5 August 2024: USDJPY from 161.95 in July to 141.7, and the Monday the Nikkei fell 12.4%

[Fairman]

5 August 2024: USDJPY from 161.95 in July to 141.7, and the Monday the Nikkei fell 12.4%

A short case file on a day when the yen did the unexpected.

USDJPY topped at 161.95 on 3 July 2024. On 31 July the Bank of Japan raised its policy rate to 0.25% and said it would cut back its bond purchases. On Friday 2 August US payrolls came in well below forecast at 114,000, with unemployment at 4.3%, and the S&P 500 lost 1.8%.

On Monday 5 August the Nikkei fell 12.4% to 31,458.42, the biggest one day drop since the 1987 crash, and USDJPY touched 141.7, a seven month low. The BIS later described a yen funded carry trade unwind, with the VIX above 60 in off hours trading. The Nikkei rebounded more than 10% the next day.

Some arithmetic for a scalper. From 161.95 to 141.7 is 20.25 yen, which is 2,025 pips in about five weeks. One lot of USDJPY gains or loses 1,000 yen a pip, so a lot long from the top would be down 2,025,000 yen, about $14,300 at 141.7.

Closer to home, a 0.5 lot long with a 10 pip stop risks about $35 at those levels. If the stop fills 100 pips away because the market gaps, the loss is about $352. Ten times the plan, from one slippage event.

Nothing on the calendar said Monday morning would be that bad. The scheduled events were earlier in the week; the damage was the market's reaction to them.

[FXS HongKong Scalper]

That Monday was brutal in Asia. The carry trade unwind hit everything funded in yen, and AUDJPY and MXNJPY fell even harder than USDJPY. From Hong Kong we watched the Nikkei open sharply lower and keep falling, and spreads on the yen crosses were very wide through the morning. One lesson I took away: positions funded in yen can look safe for months and then unwind in days once the BoJ shifts, so the size of yen crosses held overnight should reflect that risk.

[Fairman]

The BIS bulletin gives a sense of how crowded the trade was. Speculative net short positions in yen futures had reached about 2 trillion yen, roughly $14 billion, a historical peak, and were unwound during the event. Futures are only part of it, since much larger carry positions sit in over the counter products. The carry sum shows why the unwind hurts: with US rates around 5.25% to 5.5% and Japan close to zero, the yield gap is roughly 5 points a year, about 0.42% a month. USDJPY falling from 161.95 to 141.7 is 12.5%, which is around 30 months of carry income lost in about five weeks.

[Fairman]

Most of the fall was retraced fast. The BIS notes that markets stabilised almost as quickly as they broke, with the S&P 500 recovering all its Monday losses by Friday 9 August. USDJPY was back in the mid 140s within the week. A trader who sold the breakdown near the low saw that bounce, and one who bought the low needed a stop wide enough to survive the spike. Both were right about something and wrong about timing. The thing the day shows is how the speed of the move compared with how a normal stop is sized: the whole 20 yen slide happened in sessions, not months, and the final part came in hours.

[Fairman]

A pip is not a fixed amount of money. One lot of USDJPY is 100,000 units, so a pip of 0.01 yen is 1,000 yen, and converted to dollars at 161.95 that's $6.17, while at 141.7 it's $7.06. The same 10 pip stop costs 14% more in dollars after the yen rallied. A fixed lot size therefore quietly raises the risk as the pair falls. Sizing from risk each time, using the current price, keeps the dollar figure steady. It's a small point next to a 2,000 pip slide, but it's also the sort of error that goes unnoticed for months.

[Fairman]

There's a useful exercise for the account you actually trade. Export M1 bars for that Monday, the 5th, from your broker. Record the widest spread you can find, the biggest one minute range, and whether your platform's candles show gaps. If your stop was 8 pips and the largest M1 candle was 60 pips, that candle is the real stress test for your sizing, because a stop only works if there's a price to fill at. Writing that number in your plan, something like worst recent minute 60 pips, gives the position sizing a concrete figure to be measured against.