How it is tested
Every strategy goes through the same procedure. This page lists the procedure.
Research philosophy
Don’t trust the backtest. Check what happened after it. Research should be open enough to be challenged. Evidence should be open enough to be examined.
Open research, proprietary implementation
DM-Signal publishes its research and protects its implementation.
Open research. Observable evidence. Proprietary implementation.
Open to anyone: the free Basic-DualMomentum signal and every month its holding switched; the testing procedure applied to every strategy (this page); the period and conditions attached to every performance figure; and the commitment never to rewrite a signal record as it was published at the time.
For members: the advanced portfolios, their signals and weights, and additional implementation notes.
Not published: the detailed selection logic of the advanced portfolios. We do not accept requests to disclose the selection logic, and we do not answer individual questions about it beyond what is on this page.
Research should be open enough to be challenged. But transparency in research is not the same as full disclosure of a proprietary implementation. We protect the implementation so that the research can continue and its results can keep reaching members.
This is not a policy unique to DM-Signal. In “Academic Alpha: A (Re)Introduction to Style Premia Investing” (AQR, August 2026), Thomas Maloney and Tobias Moskowitz of AQR, a global investment firm, write that the ideas behind research-based investing are not secret sauce, while the measures and implementations might be. The same paper notes that a manager’s newest proprietary signals may be kept for its premium strategies.
the ideas themselves are not “secret sauce” – though some of the measures and implementations might be
a manager’s very latest proprietary signals may be reserved for its premium “full model” strategies
Source: AQR, “Academic Alpha: A (Re)Introduction to Style Premia Investing” (PDF)
Pharmaceutical companies and university research institutions also combine published research with the protection of intellectual property. Like AQR, DM-Signal follows a policy of separating published research from its proprietary implementation.
You do not need to become a member just to check the evidence.
Core backtest validation
- A1
Massive Grid Search
We search a wide parameter space and look for a plateau where results hold up, not a single best point.
- A2
Parameter-Space Robustness
We compute beta-adjusted alpha for every grid pattern and report the share of patterns with positive alpha, not just the champion.
- A3
In-Sample / Out-of-Sample
Selection uses only data up to a cutoff; evaluation uses the unseen period after it.
- A4
Multi-Window Walk-Forward
We repeat select-then-evaluate as time moves forward, across several in-sample window lengths, so one lucky window cannot carry the result.
- A5
Beta-Adjusted Alpha
We estimate market beta and evaluate what remains after market exposure is removed; strategies can and do fail this step.
- A6
Expanding Window
We also test with an expanding training window, so results do not depend on one rolling window length.
- A7
Regime Robustness
We break results down by market regime instead of relying on a single average.
- A8
Vintage OOS
We rerun the research as if we stood at an earlier date, using only information available then, and evaluate what followed.
- A9
Reject / Adopt
Strategies that pass early tests are still rejected when later tests fail; rejected research is kept on record.
- A10
Real Prospective Forward
After every backtest check, the researcher still knows the past; only signals published in advance and observed afterward remove that.
DM-Signal robustness tests
- B1
Alpha Space Check
Measures the CAGR positive-rate across the entire parameter space. Broadly positive results across the parameter space reduce the likelihood that performance comes from a single isolated parameter spike, but do not by themselves rule out overfitting.
- B2
Decay Analysis
Directly measures whether a champion selected in the past has eroded over time, across 6/12/24-month windows.
- B3
Vintage Analysis
Verifies whether out-of-sample performance reproduces across champions selected at different points in time, using selection timelines and IS/OOS comparison.
- B4
Parameter Neighborhood Analysis
Confirms whether a champion is an isolated spike or a broad peak, by comparing the performance drop-off across neighboring parameters.
- B5
Stress Test
Examines whether the apparent edge holds up against the benchmark during extreme periods such as sharp drawdowns or rate-hike shocks.
- B6
Fold Percentile Verification
Confirms whether a fixed champion maintains better-than-random performance within each walk-forward fold's population, via fold percentile.
- B7
Metric Top-Stability
Measures how well the top set is preserved as the calculation window is shortened, using overlap rate / Jaccard similarity.
- B8
Regime-Conditional Verification
Checks whether alpha remains positive across Bull/Bear/Sideways regimes, exposing apparent strength that depends on a single market condition.
- B9
Rule Freeze
Locks selection rules and parameters at a fixed point in time, with no undocumented changes afterward. This freeze point becomes the boundary against the still-unverified future.
- B10
Forward Observation
Observes and recalculates only real-time results after the Rule Freeze, kept distinct from the backtest. Forward metrics are computed using only rows from the forward period.
Historical data revisions
Historical price series can change after the fact. Adjusted prices are recalculated back through the past each time a new dividend, split, or distribution occurs, and data vendors sometimes correct past data. Recalculated research series may therefore differ from the original historical signal record. These changes do not justify rewriting the old publication record.