Pounded By The Pound: Sports Direct's Hedges and the Cable Flash Crash
Just after midnight London time on October 7th, 2016, GBPUSD (henceforth “cable”, henceforth “betty”1) abruptly plunged from 1.2620 to
1.1841 1.1938 1.1378 1.1500 1.14912 in the space of about three minutes, then just as quickly bounced back to the mid-1.20s. Explanations for the move ranged from “algorithms!” to “fat fingers!” to “stop-losses being executed in the typically thin, illiquid markets that prevail between NY and Tokyo timezones” to “hard Brexit” to, probably, “aliens landing on the grounds of Buckingham Palace!”.
Later in the morning (London time), UK retailer Sports Direct issued a terse RNS release updating its FY17 financial guidance:
In the Trading Update provided on 7 September 2016, the Company stated that it expected FY17 Underlying EBITDA to be in the region of £300m. This guidance was based on a GBP/USD rate of approximately 1.30.
In light of recent downward currency movements, the Company entered into a hedging arrangement with respect to the GBP/USD rate. Extreme movements overnight resulted in a crystallisation of that rate at 1.19, resulting in a negative impact of approximately £15m on the Company’s FY17 Underlying EBITDA expectation.
In addition, after taking into account the hedging referred to above, if the GBP/USD rate is 1.20 on average for the remainder of FY17, then the negative impact on the Company’s FY17 Underlying EBITDA expectation would be in the order of a further £20m.
The implication of this statement is that Sports Direct diligently hedged its cable exposures when spot was up around 1.3000, but a barrier embedded in their hedging kicked in when cable traded through 1.1900, resetting the hedge rate from up around 1.3000 to 1.1900. This triggered the crystallisation of the GBP15mio loss reported in the RNS message.
From this, we can attempt to guess at what the Sports Direct hedging structure was; and from that, we can make an educated guess at the effect of the embedded barrier on the cable spot market. This post attempts to estimate the size of that effect.
Sample SportsDirect hedge structure
Unless SportsDirect themselves discloses the structure of their hedge book, we can’t know the exact structure that they were using to hedge their FY17 exposures. We can guess, though, and we can use a common FX hedging structure to approximate what their hedge might have been.
The RNS statement suggests that Sports Direct was previously hedged at 1.30, and then their hedge reset down to 1.19. If an 11-big-figure move in cable cost them GBP15mio (valued at 1.19), then their hedge notional must have been about GBP 160 mio3. There are nine months left in Sports Direct’s FY17 (they run July-June), so let’s assume that the GBP 160mio notional is spread over nine months – call it GBP 18 mio/month.
Let’s assume that SportsDirect was using a simple (simple to explain, that is – not so simple to manage!) bonus-forward-style hedge. We’ll assume it was structured something like the following:
If GBPUSD never trades at or below 1.1900, Sports Direct sells GBP 18 mio per month for twelve months at 1.3000. If GBPUSD trades at 1.1900 at any time during the life of the hedge, then for all subsequent months, Sports Direct sells GBP 18 mio at 1.1900.
This is a variation of a bonus forward – a forward where the rate changes, depending on whether spot stays inside or breaks out of a range4.
This trade can be priced as a strip of long one-touches (long from the bank’s perspective), every month for twelve months. Our assumptions:
- The notional of each one-touch is USD 1.98 mio – 11 big figures (1.30-1.19) on GBP 18 mio;
- Also, the notionals are equally distributed.
- The one-touches deliver at end-of-month, and expire two business days before month-end;
- The one-touches are all pay-at-maturity;
- We’re going to use market data from October 6th, the day before the plunge.
Estimating the delta gap
One nasty feature of barrier (and touch) options is that their delta changes dramatically when you get close to the barrier: in this case, it goes to zero, because the options change from one-touch options to guaranteed USD cashflows with no FX risk5.
From the bank’s point of view, these options give the bank a short cable delta position. Simply put: if cable goes down to 1.19, the bank makes quite a lot of money (assuming they left the position unhedged, which in practice wouldn’t happen); if cable doesn’t go down, or goes down but doesn’t make it to 1.19, then the bank loses the premium it paid for the one-touches and doesn’t get its tens of millions of dollars. Effectively, the bank makes money on a mark-to-market basis if cable goes down, and loses if it goes up: if you squint, this a lot like just being short cable.
And as spot goes down toward the barrier, you get shorter and shorter and shorter delta from the options. In practice, a responsible risk manager isn’t going to leave that risk un-hedged – they’ll be buying spot all the way down to hedge against a possible bounce in sterling, and generally keeping their book delta-hedged6. This is fine on the way down, because the delta smoothly grows – though the rate of growth itself grows faster and faster as spot approaches the barrier.
Then, when the barrier triggers, the delta position abruptly snaps from “short quite a lot” to “nothing”. The responsible risk manager, who’s been buying all the way down, suddenly has a lot of cable that she’s bought, and no delta from the option against it – so she needs to turn around and sell all the delta that she’s bought. The delta needs to be sold as fast as possible, and as close to the barrier level as possible, because any slippage beyond the barrier level is negative PnL for the trader.
The easiest way to do this is a stop-loss order. The options desk says to the spot desk “If 1.1900 trades, I need to sell x million pounds as fast as possible. Don’t try to hold me in or anything fancy like that; it’s a one-touch stop, if it trades there, just do it”. Normally, no-limit stop-loss orders are a poor idea because they run the risk of uncontrolled selling into liquidity holes – but when a barrier’s triggered, and the responsible risk manager needs to sell a lot of delta, it’s a sensible way to manage the position.
The question in the Sports Direct case: how big was the barrier stop-loss order?
Here are two estimates: a worst-case scenario, and a more likely scenario.
Worst Case Scenario: GBP 240 million
The worst-case scenario occurs if the cable options trader has been diligently buying all the way down to 1.1900, then immediately has to turn around and sell the lot.
If we price these trades with spot just above the barrier, then we reprice them with spot just through the barrier, then the difference between the delta at those two levels is the amount of spot that the trader will have to sell.
Pricing these options using the prevailing market data, and a standard SLV pricing model, gives a change in delta of GBP 242,031,1867. In a worst-case scenario, this is how much the cable options trader would have to sell as soon as 1.1900 traded.
Likely Scenario: GBP 85-90 million
In practice, though, the selling was likely to be a lot smaller than this.
The aim of the stop-loss selling is for the options trader to sell out all the delta they’ve bought on the way down. Because the move downward from 1.2600 to 1.1900 and beyond happened within minutes, it’s likely that the options trader didn’t buy any extra delta below 1.2600, and only had to sell out the amount they’d bought before the plunge started.
Pricing the trades as of October 6th, with spot way up at 1.2600, the delta on the trade (smile delta, if you’re wondering) is about GBP 87 mio short. If the risk manager was hedging responsibly, this is how much she would have bought before the plunge.
Then, when the barrier was abruptly triggered, the trader would only have needed to re-sell GBP 85-90 million to clear out the delta hedges from the Sports Direct trade.
Depending on their hedging strategy, in the early hours of October 7th, the bank that took the other side of Sports Direct’s ill-fated hedge book would have had to sell GBP 85-90 million in a likely scenario, or GBP 240 mio in a worst-case scenario, on a no-limit stop-loss basis as soon as GBPUSD traded at 1.1900.
Whether the smaller or larger number is closer to reality, this would have been deeply disruptive to the cable market. The FX “witching hour” (after New York goes home, and just as Tokyo and Singapore and Sydney are having their morning coffee) has the thinnest liquidity of any time of day, and liquidity providers (automated or manual) would have been withdrawing quotes in the face of a market that had already plunged seven big figures.
It’s plausible (though impossible to prove without trading records) that this stop-loss order contributed to the final capitulation down to 1.1378 1.1500 1.1491, but GBP was already running out of control by the time this order hit the market; at the risk of descending into cliché, this order didn’t light the match; it added fuel to a garbage fire that was already roaring.
- FX market vernacular owes a lot to Cockney rhyming slang. GBPUSD -> “cable” (for the old transatlantic cable between London and New York -> “Betty Grable” (because why not?) -> “Betty”. Confession time: the first time I heard GBPUSD called “Betty” I thought it was a reference to Elizabeth II on the banknotes; this is one of those rare times when the real explanation for something is far more interesting than first imagined.) [return]
- The decentralized (and mostly OTC) nature of spot FX markets makes it tough to call an “official” low, and all of the above were cited as “the low” of the cable move. Bloomberg’s composite quote feed saw a low of 1.1841, but it definitely traded below there. Icap’s EBS platform saw a low traded price of 1.1938, but for historical reasons, EBS is not the biggest or most active trading venue for pounds. Reuters Dealing, the traditional home of cable, initially reported 1.1378 as the low trade; that trade was later busted, and the new reported low was 1.1491 in small size, with 1.1500 being the low in “market size” (i.e. multiple trades occurred at that level, for a total of more than GBP 5 mio notional). [return]
- GBP 15 mio @ 1.19 = USD 17.85mio. One pip on GBP 1 mio = USD 100 of PnL; 1100 pips on GBP x mio = USD 17.85 mio; x = GBP 162 mio. I’m assuming they’re valuing their hedge off 1.30 vs 1.19, rather than exactly marking the options to market. [return]
- They might also have been using a “smart forward”, where the customer’s long an option initially instead of a forward, but the trade again knocks into a forward if the barrier gets hit. The end result doesn’t change. [return]
- To be really picky: USD cashflows can have FX risk if you’re not accounting in USD and you choose to manage your own translation risk. We’ve assumed the bank is accounting in USD. [return]
- This is why it’s wrong to say “if Sports Direct lost GBP15mio, the bank’s exotics traders must have made GBP15mio”. The exotics desk is long a huge chunk of sterling against the one-touch option – enough that if the barrier does trigger, they’ll have lost close to GBP 15 mio that will be offset by the payout from the touch. The absolute worst-case scenario, as you can probably guess, is that spot goes down to 1.1901 but never triggers the barrier, inflicting huge losses on the delta hedge with no offsetting gain from the one-touch payout. This happened to me a lot. It’s not fun. [return]
- This figure is overly precise: realistically the trader would probably enter a stoploss order for GBP 240 mio. [return]