feat(strategy): value+momentum composite — second-edge probe - #2
Open
Dukotah wants to merge 2 commits into
Open
Conversation
… probe
Add ValueMomentumStrategy: a combined per-asset VALUE + MOMENTUM composite-score
strategy, the 'next probe' flagged in DECISIONS Session 22 / ROADMAP Phase 6 for
the second uncorrelated edge.
apex/strategy/library/value_momentum.py
- Per-asset composite = value_weight * z(value) + momentum_weight * z(momentum),
cross-sectionally Z-scored so the two legs share a scale and neither dominates
by units; degrades gracefully (Z=0) with a single live asset / zero dispersion.
- VALUE leg = long-horizon reversal -(return value_period->skip_recent) (5y, skip
last 1y) — the cheap-vs-multi-year-ago laggard scores high.
- MOMENTUM leg = weighted blend of trailing returns over multiple lookbacks
(default 21/63/126/252 ~ 1/3/6/12 months); only scored once EVERY lookback has
history so scores stay comparable across assets.
- Holds top-K by composite, long/flat, position-aware via broker-reconciled
context (cold-start correct, no pyramiding), inverse-vol sized via signal
strength only (RiskManager remains sole sizer). Optional trend-trap filter,
OFF by default to measure the pure composite's correlation to trend first.
- Configurable universe: symbols arg + DEFAULT_VALMOM_UNIVERSE (the broad 10-ETF
'expanded' pool: SPY/EFA/EEM/TLT/IEF/LQD/GLD/SLV/DBC/VNQ) + default_universe()
helper — runs on the richer 10-13 ETF pool per Session 22 next probe (a).
- Emits SignalEvents with strength 0..1, a reason string and suggested_stop_loss;
no orders, no I/O, no wall-clock; stdlib-only deterministic math.
- Docstring states it MUST clear the Gauntlet (esp. Sharpe@2x-cost) AND the
correlation gate before any capital; NOT marked deployed.
tests/test_value_momentum.py
- Validation; hand-computed value/momentum leg metrics; composite Z-score math and
the 'both cheap AND trending wins' ranking; missing-leg skipping; single-asset
graceful degradation; end-to-end buy of the cheap-recovering asset (and veto of
the long-run winner); stop/reason/strength on the signal; trend-filter veto of a
falling composite leader; default-universe shape; determinism.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…entions - Metric tests had miscounted negative indices: a k-bar return uses close[-(k+1)], so the equal/weighted momentum blends are 0.2121.. / 0.1515.. and the skip_recent value window reads close[-5]->close[-2]. Implementation was correct. - composite ranking test now sets top_k=2 so the worst-on-both asset is excluded. - end-to-end test now asserts the true behavior: the cheap-AND-recovering asset is the composite's FIRST pick and the expensive long-run winner never leads it (the 50/50 blend does not categorically veto it, so 'never bought' was too strong).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Pursues the second uncorrelated edge (the real Phase 6 blocker) — the documented next probe from DECISIONS Session 22.
Adds
apex/strategy/library/value_momentum.py—ValueMomentumStrategy: per-asset composite of a value leg (long-horizon reversal z-score) and a momentum leg (blended TSMOM over 1/3/6/12mo), each cross-sectionally z-scored then combined (AQR "Value & Momentum Everywhere" score-level approach). Holds top-K, long/flat.tests/test_value_momentum.py— hand-derived leg math, composite ranking, degenerate-case safety, end-to-end signals, determinism.Notes
🤖 Generated with Claude Code