Psi

01 / Track record

What the record says, and how it was kept.

US equity signals · 1-XH family · paper forward record

Ledger as of 2026-09-17

2026-05-29

Record starts

1,148

Closed trades

55%

Win rate, closed

+2.0%

Avg per trade

+1.5%

Median per trade

293

Open, marked daily

Question. Does a published 1-XH buy signal earn a positive return over the next 20 sessions?

Publication. Each signal is stamped with its session, published entry price and model version (xiaohan_gogo_v2, xiaohan_reborn_v3) on the day it fires.

Output at the time. A buy signal with a horizon of 20 days; the top names also carry an advisory three-agent review with a confidence score that never gates the record.

Verification. Entered the bar after the signal, marked to market daily, exits at target +15% / stop −25% / horizon; gross of costs; split-affected and unpriceable rows are flagged and excluded.

Record type. Paper forward record. It is separate from the historical backtest and cannot be fitted after the fact. Exit rules v2 date from 2026-08-09 and are applied retroactively.

Exits so far. 293 target · 71 stop · 784 horizon. A plain 20-day-hold track for the same signals is kept alongside.

Full track record → Every closed trade → Historical backtest →

03 / Directions

Two directions. One toolkit.

Direction 01 · Markets

Equity signals with a verified record.

Cross-sectional onset signals, regime detection and forward verification on US and A-share equities. This is the practice that produces the toolkit reused elsewhere.

US equity signals20-day horizon · 254-ticker universe · paper forward recordOpen →
Strategy track recordBacktest and dated forward ledger, side by sideOpen →
A-share daily picksPoint-in-time pool · Chinese deep reportsOpen →
Multi-horizon and derivatives overlaysNot yet published; no public recordIn research

Direction 02 · Frontier industries

Sector indicators for industries that lack them.

The same discipline applied to sectors where most analysis is editorial: explainable factors, disclosed coverage, and a name gated rather than guessed when the evidence is thin.

Quantum-computing sector signal13 factors · 5 families · listed universe, names redacted publiclyOpen →
Quantum-safe sector forecastBuilt with RBTX TechnologyOpen →
Advanced materials and deep-tech signalsFactor design in progressIn research
Computational-biology indicatorsScopingIn research

04 / Method

How the outputs are made and checked.

Signals are rules, not opinions.

The 1-XH family is a fixed set of onset rules run over a point-in-time universe (254 US tickers; an A-share pool built the same way). A signal fires or it does not; the entry is the bar after it fires, never the signal bar itself.

Outputs carry their own provenance.

Every published item states its data date, model version and, where a language model contributed, the agent chain that produced it. LLM review is advisory: it can annotate a signal, it cannot create or veto one.

Verification is forward, dated, and separate from the backtest.

Backtests are shown gross of costs with their accounting stated. The forward record is priced at real next-session prices, marked to market daily and closed by fixed exit rules. Rows that cannot be priced honestly (splits, data gaps) are flagged and excluded, and the count of exclusions is published.

Read the full accounting →

Coverage is disclosed; thin evidence is gated.

Sector indicators score each name only on the factors that structurally apply, show the coverage fraction next to the score, and withhold a score rather than guess when coverage is insufficient. Confidence is a number, not an adjective.

Why “Psi”. ψ describes every possible outcome weighted by amplitude. The name is a reminder that a forecast is a distribution, not a point. The mathematics on these pages is ordinary: rule-based signals, cross-sectional factors and forward verification, each documented where it is used.