H1 — Time-Series Momentum (Trend) on US equities

QuantConnect LEAN backtests on local daily data · all metrics from LEAN result JSON · generated 2026-06-30 17:13

Thesis. Time-series (absolute) momentum: an asset whose own trailing return is positive (or price > long SMA) tends to keep trending; go long on positive trend, flat/short on negative. A trailing-realized-vol target and a Faber MA overlay are layered to cut downside. We test whether this is a real, cost-surviving edge that reduces drawdown vs buy & hold, in-sample then out-of-sample. Daily bars, end-of-day decision on the prior close (no look-ahead), monthly rebalance, costs modelled as bps-of-notional.

H1 verdict: PASS

Kill criterion: OOS net-of-cost Sharpe ≥ 0.40 on SPY. Best SPY OOS variant: trend_ma → Sharpe = 0.46 (meets the bar).

McLean–Pontiff decay caveat: applying the ~×0.74 OOS post-publication discount to that figure gives an expected live Sharpe ≈ 0.34. LEAN's Sharpe is computed net of a (historically high, ~3–5% in 1998–2007) risk-free rate, which structurally depresses these numbers vs a naive mean/σ Sharpe.

Drawdown-reduction check (full period, the core defensive claim):

SPY: B&H DD=55% → Trend DD=32%, VT DD=30% (trend cuts DD vs B&H)
AAPL: B&H DD=82% → Trend DD=72%, VT DD=29% (trend cuts DD vs B&H)
BAC: B&H DD=94% → Trend DD=48%, VT DD=38% (trend cuts DD vs B&H)
IBM: B&H DD=59% → Trend DD=80%, VT DD=65% (trend did NOT cut DD)

Crypto leg deferred: the local BTCUSD daily file has a 6.4-year hole (2018-08 → 2025-01), so the crypto TSMOM leg of H1 is NOT tested here and is deferred until continuous crypto data is available. SPY itself is only continuous 1998→2021-03 in the local data (1373-day hole after), so SPY runs stop at 2021-03.

Equity curves (full period, net of 5 bps)

Variant comparison — CAGR / Sharpe / MaxDD / Turnover (IS · OOS · Full)

Sharpe cells ≥ 0.40 highlighted green. Turnover = LEAN mean daily Portfolio Turnover.

SPY

Variant	IS (1998–2013)				OOS (2014–2021-03)				FULL (1998–2021-03)
	CAGR	Sharpe	MaxDD	Turn	CAGR	Sharpe	MaxDD	Turn	CAGR	Sharpe	MaxDD	Turn
Buy & Hold	4.1%	0.11	54.7%	0.02%	13.4%	0.61	33.7%	0.04%	6.8%	0.25	54.7%	0.01%
Trend (signal only)	6.7%	0.26	20.4%	0.12%	7.7%	0.36	33.6%	0.41%	7.0%	0.30	31.6%	0.18%
Trend + Vol-Target	5.8%	0.23	18.5%	0.17%	7.0%	0.36	27.1%	0.43%	5.7%	0.24	29.9%	0.22%
Trend + MA-overlay	5.9%	0.23	17.3%	0.33%	8.3%	0.46	22.1%	0.49%	6.1%	0.26	31.2%	0.34%
Trend + VT + MA	5.3%	0.20	17.3%	0.33%	7.4%	0.42	20.0%	0.50%	5.4%	0.22	30.0%	0.35%

AAPL

Variant	IS (1998–2012)				OOS (2013–2026)				FULL (1998–2026)
	CAGR	Sharpe	MaxDD	Turn	CAGR	Sharpe	MaxDD	Turn	CAGR	Sharpe	MaxDD	Turn
Buy & Hold	30.0%	0.70	82.1%	0.02%	23.2%	0.70	38.5%	0.02%	26.9%	0.68	82.1%	0.01%
Trend (signal only)	28.0%	0.71	71.7%	0.13%	14.1%	0.46	43.0%	0.34%	20.4%	0.58	71.7%	0.28%
Trend + Vol-Target	14.1%	0.68	23.4%	0.11%	9.3%	0.38	24.9%	0.38%	11.4%	0.51	28.7%	0.27%
Trend + MA-overlay	22.1%	0.60	79.8%	0.33%	15.0%	0.53	36.2%	0.42%	18.4%	0.55	79.8%	0.39%
Trend + VT + MA	12.4%	0.60	26.1%	0.16%	10.1%	0.46	21.3%	0.41%	11.2%	0.52	26.1%	0.29%

BAC

Variant	IS (1998–2012)				OOS (2013–2026)				FULL (1998–2026)
	CAGR	Sharpe	MaxDD	Turn	CAGR	Sharpe	MaxDD	Turn	CAGR	Sharpe	MaxDD	Turn
Buy & Hold	-2.8%	0.10	93.6%	0.02%	13.4%	0.41	49.0%	0.02%	4.7%	0.20	93.6%	0.01%
Trend (signal only)	3.4%	0.07	42.2%	0.27%	2.9%	0.07	47.7%	0.47%	3.1%	0.06	47.7%	0.36%
Trend + Vol-Target	2.5%	-0.05	20.1%	0.23%	2.0%	-0.03	38.2%	0.39%	2.2%	-0.04	38.2%	0.30%
Trend + MA-overlay	3.4%	0.05	26.8%	0.35%	1.4%	-0.01	42.7%	0.67%	2.4%	0.01	42.7%	0.49%
Trend + VT + MA	—	—	—	—	—	—	—	—	—	—	—	—

IBM

Variant	IS (1998–2012)				OOS (2013–2026)				FULL (1998–2026)
	CAGR	Sharpe	MaxDD	Turn	CAGR	Sharpe	MaxDD	Turn	CAGR	Sharpe	MaxDD	Turn
Buy & Hold	6.0%	0.19	59.2%	0.02%	4.8%	0.16	43.5%	0.02%	5.5%	0.18	59.2%	0.01%
Trend (signal only)	0.3%	-0.04	67.4%	0.49%	-3.1%	-0.24	62.4%	0.73%	-1.8%	-0.15	79.8%	0.50%
Trend + Vol-Target	1.0%	-0.12	45.3%	0.40%	-2.4%	-0.34	58.1%	0.62%	-0.9%	-0.24	64.8%	0.43%
Trend + MA-overlay	-2.0%	-0.18	71.3%	0.62%	-2.2%	-0.22	59.0%	0.61%	-2.3%	-0.21	80.3%	0.58%
Trend + VT + MA	—	—	—	—	—	—	—	—	—	—	—	—

Transaction-cost sensitivity (headline Trend+Vol-Target, full period)

Config	0 bps	5 bps	10 bps	20 bps
SPY Trend+VT (full) — CAGR	5.8%	5.7%	5.7%	5.6%
SPY Trend+VT (full) — Sharpe	0.24	0.24	0.23	0.23
SPY Trend+VT (full) — MaxDD	29.8%	29.9%	29.9%	30.0%
SPY Trend+VT (full) — Fees($)	0	1986	3942	7803
AAPL Trend+VT (full) — CAGR	11.5%	11.4%	11.4%	11.3%
AAPL Trend+VT (full) — Sharpe	0.51	0.51	0.51	0.50
AAPL Trend+VT (full) — MaxDD	28.6%	28.7%	28.7%	28.8%
AAPL Trend+VT (full) — Fees($)	0	17161	34014	66811

Lookback parameter stability (Trend, full period)

Config (Trend, full, 5bps)	3M (63d)	6M (126d)	12M (252d)
SPY — CAGR	6.8%	5.4%	7.0%
SPY — Sharpe	0.33	0.20	0.30
SPY — MaxDD	19.6%	34.2%	31.6%
AAPL — CAGR	17.1%	24.4%	20.4%
AAPL — Sharpe	0.51	0.71	0.58
AAPL — MaxDD	80.6%	63.3%	71.7%

Notes & limitations

LEAN Sharpe is net of the risk-free rate; 1998–2007 T-bill yields (~3–5%) materially lower it vs a naive Sharpe.
Single-stock TSMOM (AAPL/BAC/IBM) is structurally noisier than index trend; reported as-is, not massaged.
SPY limited to 1998→2021-03 (local data hole afterward). BTCUSD crypto leg deferred (6.4y data hole).
Costs are a flat bps-of-notional proxy for commission+slippage; no market-impact/capacity modelling.
IS/OOS is a single split; only light lookback/cost sensitivity shown — full walk-forward is follow-up.

Source: /config/workspace/lean-strategy/backtests/H1TrendTsmom/ · algorithm H1TrendTsmomAlgorithm.cs · LEAN engine (net10.0).