DISCLAIMER: This note is intended for US recipients only and, in particular, is not directed at, nor intended to be relied upon by any UK recipients. Any information or analysis in this note is not an offer to sell or the solicitation of an offer to buy any securities. Nothing in this note is intended to be investment advice and nor should it be relied upon to make investment decisions. Read our full disclaimer, here.

The YX Insights Signal Family

by Alex King, CEO, Cestrian Capital Research, Inc

In today’s market, everyone has an opinion. The trouble is, most of the time the opinion is just that; opinion. The YX Insights Signal Family was developed by my colleague Yimin Xu to take advantage of machine learning systems’ ability to analyze price in an ice-cold manner. His YX Insights Signals Model underpins three signal services in the Signal Family, covering commodities, major stocks and index ETFs, and crypto.

Please take a moment to hear how he does it.

Yimin's background is unusual among our colleagues, and I think that's precisely why his models work as well as they do. Prior to starting his own research firm, Yimin had spent years as a G10 FX and Interest Rate Derivatives market maker at NatWest Markets (formerly RBS), focused on macro cross-asset flows and central bank policy dynamics. He's an ICAEW Chartered Accountant, holds two degrees from the London School of Economics, and serves as Lead FOMC Commentator at Seeking Alpha. He brings the kind of institutional-grade macro understanding that most retail signal providers simply don't have.

Like all the best quant developers I've worked with, Yimin comes at markets through a deeply practical lens. He's made prices on an institutional desk. He's traded through rate cycles, FX shocks, and liquidity squeezes in real time. That lived experience is baked into the models.

How The Models Work

The YXI models are built on a principle that resonates with how we think about markets here at Cestrian: asset prices are driven by a combination of macro forces and asset-specific dynamics, and the relative importance of those drivers shifts over time.

Each model ingests two categories of input. The first is macro factors: interest rates, credit conditions, liquidity, volatility, dollar dynamics, and more. These capture the environment. The second is asset-specific price features, such as momentum, mean-reversion tendencies, volatility regime, and trend structure. These capture how the individual asset is behaving within that environment.

The models are trained on multiple years of historical data, but they are not static. The models are continuously recalibrated on a rolling basis so they adapt to current regime shifts. A feature that mattered enormously during the 2022 rate-hiking cycle may carry less weight in a Fed-on-hold environment, and the models adjust accordingly. This makes them dynamic rather than backward-looking.

The output is a binary signal per ticker each day: Risk On (the model holds or enters the position) or Risk Off (the model exits to cash). No leverage, no options, no complex position sizing. Clean and simple.

The models were recently upgraded to reduce whipsaw and position turnover. The latest generation filters out short-term noise while still reacting quickly to genuine regime shifts. The result is fewer trades, longer average holding periods, and better net performance after transaction costs. The models remain in positions when the underlying trend is intact and exit only when macro or asset-specific evidence genuinely deteriorates.

What Gets Delivered Every Day

Signals are delivered daily via Slack before the US market open. Each morning, subscribers receive three posts in their channel:

1. Model Signals Snapshot — The current signal for every ticker in the suite: Risk On or Risk Off, with the date and any position changes clearly flagged.

2. Signal Commentary — And this, in my view, is the real differentiator. Every day, the commentary explains what the top drivers are behind the signals.

Which macro factors are dominating? Which asset-specific features are firing? Why did a signal flip, or why is it holding? This is not a black box. You see the reasoning, and over time, reading the commentary genuinely improves your understanding of cross-asset dynamics and regime shifts, regardless of whether you follow the signals mechanically. Many subscribers have told us the commentary alone is worth the subscription. I agree with them.

3. Position History Table — A running log of every signal change: entry dates, exit dates, and the outcome for each position. Full transparency on the track record.

  1. Asset Intraday Price Profile (Shared W`eekly) — A visual review of the intraday price patterns of every ticker in the past 5 days, to improve execution timing. Our subscribers have provided extremely positive feedback on how this knowledge has improved their trading.

The Three YXI Signal Services

Yimin runs three distinct services on our platform, each covering a different asset class with its own model configuration and feature set. Let me walk you through each.

YX Insights — Commodities Signals

Tickers: GLD, GDX, SLV, SIL, PPLT, PALL, USO, CPER, URA, BCI

Ten tickers spanning precious metals, energy, industrial metals, uranium, and broad commodities.

Commodities have strong regime persistence. When gold is bid on central bank demand and falling real rates, it tends to stay bid for weeks, not hours. When oil is caught in a supply shock, the trend sustains. The models exploit that structure, and the macro factor inputs are particularly attuned to the drivers that matter in commodity markets: real rates, USD strength, inflation breakevens, central bank policy expectations, and cross-commodity momentum.

Yimin updates the performance numbers every Monday morning. Since launching the service in November, the Commodities Service has seen a P&L of +25%, a Sharpe ratio of 1.85, and a Max Drawdown of -16%.

The backtest results using the latest upgraded models (they were upgraded in late March):

Learn More — YX Insights Commodities Signals

YXI Crypto

Tickers: BTC, ETH, SOL, XRP, MSTR, COIN, HOOD

Seven tickers covering Layer-1 protocols and crypto-adjacent equities.

Crypto is volatile. That's not news. What is worth noting is that the primary value of these models is keeping capital out of the worst sell-offs while capturing the majority of upside moves. The feature engineering reflects crypto's unique dynamics: the relationship between crypto and equity risk appetite, BTC dominance shifts, volatility clustering, and the fact that crypto doesn't always trade like a risk asset. Sometimes it leads equities, sometimes it lags, and sometimes it decorrelates entirely. The models adapt to whichever regime is in play.

The crypto-equity names (COIN, HOOD, MSTR) add another dimension. These trade with both crypto beta and equity market structure, and the models capture that dual sensitivity.

The backtest results using the latest upgraded models (they were upgraded in late March):

Learn More — YXI Crypto

YXI Macro & Mag-7

Tickers: SPY, QQQ, TLT, BTC, IWM, SOXX, IGV, UUP, AAPL, AMZN, GOOGL, META, MSFT, NVDA, TSLA, PLTR

Fourteen tickers covering US equity indices, duration, crypto, and the mega-cap names that drive most of the market's return. This is the broadest suite and the most useful for investors who want a systematic overlay across their core portfolio.

This suite is growing. Yimin is expanding ticker coverage, and subscribers get access to new tickers as they go live. Importantly, Yimin will split the service into two future services, each covering Macro and Equities names (with more names added to both sets). However, if you signed up today, you will retain access to both services when the upgrade happens.

Learn More — YXI Macro & Mag-7

How Subscribers Use The Signals

There are two approaches we see, and both work.

Systematic. The model signals are followed as a system. When the model is Risk On, the position is held. When it flips to Risk Off, the position is closed. No discretion, no second-guessing. This approach removes emotion entirely and has delivered the strongest risk-adjusted returns in both backtesting and live trading.

Analytical input. The signals serve as one input alongside your own research. If you're bullish on gold but the model is Risk Off, that's useful information. The daily signal commentary, which explains the top macro and asset-specific drivers, is particularly valuable here. Even subscribers who don't follow the signals mechanically report that the commentary has materially improved how they think about what's driving markets.

Get Started

The links above will get you started.

If you're considering multiple services, don't sign up at the rack rates. Contact us at minerva@cestriancapital.com and we'll put together a discounted package based on which services you want to take.

Yimin Xu is the founder of YX Insights and the developer of the YXI ML signal suites hosted on Cestrian Capital Research. He is a former G10 FX & Interest Rate Derivatives market maker at NatWest Markets, an ICAEW Chartered Accountant, and the Lead FOMC Analyst at Seeking Alpha.