PHOENIX → NOVA (current)

NOVA — Time-Stacked LETF Overnight + Intraday Ensemble

Four time-disjoint sleeves on one capital pool | LETF substitution for margin | Daily execution at 09:30 & 15:55 | No broker leverage, no continuous vol scaling

Strategy Overview

NOVA is a high-CAGR research strategy that stacks four independent alpha sleeves across time-disjoint windows on the same capital base, using leveraged ETFs in place of broker margin. Backtested CAGR is 40.4% at Sharpe 1.69 over 8.9 years, versus SPY 15.1% / 0.80. Absolute return is roughly 3× SPY with comparable realised volatility (23.9%) and a shallower maximum drawdown (−29% vs SPY’s −34% since inception).

Key idea. A trading day has three disjoint windows: overnight (15:55 → 09:30), daytime (09:30 → 15:55), and weekend (Friday close → Monday open). Each sleeve holds in one window, so the four sleeves can run at full notional on the same cash pool — this is capital stacking, not leverage. No broker margin is used.

LETF substitution. For the overnight equity/gold sleeve, NOVA holds UPRO (3×SPY), TQQQ (3×QQQ) and UGL (2×GLD) at equal dollar weight. The much-discussed LETF “decay” is an intraday phenomenon driven by daily resets; the close-to-open window captures ~2–3× the 1× overnight drift without paying broker interest. Expense ratios (0.84–0.95% per year) and swap costs are already embedded in the realised close-to-open returns.

Honest comparison vs HYDRA. HYDRA runs 21 years with full-sample Sharpe 1.58 / OOS 2.01 / CAGR ~12%. NOVA runs 8.9 years with full Sharpe 1.69 / OOS 1.61 / CAGR ~40%. NOVA trades some risk-adjusted quality (OOS Sharpe is lower than HYDRA’s) and a shorter track record for roughly 3× the absolute CAGR. Drawdown is deeper (−29% vs HYDRA’s −19%). Pick based on whether absolute compounding or stability matters more.

What it is not. Not leveraged in the account sense (no margin interest paid). Not vol-targeted (no continuous vol scaling). Not curve-fit to a hindsight sleeve subset — the four sleeves were published individually before the stack was assembled. IS/OOS split is 2022-01-01 and was fixed before the LETF substitution experiment.

How Rebalancing Works

Overnight sleeveAt 15:55 each day, if the 20-day rolling 5-min realised vol for the name sits below 0.15 annualised, enter UPRO/TQQQ/IWM/DIA/UGL at equal $ weight into the overnight window. Exit at the next day’s open (09:30). When the vol gate is off, park in BIL. TC 2 bps on every active night.
Daytime TSMOMAt 09:30, rebalance a 12-ETF time-series momentum basket (21-day lookback). Hold through the session and exit at 15:55. 5 bps TC per day.
Stock L/S overnightCross-sectional long top decile / short bottom decile of 96 liquid single names ranked by trailing overnight drift. Entry 15:55, exit next open. Dollar-neutral. 10 bps TC per rebalance.
Weekly overnightEqual-weighted long basket of SPY/QQQ/IWM/DIA/GLD held Friday close → Monday open only. 3 bps TC per week.
Signal LagAll signals are computed from prior-bar data. Overnight gates use T−1 close 5-min RV; daytime TSMOM uses T−1 close prices. No look-ahead.
Capital stackingBecause the four sleeves hold in non-overlapping windows, a single $1 of capital is fully deployed in at most one sleeve at a time. Portfolio return = sum of four sleeve returns (each applied at 1.0 notional). No margin, no borrow cost.
Concrete Example — A Tuesday in 2024:
Mon 15:55
Overnight RV gates checked. UPRO/TQQQ/UGL all pass (RV < 0.15). Buy each at 1/5 of capital along with 1/5 each in IWM and DIA.
Tue 09:30
Sell overnight basket at open. Receive proceeds. Daytime TSMOM sleeve computes 21-day momentum, rebalances into its 12-ETF basket with the same $ pool.
Tue 15:55
Sell daytime basket. Re-run overnight gates for Tuesday night. Re-enter UPRO/TQQQ/IWM/DIA/UGL. Concurrent stock L/S sleeve sets its dollar-neutral overnight positions.
Fri 15:55 → Mon 09:30
Weekly-overnight sleeve takes a long SPY/QQQ/IWM/DIA/GLD basket into the weekend window. Captures weekend gap.
Growth of $10,000 (NOVA vs 4x margin vs 1x baseline vs SPY)
Comparison bars. Thin dashed “1x baseline” shows the same four sleeves run without any leverage (no LETFs, no margin). The “4x margin” line shows the 1x baseline levered 4× with 8.25% annual interest drag netted off — the broker-margin alternative that NOVA’s LETF substitution replaces. NOVA ends meaningfully above 4x margin because it avoids ~8%/yr in interest drag while capturing similar gross exposure through the LETFs.
In-Sample vs Out-of-Sample

In-Sample (2017-06 → 2021-12)

Out-of-Sample (2022-01 → present)

Interpretation. IS Sharpe 1.77 / OOS Sharpe 1.61 — a modest OOS decay, consistent with honest construction. CAGR is essentially flat across IS and OOS (40.1% vs 40.7%). The 2022 drawdown (−12% return year) is fully inside the OOS window; the strategy recovered into 2023–2025 without needing re-tuning.
Comparison with Other Strategies on This Site
How to read this. NOVA is the high-CAGR option. It clears HYDRA on full-sample Sharpe and on absolute return, but HYDRA still wins on OOS Sharpe and on drawdown. Pick NOVA if absolute compounding matters more than stability; pick HYDRA if you want the best risk-adjusted track record. Backtest windows differ — HYDRA’s 21-year run stresses the strategy across more regimes than NOVA’s 8.9 years.
Walk-Forward — Rolling 3-Year Windows
Three non-overlapping 3-year windows. NOVA is net-positive in all of them and beats SPY on Sharpe in 2 of 3.
Trailing Returns
Performance & Risk Statistics

Performance (full sample)

Risk (full sample)

Drawdown
Rolling 6-Month Sharpe
Calendar Year Returns
Monthly Returns Heatmap
Sleeve Decomposition
Methodology Notes & Caveats
Backtest disclosure. All figures are from a daily-bar simulation from 2017-06-01 to the most recent close shown above. Overnight 5-min RV gates and close-to-open prices are computed from minute-level bars; the LETF overnight returns come from daily open/close published by each sponsor and therefore embed LETF expense ratios and swap costs. Per-sleeve transaction costs are applied on turnover as stated above. Signals use T−1 data applied from T onward (one full bar lag). The strategy uses leveraged ETFs — these products carry compounding risk on multi-day holds; NOVA’s single-overnight holding pattern largely avoids that risk in backtest but does not eliminate the possibility of a −60% intraday LETF move in a large equity gap (March 2020 precedent). Past results are not a guarantee of future performance. This factsheet is for informational purposes only and is not investment advice.