is a tent with one long side: flat at the debit below 10,800, up to the peak at 11,000, and down through the upper breakeven to a floor 100 points below where it started. The two curves before expiry are nearly flat by comparison. That gap is the subject of the next section.
The 11,500 row is the maximum-loss formula worked leg by leg: the 10,800 call settles for £7,000.00, the two 11,000 calls cost £10,000.00, the 11,300 call pays £2,000.00, a net −£1,000.00 before the debit and commission. Every level above 11,300 gives the same figure, which is why a 20% rise costs no more than an ordinary 5% one.
Why the tent is out of reach until the last week
A butterfly earns its peak only at settlement, and the chart shows how late that value arrives. Each curve is the position valued with that many days left, each strike keeping its volatility.
With 30 days left the best the structure can show anywhere is £189.63, about a tenth of the maximum, and it shows that near 10,695, not at the sold strike. Even with the index sitting exactly on 11,000, the mark is £26.53 with 30 days left and £581.80 with a week left. The time value in the two sold calls is what stands in the way: it is at its largest where the index needs to be, and it only runs off at the end. For the same reason a profit target set as a share of the maximum cannot fire early. A quarter of the maximum, £484.55, is not available anywhere with 14 days left, and on a steady climb to 11,000 it first appears on Wednesday 7 October 2026.
Stress: a crash costs the debit, an ordinary rally costs the maximum
For this family the stress test comes first, and here it reads unusually. The index jumps at once on 17 August with the stated volatility shift; the columns give the mark, the settlement result if the index stays there, the requirement and the new delta. The general method is on the Level 3 page.
A 20% fall of the kind listed in the Level 3 page's gap history costs this structure its £61.80, and the requirement does not move. The row to read is +1 SD. An ordinary 5.8% rally takes the index past the far wing, so settlement there is the full £1,061.80 loss, yet the mark straight after the move is −£529.93. The four legs still hold a net £525.07 of time value, so the screen shows £531.87 less loss than settlement would (the other £6.80 is commission) and looks recoverable. Unlike a short strangle, whose worst case needs a large move, this structure's worst case is a move the model gives about a one-in-five probability.
Four paths, with the worked plan's conventions firing first
The worked plan's teaching conventions (methods page) are: take profit at a quarter of the maximum, £484.55; close if the index trades at or above 11,300, the far wing; and close on Tuesday 13 October, three days before settlement, whatever the mark. They are illustrations with a cost in pounds, not rules. Closing figures include commission and half the quoted spread both ways, £93.60 on four legs.
Path D is where an adjustment comes up: rolling the wing in by buying the 11,200 call and selling the 11,300, which turns the position into the symmetric butterfly. At 11,240 with 21 days left that call spread costs 52.5 points, £528.40 with commission, which lowers the maximum loss to £590.20 and the maximum profit to £1,409.80. The roll removes the tail by paying the model's price for it: from 11,240 the model gives a 46.4% probability of settling above 11,300, and the roll costs more than twice the £220 the broken wing saved on day one. The rolling page sets out close-or-roll decisions in general.
Greeks: short volatility on day one, long it in the final week
The entry value is £3.27 because the fills sit slightly better than the model values. The Greeks page explains position Greeks in pounds.
The payoff chart looks bullish, but the first column is not: delta is −£0.49 a point, theta +£4.49 a day and vega −£41.04 a volatility point. On the day it opens this is a short-volatility position that earns a little from time, with a slight bearish lean. It becomes the shape on the chart only near expiry: a week out with the index on the tent, theta is +£62.14 a day and vega −£66.90, while with the index still at 10,750 the signs have turned (−£7.45 a day, vega +£4.38), because the only option still worth much is the bought 10,800 call. Because the structure is short vega at entry, the level of volatility matters. The example's 14% sits near the bottom of the 60-day FTSE 100 IVI's range over the year to 30 June 2026 (13.54 to 25.59, average 17.24; FTSE Russell); with every strike 3.24 points higher, at that average, the same model butterfly would open for a 5.25-point credit rather than a debit.
UK tax: three computations, netted in 2026/27
The two sold 11,000 calls are one grant, a disposal on 17 August (TCGA 1992 s144(1)), read as outside the same-series pooling in CG55535 because they are written (an inference; HMRC's page does not draw the distinction). The bought calls are different series, each its own computation. All three end in 2026/27 and net off; other gains are assumed to use the £3,000 exempt amount.
So the £2,596.60 grant gain is not taxed on its own when all three legs end in the same tax year: the net across the three computations is what counts. The timing risk needs 5 April in between: opened on Monday 15 February 2027 on the April series (cross-year examples usually open in March; this one opens in February so that its 60 days end at the April expiry) and settled below 10,800, the £2,596.60 grant gain falls in 2026/27 (£467.39 at 18%, £623.18 at 24%) while the £2,658.40 of losses on the bought calls is dated 16 April 2027, in 2027/28: relief deferred, and lost only if never used. HMRC gives no worked example (tax page: written options, bought options, across 5 April). Options sit in a general account and go on the SA108; an ISA cannot hold them and no SIPP administrator permitting them was found (checked 26 September 2026; wrappers).
Requirement, permission and costs
Requirement. Taken apart on a strategy-based schedule (Cboe, checked 26 September 2026), the lower half is a bull call spread whose £910.00 debit is paid in full, and the upper half is a short 11,000/11,300 call spread charged at its width, £3,000.00. That is a conservative figure against a true worst case of £1,061.80. A schedule that charges the whole position its maximum potential loss at the strikes, as FINRA Rule 4210(f)(2)(H)(i) does for US accounts (FINRA Rule 4210, checked 27 September 2026), would ask for £1,000 beyond the debit already paid. IBKR's permission list names "unbalanced" butterflies alongside long and short ones at Options Level 3 (IBKR options trading permissions, checked 26 September 2026), so a broker may treat the four legs as one position and ask for less; its order preview is the figure that applies. Funded with £4,061.80, the decomposed requirement plus the whole maximum loss, no move can take equity below it. Every written call is covered, so no uncovered permission is involved, but at IBKR spreads need a margin account (spread margin; accounts and permissions).
Costs. Four contracts cost £6.80 in commission to open and nothing more at cash settlement. Half an illustrative 2-point quote is £10 a contract each way: £40 to open and £40 to close, together 1.5 times the £55.00 debit. On a structure this cheap, the spread and commission are the main cost of any trade before settlement.
Contract and access. ICE FTSE 100 options are European and cash-settled on the EDSP, so no leg can be assigned early and there is no SDRT (FTSE 100 contracts, contract sizes). On ICE single-stock options the same shape is American and physically settled, and a sold call can be exercised against the writer before an ex-dividend date when the dividend exceeds its time value (early exercise before a dividend).