Methodology

The machine does not have opinions. It has a process.

A multi-layer systematic pipeline processes global market data after every close. Every signal source is independently validated. Every prediction is out-of-sample. The record is open.

0.1004
Mean IC
6.13
T-Statistic
147
OOS Months
74.1%
Positive Months
Architecture

Four layers. Each validated before the next runs.

Raw market data enters the pipeline every evening and passes through four sequential stages. No layer can pass forward data that fails its validation check. The ensemble never receives a signal source it has not independently confirmed is working.

Layer 0
Feature Generation
Macro regime classification, credit conditions, yield curve structure, volatility regime, cross-asset correlation, momentum persistence — the market’s environmental state, computed fresh each night.
Layer 1
Specialist Signal Sources
Ten proprietary models, each targeting an orthogonal source of alpha. Each source carries an independent IC history. Each is validated separately before contributing to the ensemble.
Layer 2
Meta-Learner Ensemble
A proprietary ensemble model combines signals from active sources, weighted dynamically by rolling predictive performance in the current regime. This is the output published as meta-alpha.
Layer 3
Allocation Output
Converts ensemble alpha to target positioning. Applies volatility scaling, exposure discipline, and onboarding discounts for instruments without sufficient OOS history.
Internal model architecture, signal construction, ensemble weighting, and sizing parameters are proprietary and not disclosed.
Validation

Walk-forward. Out-of-sample. 12 years of genuine predictions.

Every prediction in the performance record was made before the outcome was known. The methodology is strict walk-forward: the model trains on a fixed historical window, makes predictions on data it has never seen, then steps forward. The IC record is the continuous output of that process.

Information Coefficient
The statistical measure of predictive accuracy
0.1004 Mean IC · 147 OOS windows
IC measures rank correlation between predicted and realized forward returns. Zero is random. A sustained IC of 0.10 across 147 out-of-sample windows means the system consistently identifies which assets will outperform which others — not perfectly, but with a statistically significant edge.
Statistical Confidence
The probability this edge is luck
6.13 T-statistic · institutional threshold: 2.0
A t-statistic of 6.13 means the probability this IC is the product of random chance is less than one in a billion. Institutional-grade factor research typically requires a t-stat above 2.0 to be considered publishable. We are three times that threshold.
Lower Confidence Bound
The floor on the edge, not just the average
0.0729 IC lower 95% confidence bound
The 95% confidence interval lower bound on the IC is positive: 0.0729. This means we are statistically confident the true predictive edge is non-zero — even accounting for variance across the 147 prediction windows in the OOS record.
The 2022 Test
When every traditional portfolio broke
+42.1% 2022 return · SPY: −27.2%
In 2022, equities and bonds collapsed simultaneously. The regime framework had been reading late-cycle stress conditions all year and positioned accordingly. This was the system’s best year on record — produced by the same pipeline, the same process, the same discipline.
Signal Discipline

Sources go silent when they stop working.

Each specialist model in the ensemble carries an automatic exclusion mechanism. The system monitors rolling predictive accuracy for every model in every macro regime. When a model’s IC falls below threshold — indicating the edge has degraded in the current environment — that model is excluded from the ensemble automatically. No discretionary override. No second chances until performance recovers.

This matters because the alternative is what most systematic funds do: leave every model running at all times, let the underperforming ones dilute the strong ones, and report the blended result as if it were coherent. SQ Research does not do that.

The status of every specialist model is published daily on the Signal Feed. In the current late-cycle regime, 1 of 10 models meets the activation threshold. The ensemble is running on the one model with a verified positive IC in this environment. That is the correct answer — not the comfortable one.

Specialist Models · Today Late Cycle Regime
Alpha Model
IC lower CI = +0.053 > 0
Active
Volatility Model
IC lower CI = −0.108 ≤ 0
Excluded
Precious Metals Model
IC lower CI = −0.147 ≤ 0
Excluded
Correlation Model
IC lower CI = −0.115 ≤ 0
Excluded
Macro Model
IC lower CI = −0.138 ≤ 0
Excluded
Sentiment Model
IC lower CI = −0.017 ≤ 0
Excluded
+ 4 further models
All excluded in late cycle regime
Excluded

Live source status updates nightly · Full feed →

Universe

55 liquid, exchange-traded instruments across every major asset class.

All signals are published in terms of instruments that can be traded directly — no translation required. The universe is designed for breadth of regime coverage, not concentration in any single market.

US Equity
SPY · QQQ · IWM · DIA
XLK · XLF · XLE · XLV
XLI · XLB · XLU · XLC
XLRE · XLY · XLP
XHB · XOP · KRE
Intl Developed
EWJ · EWG · EWL
EWP · EWQ · EWU
Intl Emerging
EEM · EWH · EWT
EWY · EWZ · FXI
INDA
Precious Metals
GLD · SLV · GDX
GDXJ · SIL · SILJ
Energy
OIH · XOP
Natural Resources
COPX · LIT · URA
URNM · URNJ
Agriculture
WEAT · CORN · SOYB
DBA · MOO · TAGS
Fixed Income
TLT
Digital Assets
BITO
Coverage Summary
54 total instruments
All exchange-traded
All liquid (>$50M ADV)
Daily signal output
Directly implementable
What We Publish

Signal intelligence. Not advice.

SQ Research is a publisher, not an adviser. The distinction matters legally and practically. We publish what the system produces. You decide what to do with it.

What we publish
+Daily macro regime state and regime tags
+Directional signal and confidence score per instrument
+Multi-timeframe momentum analysis (3-day through 36-month)
+Signal source status and kill-switch transparency
+Allocation weights and vol-adjusted sizing (Professional)
+Factor attribution and regime performance breakdown
+Rolling IC report — monthly predictive accuracy update
+Full OOS walk-forward performance tearsheet
What we do not publish
Individual investment advice
Price targets
Discretionary commentary or opinion
Portfolio construction advice for specific accounts
Tax or legal guidance
Anything not produced by the systematic pipeline
The system ran last night. See what it produced.
Daily macro regime, full SPY brief, and signal preview — free, no card required. Or see the full 12-year performance record on the tearsheet.
See Today’s Signal Feed Read a Sample Daily Brief View Performance Tearsheet