The Dispatch
Notes on the strategies we're working on, markets we're watching, and what's happening inside the firm.
Washington declared an “economic D-Day” on Iran and doubled its long-bond buyback in the same week. The dollar shrugged, oil fell, and gold - near a record - barely blinked. It was listening to the quieter announcement: a buyback that hasn’t bought a single bond yet.
Fintwit says the Bloomberg Commodity Index just cleared a multi-year cup-and-handle. On the tape, its proxies are still 2-4% below their May highs - pressing resistance, not through it. What's really driving it, and what would actually confirm it.
The 30-year Treasury made a new post-2007 high — while the two-year fell. That combination is the whole story.
July core CPI helped. It did not settle the higher-for-longer question.
The Fed held rates and hiked its rhetoric. But the real tell was the bond market: Treasuries sold off as stocks fell - an inflationary, higher-for-longer repricing, not a growth scare. Our read on the July FOMC.
The Iran ceasefire collapsed and crude spiked. Equities shrugged — and the demand side explains why.
The equity market is pricing in a soft landing. Credit, banks, and commodities are not. Our May 2026 cross-asset read on where the signals align — and where they break.
A popular rule says buy your favorite stocks when the Nasdaq drops 10%. We backtested 27 years of QQQ — a standing weekly buy that never looks beat it by roughly two to one.
A 50-EMA session breakout on forex and gold, tested four ways across 16 years of hourly data — plus the one statistical test that tells you an entry is worthless before you spend a month tuning it.
An RSI-divergence idea on the Nasdaq 100, and a case study in how a backtest lies before it tells the truth — data artifacts, Monte Carlo blind spots, and the search for a real edge.
Shardul Mehetar joins our Systematic Strategies team, bringing a healthy suspicion of backtests that look too good to be true.
Chang and Alice maintain their Hold ratings on $COIN and $NVDA heading into back-to-back earnings in May.
If you've spent meaningful time in systematic or quantitative strategies, we'd love to connect you with Chang and Alice.
Chang Liu starts Coinbase at Hold and Alice Li starts Nvidia at Hold - two theses, twelve weeks to see how they play out.
Applying Karpathy's autoresearch loop to quantitative strategy development — 30 autonomous experiments, zero human intervention.
From internal research tool to open-source project - the story behind July Backtester, our professional-grade Python backtesting engine.
Chang Liu and Alice Li join our Quantitative Research group this spring.