Understanding the different signal types

Strategy labels are provenance, not predictions

A recommendation's strategy label tells you which documented research lens produced it. The label does not mean the security will rise, that a target will be reached, or that the idea is suitable for a particular person.

The public Markov methodology currently describes six factor-oriented lenses.

Persistent quality

Ranks businesses using the persistence of measures such as profitability, leverage and operating stability. Historical company metrics can change and do not eliminate valuation or market risk.

Residual momentum

Compares the stock-specific portion of recent relative performance after estimating broad-market and sector effects. Model estimates are imperfect and historical momentum can reverse.

Sector-neutral momentum

Compares momentum within sectors so that one strong sector does not dominate the selection solely because the whole group rose together.

Low-volatility quality

Combines business-quality measures with a realised-volatility screen. Low historical volatility does not prevent future gaps, drawdowns or losses.

Volatility-targeted basket

Adjusts a basket's model exposure as estimated volatility changes. An estimate can react late, and a target volatility is not a maximum loss.

Quality-momentum / QMJ

Combines quality, momentum and an inverse junk rank in Markov's internal composite. It should not be treated as identical to the academic Quality Minus Junk portfolio.

Read the complete card

The strategy label is only one field. Check the dated status, direction, illustrative entry area, structural stop and indicative target together. If a price has moved beyond the published structure, do not infer from the label that an old entry remains current.

Anything described as experimental or research-only is not a live recommendation. Use the current app state as the source for a dated item and the How it works page for the evergreen methodology.

For academic background, read Factor strategies, explained for non-quants.