Methodology

Architecture

Universe: point-in-time constituents of S&P 500 ∪ Nasdaq-100 ∪ Dow-30 (delistings included, no survivorship bias), ~500+ US large caps.

Selection: 12-1 momentum factor, monthly ranking, Top-10. One of the most academically validated cross-sectional anomalies (Jegadeesh & Titman 1993 onward).

Sizing: Soft Actor-Critic RL ensemble (3 seeds) produces per-name weights plus an explicit cash dimension; 25% single-name cap.

Tail hedge: ~30-day 5%-OTM SPY puts sized to NAV; premium booked to NAV.

Execution: signal generated at month-end close, executed next trading day via full-session VWAP algo orders (postponed on holidays). ~100% monthly turnover; all holdings are US large caps — conservative capacity estimate >$500M without meaningful impact. Dividends: backtest uses adjusted prices (reinvested); live dividends accrue to cash.

Backtest basis

· All reported performance is post-2022 walk-forward, not in-sample fit.

· Fill assumption: next-day execution after signal; commissions, slippage and option premium fully charged in the NAV.

· Why the sample starts in 2022: point-in-time constituent data was rigorously cleaned from 2022 onward; earlier coverage does not meet our publication bar. The ~4.8y sample is the strategy's main limitation, though it spans a full bear market (2022) and a momentum drawdown regime (2025).

· Every ML/DL enhancement (supervised ranking, MASTER transformer, Kronos foundation model) was rejected on out-of-sample RankIC vs the momentum baseline — we do not add complexity for its own sake.

· Current stage: IBKR paper-account forward validation; live-vs-backtest divergence is tracked on the overview page.

Hedge disclosure

The monthly rolling OTM SPY put hedges gap-crash tail risk, not trending drawdowns — the 2025-01 −31.7% max drawdown was grind-type, where puts contribute little. Premium is ~0.3–0.5% of NAV per month (~3–5% annualized), charged to NAV as an explicit insurance cost. Current sector concentration (semis + hardware) is a natural outcome of the momentum signal, not a constraint target.

Governance & Data Integrity

Our commitments: automated daily sync, fill-level disclosure, backtest vs live separation, NAV commitment chain with blockchain timestamps (see ZKP proposal). Every figure is independently verifiable.

FAQ

Why monthly rebalancing?
Monthly frequency balances momentum decay against transaction costs; intraday noise never enters decisions.

Why only 10 names?
Momentum returns concentrate in the top ranks; over-diversification dilutes alpha. Top-10 with a 25% cap is the academic-empirical compromise.

Why hold cash?
Cash is the model's de-risking tool — in weak-signal regimes it prefers underweight over forced full investment.

Manager / Contact

AlphaGo Capital — systematic quantitative research. For institutional due diligence, data packs and monthly letters: research@alphago.example (placeholder). Legal entity and registration details will be disclosed upon fund formation; figures shown are a strategy research record, not a registered fund product.

Data integrity — ZKP Phase 1 live: the full published NAV series is committed in a SHA-256 Merkle tree, signed with Ed25519. Any institution can independently recompute the root and verify a leaf offline — proving the track record was not rewritten. View the signed commitment (JSON) · ZKP proposal (Word)

Signing pubkey (Ed25519): 2a5b2ba3befc2a079fdbc60bebd1d3c04c98a29a5e544367183a04119e50f648
To verify: authenticated GET /api/zkp/proof?date=YYYY-MM-DD, then recompute root + check signature offline with verify_zkp.py. The root alone reveals no NAV data.

Strategy Version — AGM-1.4 mom12-1 · top10 · sac-ens3 · cash · spy-put · 2025-10

Strategy Version	When	Changelog
v1.0	2024 Q3	3-seed SAC ensemble + 12-1 momentum Top-10 + monthly rebalance finalized
v1.1	2024 Q4	25% single-name cap + explicit cash allocation dimension
v1.2	2025 Q2	Monthly rolling 5% OTM SPY put tail hedge introduced
v1.3	2025 Q3	PIT index universe rebuilt; survivorship-bias audit passed
v1.4	2025 Q4	Cost model recalibrated ($0.005/sh + VWAP slippage); IBKR paper forward validation launched