Flagship XS-01 · Adaptive Multi-Horizon Rank Ensemble
One rank ensemble, every dataset we support — crypto, US, China, Hong Kong, UK, FX and commodities — market-neutral, rebalanced on a 10-day horizon.
Alpha performance report
Walk-forward backtest
Nasdaq-100 constituents, point-in-time membership, daily bars
A cross-sectional rank ensemble that blends momentum, reversal, volume and volatility factor families into a single daily score for every name in the US · Nasdaq-100 universe, rebalanced on a 5-day horizon. Factor weights adapt per walk-forward window via a regularised rank regression — no look-ahead: every signal uses only data available at decision time. In-sample 2016-01-04 → 2020-12-31, validation 2021-01-04 → 2022-06-30, out-of-sample 2022-07-01 → 2026-06-30. Headline figures are net of 5 bps per side. Demonstration run — figures are illustrative.
Signal quality · shared by both books
Signal IC
0.0466
IC info ratio
1.026
% positive
87.5%
Horizon
10 days
WF windows
24
OOS 2022-07-01 → 2026-06-30 · $1.00M book · market-neutral L/S vs long-only
Statistical robustness
Deflated verdict
Adjusts the headline Sharpe for multiple-testing and non-normality, so a high in-sample number is not mistaken for a real edge.
Sharpe survives multiple-testing deflation — the edge is unlikely to be luck.
Probabilistic Sharpe
99.6%
Deflated Sharpe
98.2%
Sharpe t-stat
4.22
Trials tested
128
Expected max Sharpe
1.04
Book Sharpe (ann.)
1.92
Signal quality score
0.91
Expectancy
0.064
SQS components
Institutional checklist
Quality gates
Each gate is an independent pass/warn/fail threshold an institutional reviewer would apply before allocating.
≥ 1.00
≥ 95%
≥ 2.0
≥ −25%
≥ 10 bps
≥ 0.50
Signal analysis
Information coefficient & signal quality
How well the cross-sectional score ranks forward returns — IC by horizon, walk-forward stability and the quintile sort.
Best-horizon IC
0.0466
Best IC-IR
1.026
% positive windows
87.5
Quintile monotonicity
0.98
Q5 − Q1 spread
0.668
IC by horizon
Mean IC (bars) and information ratio (line) at each forward horizon.
Quintile sort
Mean forward return of each signal quintile — monotone Q1→Q5 is the goal.
Walk-forward IC stability
IC at each horizon across out-of-sample windows — flat/upward is stable, downward decays.
IC by horizon — detail
| Horizon | Mean IC | Std | IC-IR | Hit % | Windows | Stability |
|---|---|---|---|---|---|---|
| 1d | 0.0269 | 0.017 | 1.611 | 91.7 | 24 | 0.74 |
| 5d | 0.0445 | 0.035 | 1.258 | 87.5 | 24 | 0.67 |
| 10d | 0.0466 | 0.045 | 1.026 | 87.5 | 24 | 0.59 |
| 20d | 0.0393 | 0.026 | 1.481 | 91.7 | 24 | 0.72 |
| 40d | 0.0236 | 0.020 | 1.165 | 87.5 | 24 | 0.64 |
Portfolio construction
Tradeable books
The signal turned into portfolios: a market-neutral long/short and a long-only variant, each marked to an out-of-sample equity curve.
Out-of-sample equity
Growth of the initial book over the walk-forward test period.
Long / Short
L/STotal return
+78.0%
Ann. return
+14.9%
Sharpe
1.92
Sortino
2.99
Max DD
-7.9%
Win rate
55.5%
Profit factor
1.35
Turnover
12.4×
Drawdown
Underwater curve from the running peak.
Daily return distribution
Histogram with mean, ±1σ band and VaR₉₅.
Rolling Sharpe
Annualised Sharpe over a trailing window.
Rolling volatility
Annualised volatility over a trailing window.
Monthly returns
Calendar of monthly P&L with annual totals.
| J | F | M | A | M | J | J | A | S | O | N | D | Yr | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2022 | 2 | 7 | -1 | 1 | -0 | -0 | +8 | ||||||
| 2023 | 3 | 1 | 3 | 4 | 3 | -3 | 2 | 1 | 4 | 2 | -0 | 3 | +23 |
| 2024 | -0 | -2 | 2 | -1 | 1 | 2 | 2 | 0 | -4 | -2 | -0 | 0 | -1 |
| 2025 | 2 | 1 | -1 | 5 | 2 | 5 | 4 | -2 | 2 | 1 | 0 | 7 | +30 |
| 2026 | -2 | 1 | 2 | 0 | 1 | 3 | +5 |
Long Only
L/OTotal return
+141.7%
Ann. return
+23.8%
Sharpe
1.34
Sortino
2.02
Max DD
-17.0%
Win rate
53.9%
Profit factor
1.23
Turnover
5.4×
Drawdown
Underwater curve from the running peak.
Daily return distribution
Histogram with mean, ±1σ band and VaR₉₅.
Rolling Sharpe
Annualised Sharpe over a trailing window.
Rolling volatility
Annualised volatility over a trailing window.
Monthly returns
Calendar of monthly P&L with annual totals.
| J | F | M | A | M | J | J | A | S | O | N | D | Yr | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2022 | -0 | 10 | -2 | 0 | -5 | -3 | -2 | ||||||
| 2023 | 14 | 10 | 1 | 2 | 0 | -1 | -4 | -2 | 4 | 3 | -2 | -5 | +21 |
| 2024 | 1 | -1 | 7 | -5 | -0 | -2 | 5 | 5 | 3 | 6 | 4 | -3 | +19 |
| 2025 | 4 | 14 | -2 | 2 | -1 | -2 | 10 | -2 | 6 | 5 | 2 | 1 | +40 |
| 2026 | 5 | 10 | 8 | -1 | -1 | -1 | +22 |
Head to head
Long/short vs long-only
The same signal expressed two ways — compared on growth, drawdown, the risk-adjusted profile and year by year.
Cumulative return
Growth of each book over the test window.
Drawdown
Underwater curves overlaid.
Risk-adjusted profile
Sharpe, Sortino, Calmar, win rate, profit factor and low-vol, normalised.
Annual returns
Year-by-year return for each book.
Tradeability
Transaction-cost sensitivity
How performance survives realistic trading costs — and the breakeven cost at which the edge disappears.
L/S breakeven
58.0 bps
L/O breakeven
79.0 bps
L/S turnover
12.4×
L/O turnover
5.4×
Net Sharpe vs cost
Annualised Sharpe net of trading cost at each bps tier; dashed line marks breakeven.
Net return vs cost
Total return net of trading cost at each bps tier.
Risk & costs
Risk profile
Tail risk, distribution shape and trading cost intensity for the headline book.
Ann. volatility
7.4%
Daily volatility
0.47%
Skewness
-0.04
Kurtosis
2.9
VaR 95%
-0.70%
CVaR 95%
-0.90%
VaR 99%
-1.00%
Max consec. losses
6
Avg drawdown
-1.64%
DD recovery
91d
Total trades
19,039
Annual turnover
12.4×
Reproducibility
Backtest configuration
Exactly how the run was specified — universe, frequency, walk-forward scheme and capital.
Universe & method
Walk-forward & capital
Execution rules
Five to ten trading days per position, rebalanced on the 10-day forecast horizon.
Final value $1.78M from $1.00M over 4.14 years · generated 2026-07-14.
How it works
Methodology & next steps
The mechanism behind the signal and the improvements the research pipeline proposes next.
Algorithm
- Compute per-name factor z-scores (12-1 momentum, 5d reversal, volume surprise, range-based volatility) each day.
- Rank-transform each factor cross-sectionally to unit-uniform scores.
- Fit a ridge rank-regression of forward 10-day returns on the factor ranks over the trailing training window; hold coefficients fixed through the test window.
- Score every name daily; enter the top/bottom quintiles with turnover-penalised weights, rebalance every 5 trading days.
- Repeat the fit/test cycle every 2 months, walking forward with no overlap between train and test.
Key features
- Four factor families fused by a regularised rank regression, re-fit each walk-forward window
- Same specification runs unchanged on every supported dataset — crypto, US, China, HK, UK, FX and commodities
- Market-neutral L/S book with a long-only variant for allocators with no shorting mandate
- Turnover-penalised optimiser keeps annual turnover near 12× with a 50+ bps cost breakeven
Known limitations
- Capacity is universe-dependent — the crypto book saturates earlier than the S&P 500 book
- Factor weights adapt every two months; a regime break inside a window is carried until the next re-fit
Proposed improvements
Intra-window regime gate
A volatility-regime classifier can de-risk the book between re-fits.
Impact. Lower tail drawdowns in fast regime breaks.
Cross-dataset signal transfer
Factor weights learned on deep US cross-sections prime the smaller universes.
Impact. Faster warm-up and higher IC stability on the 10–50 name universes.
Jump to section