QU100 Top-N Backtest — with a QQQ 55-day VWAP Regime Gate

Same strategy as the base sweep — each ranking day at the close, hold an equal-weight basket of the QU100 money-flow top-N names, rebalance daily — but now overlaid with a single market-regime switch: stay invested only when QQQ is above its 55-day rolling VWAP; otherwise sit in cash. Start $100,000. Gross of trading costs (0 bp).

Window: 2020-10-01 → 2026-06-04 · 1,402 ranking-days (~5.7 yrs) · rankings from QU100 top100 · basket prices = yfinance adjusted close · N swept 1 → 100. Gated vs. ungated shown side-by-side.

hlc3      = (high + low + close) / 3                       # typical price, per bar
VWAP55[t] = sum(hlc3 * volume, last 55 bars) / sum(volume, last 55 bars)

1 · Per-N table — gated, with ungated reference (0 bp, gross, N=1..100)

First eight metric columns are the gated strategy; the last three (lighter) are the ungated baseline pulled from the prior sweep for side-by-side comparison. Green row = best gated Sharpe (N=44); blue-tinted row = best gated return (N=6). SPY/QQQ rows are buy-and-hold over the identical window (always 100% invested, no gate). Max drawdown is close-to-close at ranking-day granularity. Time-in-market is 71.6% for every N — the gate decision is the same QQQ signal regardless of basket size.

2 · Total return vs N — gated vs ungated

The gated line sits below the ungated line almost everywhere — the cash periods cost raw return. The gap is widest in the high-return mid/upper range (~N=24–100), where the ungated book rode the full recovery while the gated book sat out chunks of it. Both clear SPY across most of the range; the gated book trails QQQ on raw return at high N but wins on risk-adjusted terms (next chart).

3 · Max drawdown vs N — gated vs ungated

Less-negative is better. The gated drawdown line is shallower than ungated across the whole sweep — the cash periods (chiefly 2022) clip the deepest losses. The improvement is largest in the mid-range (N≈24–55), where gated max DD tightens to roughly −40% vs the ungated −46–48%.

4 · Sharpe vs N — gated vs ungated

The gated Sharpe line sits above ungated across essentially the entire range — lower return but proportionally much lower volatility and drawdown nets out positive on a risk-adjusted basis. Gated Sharpe clears 1.0 across roughly N=30–90 and tops out at 1.07 (N=44), above QQQ buy-and-hold (0.90) and edging SPY (1.02) at the peak.

5 · Headline side-by-side

At N=44 the gate trades ~56 points of gross return for a 7-point shallower drawdown and a 0.07 higher Sharpe. The gated best-Sharpe basket beats QQQ buy-and-hold on Sharpe (0.90) and clears SPY's drawdown discipline far better than the ungated book, though still at higher volatility than either index.

6 · Per-year performance (calendar-year returns, gross, 0 bp)

Per-calendar-year total return for the ungated basket vs the QQQ 55-day VWAP gate basket, alongside SPY and QQQ buy-and-hold, at the two representative basket sizes. Each year = equity at the last ranking-day of the year over the prior year-end (2020 = Q4 partial from the 2020-10-01 start; 2026 = YTD through 2026-06-04). Year-by-year returns compound to the headline totals shown above.

N=6 — best gated raw return

N=44 — best gated Sharpe

The 2022 row is where the gate earns its keep. Holding cash for ~71% of the bear year, the VWAP gate cut the ungated basket's 2022 loss sharply (N=6: −47.6% → −26.0%; N=44: −27.7% → −10.6%) — beating QQQ's −32.6% in both. The cost is up-year give-back: the single-threshold VWAP whipsaws in and out, so in trending years it lags the ungated basket (N=44 2021: +18.6% → +2.6%; 2023: +23.2% → +15.0%; 2026 YTD: +55.8% → +42.8%). You pay in bull-market upside to buy a smaller drawdown in the bear — net positive on Sharpe across the window, but a larger raw-return give-up than the MA gate.

7 · Bottom line

• The gate works as intended. Across nearly all N, the VWAP55 overlay raises Sharpe and shrinks max drawdown versus ungated, at the price of lower gross return. It is a risk-management overlay, not a return amplifier.
• 2022 is the whole story. The gate sits in cash ~71% of 2022's ranking-days, which is precisely the bear market that produced the ungated book's deepest drawdowns. Sidestepping it is where the −47% → −40% (best-N) drawdown improvement comes from.
• Best-N shifts. Best Sharpe stays at N=44 (1.07); best raw return moves down to N=6 (+241%) — with the worst market stretches removed, modest concentration is less punishing.
• The cost of a single threshold: whipsaw. 74 regime flips over 5.7 years, with 71.6% time in market. A buffer/hysteresis band would cut flips (and turnover) but is explicitly out of scope here — this is the raw single-threshold result.
• Costs still excluded. 0 bp throughout. The gate adds turnover (full liquidation on every risk-off flip), so a costed run would erode the gated edge further than the ungated one. Lower-turnover gate variants are the natural next study.