Macro themes, CPI and FX Carry

CPI tomorrow

The FX setup into CPI looks symmetrical to me as market positioning is not extreme and conviction is low. Out of office replies are near their seasonal maximum and then we will soon enter the period where everyone starts to gear up for the stretch run and markets re-engage and look for themes. Right now, the macro themes are somewhat familiar.

• US economy on autopilot. Antifragile to the point of seeming almost indestructible as AI Capex and monster deficits provide fuel.
• US dollar midrange while FX carry is rolling.
• Inflation is above target but not ripping higher as oil settles into its new range following the war in the Middle East.
• AI Capex mania dissipated in the stock market, memory stocks have been cut nearly in half, but AI Capex continues to grow. Companies are still pouring incomprehensible amounts of money into AI infrastructure. The old money printing machines like Google, Amazon, and Meta are flipping to cashflow negative capex machines.
• Issuance deluge and mad deficits keep the pressure on yields, but bond vigilantes don’t get much love as US 10-year yield is trading at levels we already saw in 2023, 2024, and 2025. Hardly terrifying stuff. More like: Normalization back to pre-GFC yield levels. If the US 10-year were to take out 5.0%, the complacency would vanish quickly, but for now it feels like bonds found their new equilibrium level three years ago and that equilibrium is astonishingly stable (3.80%/4.80% range with a few short expeditions outside of that).
• Fed is thinking about maybe fiddling with interest rates but there is no meaningful rate hike cycle on the horizon. They might not even hike rates once. Warsh wears wolf’s clothing but is sheepish about hiking.
• Equity breadth was bad for many years, and now it’s good. Bull market either way. Breadth isn’t a directional or predictive indicator. It’s a descriptive one.
• Coordinated intervention in JPY makes for some fun tactical trades, but US Treasury does not have enough credibility to turn the tide right away. To be fair, the 1998 and 2011 coordinated interventions did not turn the tide right away, either. Sell zone is 160.00/50.

So that is the basic setup here going into CPI. September is priced 60/40 for no change, which is quite something given July got near 50% at one point. In other words, we could have had two hikes by the end of September and now we’re probably getting zero. The entire “new sheriff in town” story was a red herring. In the next section, I will go through some thoughts about CPI.

---

CPI quick thoughts

1. As a starting point, the base case should be that Core CPI comes in as expected. That is the case around 37% of the time for both series.
2. Expectations have been softening over time. When last month’s CPI came out, this month’s repriced, and you can see that odds of 0.2% have been rising while odds of 0.3% have been falling. Given the volatility of policy and oil prices, it’s not crazy to be ready for 0.0% or 0.4%. Out of 64 economists, exactly one is calling for 0.0% and one is calling for 0.4%. Meanwhile 48 of 64 are calling for the 0.2%.

3. I had a theory: The last CPI release before FOMC should move USDJPY and gold (for example) more than CPI releases that do not precede an FOMC meeting. What I mean is, tomorrow’s CPI is not the last report before the Fed meets, so it should have less impact than the one next month because it’s right at decision time.

Sadly, my theory is not wrong, but it’s not right enough to make a difference. Here is the average move in USDJPY and in the six hours after a CPI release, split by pre-FOMC CPIs or not pre-FOMC CPIs. You can see there is no meaningful difference whether CPI is right before FOMC, or not. Just thought you should know. Same deal with gold.

---

Some quick points on FX carry

FX carry is boring to talk about at parties, but it can generate significant uncorrelated returns for patient investors. For example, there is the quite stunning fact that MXNJPY (including carry) outperformed the NASDAQ over the past five years. Most of this outperformance came when the Fed was hiking, but it’s still a crazy factoid (via Tim Power).

This analysis comes with the benefit of hindsight, of course, and the yen carry trade could be over as the BOJ hikes rates and the powers that be attempt to defend the JPY. Moving forward, what might be the next high-Sharpe carry trade that you can add to your portfolio for diversification and high nominal returns? Here are some very quick thoughts on some funders and high-yielders.

This is another good example of: I write to find out what I think. Two hours ago, I had no view on AUD and certainly no view on AUDCNH. Now, I like that pair as a nice carry trade and you can do 2-month AUDCNH options (for example) if you work at a hedge fund. This view will always be subject to change if PBoC policy preferences change, but for now CNH is a low-vol, low-yielding thing that isn’t going anywhere, and AUD is a high-yielding thing that probably isn’t going anywhere but could easily go higher.

---

Final thoughts

1. If you are a short-term trader, collect daily data on your trading. That’s the best way to really, truly understand variance. But this table might help a tiny bit. If you have a 50% daily win rate and trade 252 days/year, there is a 99% chance you will have a 5-day losing streak in a given year. There is a 63% chance you will have a 7-day losing streak. One in ten years, you will have a 10-day losing streak.

Two things. First, a 5-day drawdown is essentially guaranteed at any hit rate below 60%. It carries zero information about whether your edge is broken. It’s guaranteed variance. Second, the curve is brutally sensitive: dropping from a 60% win rate to 40% doubles the expected worst streak, but the tail moves much more. A 12-day skid is a 1-in-400 event at 60% and a 1-in-5 event at 40%.

Caveat on the independence assumption: P&L streaks can cluster if your trading is regime-dependent. Empirical streaks can run longer than this table implies in that case. Treat these as a floor on what to expect. If you have enough data, you can eventually figure out if your trading has serial correlation or if each day is truly independent. Some people even have negative serial correlation because they trade better when they are in drawdown.

That is the case for me. I trade better when in drawdown because I stick to my process and I am more selective. When I am overearning, or at high water, I sometimes take negative EV risks for fun, or oversize due to having cartoon dollar sign eyeballs. So, my losing streaks are shorter than math would suggest, but my winning streaks often end with a bang.

2. If you have been following the VCX law of one price violation story… The lockup on that stock expires August 13. Unlike SPCX, there is no reason for VCX to bounce once the lockup expires—it should simply compress right towards NAV and stay there. NAV is around $18/$20. Wild ride for that thing as it peaked once above $500, once above $300 and now trades at $32. Daft.

I was listening to Nighttrain by Guns n’ Roses yesterday, and it got me wondering about the relationship between number of cowbell strikes and number of Spotify streams. The scatter above shows it.

61% is the r-squared. Once again proving a point made many times in am/FX. R-squared in isolation is one of the stupidest measures in statistics.   :]

Congratulations to Wild Cherry and Twisted Sister for maximizing their streams/cowbell ratio with “Play That Funky Music” and “We’re Not Gonna Take It” respectively.

good luck ⇅ be nimble