Rolling and adjusting options: mechanics, costs and UK tax

A roll is two trades on one ticket

Rolling an option means closing it and opening another on the same underlying in a single order: buying back a put that was written and writing a later or lower one, or selling a call that was bought and buying a different one. This page works through what each kind of roll changes, what it costs in pounds, how the two halves are taxed in the UK, and when a roll does no more than reshape a loss that has already happened. The figures come from the site's pricing engine on ICE and FTSE 100 (ESX) contracts, with one US example in dollars.

The order ticket shows one net price, but the account and the tax return see two separate trades. The first ends the old position at today's price, so whatever it has made or lost so far is fixed at that moment. The second opens a new position with its own strike, expiry, breakeven and maximum loss. Nothing carries across except cash. A roll "for a credit" is a new option that sells for more than the old one costs to buy back; a roll "for a debit" is the reverse. Neither says anything about whether the combined campaign is making money.

Seen that way, every roll can be read as two decisions taken at the same price and the same moment: would the old option be closed here, and would the new one be opened here by someone who had never held the first? The worked plans in this library use that split to judge a roll, because it separates the result already banked from the risk newly taken on.

Seven ways to move a position, and what each changes

"Tested" means the side the underlying has moved towards, "untested" the side it has moved away from. Rolls are not only defensive: a covered-call writer may roll up because the shares have risen and the writer would rather keep them, and a holder of a long call may roll up to take cash off the table. The mechanics and the tax are the same in every case.

Close or roll: a Tesco 430 put with 21 days left

The worked example follows one written put from the model date to the day a decision falls due under the library's 21-day time stop, a teaching convention whose origin and cost are set out on the methods page.

Model inputs. Entry on Monday 17 August 2026: Tesco at 450p (a model level; Tesco closed at 447.8p that day), one ICE standard Tesco put, strike 430p, expiring Friday 16 October 2026 (60 days), IV 22%, Bank Rate 3.75%, and an assumed interim dividend of 5.08p going ex on Thursday 15 October 2026 (35% of last year's 14.5p total; Tesco sets the real figure with its interim results on 8 October 2026). American puts are priced on the binomial tree: model value 8.30p, filled at 8.25p, or £82.50 on the 1,000-share contract. Decision day is Friday 25 September 2026, 21 days before expiry. Tesco is assumed to have fallen to 425p by then (a path chosen for the lesson; Tesco actually closed at 473.2p that day, price data: Yahoo Finance), and IV is left at 22%. The 430 put's model value is 14.27p, filled at 14.25p. ICE authorises strikes share by share, and we could not find a published table of strike intervals (checked 26 September 2026), so the strikes here are illustrative and the live chain decides which exist. Tesco is used as a model underlying; this is not a view on Tesco.

With 21 days left the put is 5p in the money and worth 14.25p against the 8.25p it was sold for. Closing it fixes a loss of 6.00p a share, £60.00 on the contract, before costs. Every roll below starts with that same buy-back and adds a new put to it, so the table is really a menu of new positions, each carrying the £60.00 with it.

How to read it. "Premium kept in total" is the old sale, less the buy-back, plus the new sale: the most the whole campaign can make, reached if Tesco finishes at or above the new strike. "One SD lower" is the model's one-standard-deviation fall from 425p to that row's expiry (lognormal, IV 22%), so later expiries are tested against lower prices, which is the fair comparison: a longer roll has longer for the share to fall. Assigned shares are counted at that day's price. Model values of the new puts: November 430 18.97p, November 420 13.43p, December 420 16.17p, November 410 9.03p, January 410 13.86p.

What the menu shows

• Rolling out at the same strike is paid for time, not for being right. The November 430 put brings in +£47.50 and lifts the best case to £130.00, but the delta barely changes (−648 to −577) and the one-SD result is worse than letting the October put run, because the same in-the-money obligation now has 56 days to go wrong instead of 21.
• Moving the strike down costs money now and saves it later. The November 410 put is a £52.50 debit and caps the best case at £30.00, yet it has the smallest one-SD loss of any roll, −£170.90, and the lowest delta of any roll.
• Going further out makes a strike change look free. January 410 costs only £5.00 and keeps £77.50, which looks better than November 410 at every fixed share price. Tested at its own one-SD price (376.2p in January), it loses −£260.12.
• Doubling the number of contracts buys the credit with size. Two November 410 puts pay +£37.50 on the roll, and the obligation becomes 2,000 shares at 410p, £8,200 of stock instead of £4,300. It has the worst one-SD result of any row. Position size is set out on the sizing page.

A roll carries the loss forward; it does not erase it

The chart puts two of the rolls on one axis, Tesco's price on 20 November, next to the flat line of closing on 25 September. Each roll line is simply the new put's payoff moved down by the £60.00 already lost on the October put.

Closing turns an uncertain result into a known −£60.00. Rolling to the November 430 swaps that for a line that does better than closing if Tesco ends above 411p and worse below it, losing £10 more for every penny lower. The November 410 roll crosses the closing line at 401p, and the two rolls meet at 420p. None of the lines is better everywhere, and on the model's own prices none of them adds value: each new put is priced so that, before costs, what the writer collects equals the model's discounted value of what the put may pay out, as the methods page explains. A roll decides where the pounds land across the possible outcomes. It cannot change their average, and every roll adds costs.

The roll map shows the same menu as moves on the option chain: along the expiry axis the writer is paid for extra time, and down the strike axis the writer pays to move the obligation away from the share price. Filled points are rolls for a net credit, open points rolls for a net debit.

Six questions that separate a roll from a close, answered for the Tesco put

The model prices Tesco's results day like any other day. A results announcement usually lifts the implied volatility of the expiry that spans it, which would make both the buy-back and the new put dearer than shown; the results-day page sets out how that event premium is measured.

Rolling for a credit, and rolling for time

US retail options education commonly teaches a convention of rolling a short-premium position only when the roll pays a net credit. The reasoning behind it is that a trade entered to collect premium then never turns into one that keeps paying out, and a credit roll never requires adding cash to the position. The library shows rolls for a net credit where one exists and, where it shows a debit roll, adds the debit to the maximum loss; the methods page records where the convention comes from, and we could not find a study that tests it (checked 26 September 2026).

The trap is in what a credit can buy. On 25 September, with Tesco at 425p, a credit is available only by staying near the old strike (November 430, +£47.50) or by going out to December for a 10p move (+£20.00). No roll to the 410 strike pays a credit even four months out: January 410 is a £5.00 debit. A writer holding to the credit-only rule therefore keeps the strike close to the share price and lengthens the exposure, and the one-SD column shows what that costs: −£270.90 and −£273.19 against −£170.90 for the debit roll to November 410. The other way to find a credit is to add contracts, the worst row in the table. Here a £52.50 debit roll lowers the strike by 20p and would cut the maximum loss, if Tesco went to nothing, from £4,170.00 on the November 430 roll to £4,070.00, and leaves the campaign £100.00 better off at any price below 410p.

Rolling "for time" is the other label in common use: a roll whose main purpose is more days for the share to recover, usually out at the same strike. On a written option the extra time is paid for in the price of the new option, so it arrives as a credit; what it costs is exposure. On a bought option the reverse holds: rolling a long call out to a later month costs a debit, which is the price of the extra time. A debit roll of a spread raises its maximum loss by the debit, as the condor section below shows in pounds.

A roll only reshapes a losing position, and never repairs it, whenever the new option is fairly priced, which on the model it always is. The things a roll can genuinely change are the date of the risk, its size and its direction. The menu above prices those changes; which shape suits a given writer, if any, depends on circumstances this page cannot know.

What a roll costs in pounds

A roll crosses two bid-ask spreads and pays two commissions, before any assignment. The library's cost convention (set out on the methods page) counts half the quoted spread on each leg, each way, with commissions at IBKR UK's published rates (checked 26 September 2026). The spreads below are illustrative: live quotes on ICE single-stock options vary by share and by strike.

Two points stand out. First, on a 100-share mini the commission barely shrinks while the premium falls tenfold, so costs take a much larger share: £4.40 a roll against £13.00. IBKR's commission page shows no separate rate for minis (checked 26 September 2026), so £1.70 is the library's placeholder; ICE lists the mini, but whether a broker offers it and quotes a two-way price has to be checked. Second, costs repeat: a position rolled every month pays the two-leg cost each time, on top of the cost of opening the original trade. A roll moves no shares, so in normal ICE dealing no SDRT is charged on it; the 0.5% arises only if a put is assigned, when the writer buys the shares and pays 0.5% of the strike, £21.50 on one Tesco contract at 430p (who pays SDRT). Brokers generally let the two legs go in as one combination order at a net price; whether that fills nearer the middle of the two spreads depends on the market maker's quote on the day. The account type matters here: IBKR's account table lists covered calls and cash-backed puts as the option writing a cash account allows, and places strategy-based combinations under its margin account (checked 26 September 2026), so a writer in a cash account may have to place the two legs as separate orders, with the legging risk described below (account types and permissions).

Early assignment and legging risk in the middle of a roll

Two things can go wrong between deciding to roll and completing it.

The old option is assigned first. An ICE stock option can be exercised by 18:30 on any business day, so a writer of an American-style option can find the shares delivered before the roll is placed. For the Tesco put the tree puts the value of early exercise at 0.03p a share on 25 September (0.026p before rounding, £0.26 a contract): the 5.08p dividend still to come on 15 October makes waiting worth more to the holder than exercising, and the put still holds 9.27p of time value. After the ex-date, with the put deep in the money and little time value left, the balance turns. Calls behave the other way, with early exercise most likely just before an ex-date. The assignment page works the put thresholds and the call and dividend test in full, including the SDRT a private holder would pay. European-style FTSE 100 options (ESX) cannot be exercised early, so a FTSE roll carries no early-assignment risk.

Only one leg fills. Placed as two separate orders, a roll leaves a gap. Buying back first leaves the writer flat, with the new put's price free to move before it is sold: on the model, a 5p rise in Tesco to 430p would cut the November 430 put from 18.97p to 16.22p, and a 5p fall to 420p would lift it to 21.99p. Selling first leaves two written puts for a moment, twice the exposure and twice the collateral. A combination order avoids the gap, at the price of waiting for both legs to trade together. On a spread, an assignment on one leg mid-roll leaves shares plus the other leg overnight, which the spread-leg page works through.

Adjusting spreads and iron condors

A vertical spread is rolled as a unit: both legs are closed and a new pair opened, four legs in all. The width usually stays the same, so the maximum loss after the roll is the width less all the credit collected so far, including the loss taken on the old spread; a debit roll raises it by the debit. Margin for the new spread is set as for any vertical (spread margin). An iron condor gives more choices, because its two sides can be moved separately.

Model inputs. Entry on Monday 17 August 2026: FTSE 100 at 10,750 (a model level; the index closed between about 10,600 and 10,900 in August and September 2026), ICE FTSE 100 options (ESX, £10 a point, European, cash-settled on the EDSP) expiring Friday 16 October 2026 (60 days), Bank Rate 3.75%, dividend yield 3.05% (FTSE Russell factsheet, 28 August 2026), IV from the library's FTSE skew surface, IV(K) = 14.0% − 0.40 ln(K/10,750), held sticky by strike. The condor buys the 10,050 put, sells the 10,250 put, sells the 11,150 call and buys the 11,350 call: model values 56.96, 87.29, 79.60 and 35.53 points, filled on the 0.5-point tick for a credit of 74.5 points, £745.00. Adjustment day: Monday 21 September 2026, 25 days left, FTSE 100 assumed at 10,250, exactly on the short put strike. Commission of £1.70 a contract is left out of the table and counted in the costs section.

On 21 September the condor is almost flat: it would cost 75.0 points to close, a loss of £5.00, because 35 days of time decay have offset the fall. What has changed is the risk. At entry the position's delta was close to zero; now it is £1.67 a point long, so every 100-point fall in the index costs about £167, and the short put is at the money with 25 days to go.

Probabilities are model probabilities (risk-neutral, lognormal, from the model skew surface), from 10,250 with 25 days left. Model values of the new spreads: calls 10,550/10,750 35.96 points, calls 10,250/10,450 88.31, November puts 9,950/9,750 47.50. Each new spread's value comes from the same skew surface, so the new call strikes are priced at their own, lower implied volatilities. After the £6.80 commission for opening the four legs, the unchanged condor's maximum profit is £738.20 and its maximum loss £1,261.80; each adjustment in the table trades four more legs, another £6.80 before half-spreads.

The trade-offs are visible in the chart. Rolling the untested calls down collects £355.00 more, taking the total credit to 110.0 points, which lowers the maximum loss on both sides from £1,255.00 to £900.00 and brings the delta back to +£0.38 a point. The price is a profit zone that shrinks from about 1,050 points wide to 520, and a new loss if the index rebounds above 10,660 by expiry, a level the original condor would have treated as comfortable. Converting to an iron butterfly takes this to its limit: the total credit reaches 162.5 points, the maximum loss falls to £375.00, but the full £1,625.00 is paid only if the index settles exactly at 10,250, and the model probability of ending anywhere between the breakevens drops to 29.7%. Rolling the tested puts down and out is the one debit: £270.00 buys another month and 300 points of room, and raises the maximum loss to £1,525.00.

None of these is a repair. Each is a different bet placed on 21 September, and the chart and the table are the whole of the comparison. Stress testing such positions against gaps and volatility rises is set out on the Level 3 page, and the sticky-strike assumption behind these prices on the implied-volatility page.

UK tax of a roll: two trades, two rules

For a UK individual investing (not trading), listed options fall under capital gains tax, and the two halves of a roll are taxed under different sections of the Taxation of Chargeable Gains Act 1992 (TCGA). The canonical statement of each rule, with HMRC's manual references, is on the tax worked-examples page; this section applies them to rolls.

• Buying back a written option is not a disposal. For a traded option, the closing purchase is disregarded and its price and commission count as extra incidental costs of the original grant (TCGA 1992 s148; HMRC CG55545). The grant's gain falls or becomes a loss, and the result stays dated on the day the option was written, in that tax year. The relief applies only to traded options, meaning options listed on a recognised stock or futures exchange (s144(8)), so not to CFD-style options.
• Writing the new option is a new grant, a disposal of the option itself dated on the day it is written (s144(1); CG55536). If it lapses, that gain stands; if it is bought back in turn, s148 revises it; if a put is assigned, the grant and the share purchase become one transaction and the premium reduces the cost of the shares (s144(2)).
• Selling a bought option is an ordinary disposal on the sale date, and a bought option that lapses is a disposal at expiry for a loss equal to its cost (s144(4); CG55415). The bought legs of spreads and hedges follow this rule.
• So each roll of a written option adds one computation and revises another. The counting rule for whole positions is on the tax page.

The Tesco roll on the tax return

The computations below follow the November 430 roll. Both grants fall in 2026/27, with £1.40 commission on each trade.

The assigned branch shows a point specific to UK rules. Economically the campaign is one trade, but the tax splits it: the October put's loss is an allowable loss now, set against other 2026/27 gains, while the November premium sits inside the base cost of the shares until they are sold. Counting SDRT in that base cost is the library's reading of HMRC's incidental-costs rule, which names stamp duty but not SDRT. If the writer had sold Tesco shares in the 30 days before 20 November, the shares received on assignment would be matched with that sale first, changing its gain or loss (s106A; CG51560).

A FTSE 100 put spread rolled across 5 April

The rules produce their least intuitive result when a roll crosses the end of the tax year, because the two legs of one spread can land in different years. The tax page works a single written put bought back after 5 April (Example 2); a spread adds a second leg that follows a different rule.

Model inputs. Monday 15 March 2027: FTSE 100 assumed at 10,750 (the model level carried forward; no forecast), ESX expiring Friday 16 April 2027 (32 days), the same skew surface, Bank Rate and dividend yield as above. The spread sells the 10,450 put (model 74.12, filled at 74.0) and buys the 10,250 put (model 39.19, filled at 39.0) for a credit of 35.0 points, £350.00. Roll day: Thursday 8 April 2027, 8 days left, in the new tax year. Commission £1.70 a contract.

If the index has fallen to 10,350, the spread is tested. Its legs are worth 150.21 and 54.29 points (filled at 150.0 and 54.5), so closing costs 95.5 points, a result of −£605.00 on the April spread. The roll writes the same strikes for Friday 21 May 2027 (43 days) at 263.0 and 173.5 (model 263.20 and 173.51), a credit of 89.5: a roll for time at a net debit of 6.0 points, £60.00.

If the index has risen to 10,950, the short put is bought back for 1.5 points (model 1.58), and the long put, worth 0.18 of a point, has no bid on the 0.5-point tick, so it is left to lapse on 16 April. The roll writes a May 10,700/10,500 put spread for 43.0 points (model 43.31).

In the down branch the written leg's −£763.40 travels back to 2026/27, where it is set against that year's other gains or carried forward, while the bought leg's +£151.60 lands in 2027/28. The online return for 2026/27 is due by 31 January 2028 (GOV.UK filing deadlines: 31 January after the tax year ends), so on 8 April 2027 it has normally not been filed and the revised March grant simply goes on it; a correction would be needed only if that return had already gone in. In the up branch the timing works against the writer: the £721.60 gain stays in 2026/27, taxable at 18% or 24% (£129.89 or £173.18, assuming the £3,000 annual exempt amount is used by other gains), while the long put's −£391.70 arises in 2027/28. Capital losses cannot be carried back, so that loss is worth £70.51 or £94.01 only when set against 2027/28 or later gains; it is deferred relief, not lost relief, if such gains arrive. We could not find an HMRC worked example of a buy-back that falls in a later tax year than the grant (checked 26 September 2026); the treatment here is derived from s148, which dates the buy-back cost with the grant, and from the ordinary rule for bought options. More on positions that straddle the year end is on the tax page.

A US covered call rolled up: two dates, two exchange rates

Model inputs. A hypothetical US share (not a real company) held as 100 shares, trading at $150 on Monday 17 August 2026. The holder writes the $160 call expiring Friday 16 October 2026 (60 days), IV 30%, US rate 3.625% (the midpoint of the Federal Reserve's 3.50% to 3.75% target range on the model date), no dividend in the window: model value $3.82, filled at $3.80 on the $0.10 tick. Roll day: Friday 18 September 2026, share assumed at $168, 28 days left, IV and the US rate held at their model-date values (the Federal Reserve raised its range to 3.75% to 4.00% on 17 September; at that rate both fills would be the same). The October call is worth $10.66, bought back at $10.70; the November $175 call (63 days) is worth $5.86, written at $5.90. Commission $1.00 a trade (the order minimum). Exchange rates are illustrative: 1.3559 dollars per pound on 17 August and 1.3344 on 18 September, the cross of the European Central Bank's euro reference rates on each day; the broker's own rate on the day is an equally reasonable source.

The roll lifts the price at which the shares can be called away from $160 to $175, for a net debit of $4.80 a share, $480 on the contract. There is no reason for early exercise here: the October call still holds $2.66 of time value and no dividend is due, but an in-the-money call with an ex-date before expiry is a different matter (see the early-assignment section above). If the call were assigned instead, the premium would be added to the sale proceeds of the shares under s144(2)(a) and no buy-back would take place.

Converting the net −$692.00 at one rate is the method HMRC rejects (CG78310): at 18 September's rate it gives −£518.59, £4.50 less loss than the correct figure. The gap is small here because the pound moved only 1.6% in a month; it grows with the size of the trade and the move in the exchange rate. HMRC prescribes no source for the rate, only a reasonable and consistent one, as the tax page's FX section explains. Note also what the roll does to timing: the −£523.09 loss on the call is realised in 2026/27, while the matching rise in the shares stays unrealised until they are sold. The targeted anti-avoidance rule can disallow a loss that comes from an arrangement whose main purpose is a tax advantage (s16A, CG13350). Trading US options from the UK, including hours and withholding, is covered on the US options page.

Matching rules when a leg is bought back in the same series

A roll to a different strike or expiry creates a different series (one type, on one share, at one strike and one expiry), so the new option never matches with the old. The matching rules bite when a bought leg is sold and the same series is bought again, which happens when a long leg is closed to take a loss and then re-established. HMRC's manual treats options of one series as a single pooled holding, which brings them within the same-day and 30-day identification rules (CG55535). If a FTSE 10,250 put bought earlier is sold at a loss on 30 March and the same series is bought again on 8 April, the 30 March sale is matched with the 8 April purchase rather than with the original cost, so the loss the investor expected to bank in 2026/27 becomes the difference between those two prices. The manual is silent on grants, so reading the pooling rule as covering bought options only is an inference, not HMRC text. Worked cases, including shares received on assignment inside 30 days, are on the tax page.

None of this applies inside an ISA, which cannot hold options at all; the wrappers page sets out where options can be held. The reporting boxes for these computations are on the SA108 page. Because the cross-year treatment above is derived from the statute rather than taken from an HMRC example, a tax adviser's view may be worth having on a large cross-year position.

Strategy pages that apply these rolls

Each of these pages keeps only its own roll triggers and numbers, and links back here for the mechanics and the tax:

How these numbers are calculated