Focused Hurst MR Strategy Eval

Review Decision

Official strategy-eval P0: FAIL / do not promote. The generic shift-lookahead heuristic fails and the post-2026-05-12 window decays materially. The added truncate replay below passes, so this looks more like a shift-test false positive for mean reversion than proven leakage.

5m net return13.77%

Fixed-capital net P&L$11,699.26

5m net Sharpe5.343

Max DD-2.02%

Avg exposure12.00%

Trades300

P&L Curve

Shift Test vs Causal Replay

The shift heuristic tests whether using future positions improves Sharpe. This is useful for many momentum/ranking systems, but it can fail on mean-reversion systems because tomorrow's entry often follows today's adverse move, so the lead-position cheat can make performance worse instead of better.

Official shift passfalse

Lead-1 gain-1.622

Truncate replaytrue

Compared bars43515

Position diffs0

PASS - positions before the cutoff are identical when future bars are physically removed.

ticker	compared_bars	position_diff_count	max_position_diff
FOXA	4835	0	0.0
WMB	4835	0	0.0
HON	4835	0	0.0
QCOM	4835	0	0.0
INTC	4835	0	0.0
FTV	4835	0	0.0
KMI	4835	0	0.0
VLTO	4835	0	0.0
MGM	4835	0	0.0

Why Return And Drawdown Look Low

The strategy is not usually fully invested. Average gross exposure is only 12.00%, max observed exposure is 66.67%, and at least one symbol is active on 60.34% of bars. So dollar P&L and drawdown are low because most capital is idle most of the time.

Return on total NAV13.77%

Return / avg exposure114.71%

Avg gross exposure12.00%

Max gross exposure66.67%

Scaling notional linearly would scale dollar P&L and dollar drawdown roughly linearly while percent return on the stated capital base rises only if we intentionally use more of the idle capital. Sharpe is scale-invariant only before market impact, borrow/financing constraints, liquidity caps, and fill slippage.

Assumptions

Data: Alpaca SIP 5-minute bars, regular US session only, 5850 bars per symbol.
Execution model: signal evaluated after bar close; position applies on the next bar.
Costs: 2 bps per position change. No bid/ask queue, latency, or partial-fill simulator.
Capital: fixed $10,000 per symbol, $90,000 total. Official eval also shows compounded NAV returns.
OOS split: 2026-05-12, representing post-original-research/paper-observation period, not a hidden holdout.

OOS Split

IS net Sharpe6.218

OOS net Sharpe0.971

IS net return13.34%

OOS net return0.38%

OOS max DD-2.02%

Official OOS decay-0.594

Current Paper Execution Rules

Baseline P&L$11,699.26

Controlled P&L$10,605.72

Controlled Sharpe5.053

Order reduction50

Blocked entries133

case	ticker	strategy	orders	entries	exits	blocked_entries	exposure_pct	pnl_usd	return_pct	sharpe	max_drawdown_usd
baseline_portfolio	PORTFOLIO	focused_basket	600	300	300	0	NaN	11699.26	13.00	5.343	-1829.02
cooldown30_max4_portfolio	PORTFOLIO	focused_basket	550	275	275	133	NaN	10605.72	11.78	5.053	-1469.15

Strategy-Eval Gate

Official operational reliability: shift-lookahead heuristic failed. Truncate replay passes, but the official P0 gate remains conservative.
DSR/PBO: needs input; no candidate trial panel or walk-forward panel supplied.
Factor exposure: computed against SPY_market only; beta 0.148, R2 0.072. Full factor model still missing.
Portfolio interaction: needs existing strategy returns; no max-correlation conclusion yet.
Paper/fill readiness: needs paper return stream and fill/latency/spread model.

Open official eval dashboard · Bias report · Internal report · Summary MD · Causal replay JSON

Per-Symbol 5m Backtest

ticker	sharpe	total_pnl_usd	total_return_pct	max_drawdown_usd	trades	win_rate	exposure_pct
FOXA	4.380	1389.58	13.90	-186.93	33	0.758	11.38
HON	3.559	1100.58	11.01	-246.03	29	0.793	10.09
FTV	3.558	1325.08	13.25	-433.62	32	0.750	10.12
WMB	3.246	788.12	7.88	-292.03	26	0.846	9.52
MGM	2.529	1217.52	12.18	-519.48	34	0.706	10.09
INTC	2.378	4054.73	40.55	-1011.09	40	0.625	13.76
VLTO	1.954	1009.68	10.10	-499.56	38	0.737	16.26
KMI	1.401	380.23	3.80	-536.27	30	0.667	12.53
QCOM	0.933	433.74	4.34	-651.95	38	0.579	14.29

CSV

Common Daily Horizons

Daily multi-horizon scan from the existing research artifact; this is a broader daily proxy, not the 5m SIP strategy itself.

horizon	tickers	positive_tickers	avg_trades	total_pnl_usd	return_pct	sharpe	max_drawdown_usd
1m	80	29	0.71	11784.21	1.47	4.511	-4074.71
3m	80	39	1.50	22371.22	2.80	2.780	-14283.83
6m	80	50	2.62	30705.48	3.84	2.158	-13995.05
1y	80	58	4.17	65193.33	8.15	2.591	-11579.68
3y	80	64	10.18	157914.51	19.74	1.462	-29418.13