Apex Rotation: an EMA-ranked market-leader rotation with staggered sleeves and a margin-funded TQQQ sleeve

Working paper, October 2026. Backtests cover 2010-02-16 to 2026-09-18. Not a deployment instruction and not financial advice.

1. Strategy specification

Universe: the point-in-time union of S&P 500 and Nasdaq 100 members. A name is in the universe on date d if it is a member of either index on d, using membership intervals as known at the time.

Prices: daily open, high, low, and close, with closes adjusted for splits and dividends. All indicators are computed on the adjusted close series. Each name needs at least 200 trading sessions of history before its first decision date.

Eligibility, evaluated on each decision date d from the prior close: the name is in the universe on d; its adjusted close exceeds its 200-session simple moving average; and that average exceeds its own value 20 trading sessions earlier.

Ranking: eligible names are ranked by EMA50 divided by EMA200, where EMA_W is the exponential moving average with span W in the standard recursive form (smoothing factor 2/(W+1), seeded over at least W sessions). Higher ratios rank first.

Selection: the strategy holds up to four names. Buffer rule: a name held since the previous rebalance (an incumbent) is kept while it remains eligible and ranked within the top 12. Remaining slots are filled by the highest-ranked eligible names that are not incumbents.

Weighting: each held name is weighted in proportion to its estimated market capitalization on the decision date (adjusted close times shares outstanding) raised to the 1.5 power, renormalized across the names held. If all names lack a share count on a date, the sleeve falls back to equal weight for that rebalance.

Sleeves and timing: account equity is divided into four equal sleeves. Each sleeve rebalances every 20 trading days. The sleeves are phased 0, 5, 10, and 15 trading days after the anchor date (the first trading day on or after 2010-02-16), so roughly one quarter of the basket trades each week. Decisions use the prior close; trades execute at the next session's open. Every trade is charged 5 basis points of traded notional per side.

TQQQ sleeve: on a fixed monthly schedule (10 trading days after each month's first session), the TQQQ position is resized to 0.5 times account equity and held on margin.

Open and close logic: at each session open, in order, (1) any forced delever flagged by the prior close executes first, selling a pro-rata slice of every sellable rotation position and TQQQ at the open to move the account toward a 40% equity ratio, with 5 bps per side; (2) rotation sleeves scheduled to rebalance sell removed names at the open, then buy or resize kept names to their power-1.5 cap target notionals at the open, with 5 bps per side; (3) if the TQQQ schedule fires that day, the TQQQ position is resized toward 0.5x equity at the open, capped by the 75% initial margin requirement, with 5 bps per side. At each session close, in order, (1) all positions are marked to the adjusted close; (2) borrow interest accrues at 6%/360 on any cash debit; (3) the blended maintenance requirement is checked (25% of rotation market value plus 75% of TQQQ market value); a breach flags a forced delever for the next open, it does not trade at the close itself.

for each trading day d, in order:

  OPEN
  if a full liquidation was flagged at the prior close:
      sell every position at the open; the account stops
  elif a forced delever was flagged at the prior close:
      value the account at the open (open price, or adjusted close when the open is missing)
      sell a pro-rata fraction of every sellable rotation name and of TQQQ at the open,
      minus 5 bps per side, until the equity ratio moves back toward 40%
      if equity is not positive at the open: flag a full liquidation instead
  for each sub-sleeve rebalancing today:
      sell removed names at the open, minus 5 bps per side
      buy or resize kept names to their power-1.5 cap target notionals
      (each sleeve targets account value / 4), minus 5 bps per side
  if the TQQQ monthly schedule fires today:
      resize TQQQ toward 0.5x account value at the open, minus 5 bps per side,
      capped so the combined initial-margin requirement does not exceed account value

  CLOSE
  mark every position to the adjusted close
  accrue borrow interest at 6%/360 on any cash debit
  equity = cash + gross market value
  requirement = 25% of rotation market value + 75% of TQQQ market value
  if equity is not positive: flag a full liquidation for the next open
  elif equity < requirement: flag a forced delever for the next open
  no trades happen at the close

Margin model: rotation names carry 50% initial and 25% maintenance margin; TQQQ carries 75% initial and 75% maintenance, which caps the sleeve below 0.5x equity under normal conditions. Debit balances accrue 6% annualized (6%/360 per day). A blended account-level maintenance check runs daily; after any breach the account delevers pro-rata toward a 40% equity ratio at the next open.

Rolling evaluation: 151 four-year windows. Window starts are monthly from March 2010 through September 2022, each set 10 trading days after the month's first session; each window ends four calendar years after its start. CAGR is (end/start)^(1/years) - 1 with years measured in days/365.25; Sharpe is the daily log-return Sharpe annualized by sqrt(252); maximum drawdown is the minimum of equity divided by its running peak, minus one.

2. Assumptions

• Trading costs are 5 basis points of traded notional per side, covering commissions and slippage. There is no separate market-impact model; the backtest assumes the strategy's own trading does not move prices.
• Orders execute in full at the printed next-session open.
• Fractional shares; no lot-size or minimum-trade constraints.
• Returns are total-return: the adjusted close series reinvests dividends implicitly.
• Margin terms are fixed across the whole sample: 6% annual borrow rate, 50%/25% initial/maintenance on rotation names, 75%/75% on TQQQ. No margin calls beyond the modeled breach and pro-rata delever.
• No taxes and no account fees.
• TQQQ exposure is the actual TQQQ price series, so its fees and daily-reset decay are in the numbers, not modeled separately.

3. Data requirements

1. Point-in-time index membership for the S&P 500 and the Nasdaq 100: dated membership intervals per constituent, as known at the time, covering 2010 to 2026. Survivorship-biased or backfilled membership histories will inflate results.
2. Daily open, high, low, and close prices with split- and dividend-adjusted closes for every name that ever enters the universe, from at least mid-2009 (indicator warmup) through 2026-09-18. Both closes (for decisions) and next opens (for execution) are required.
3. Daily open, high, low, and close for TQQQ over the same span.
4. Shares outstanding per name for capitalization weighting, or equivalently a market-capitalization series. The results below use a recent share-count snapshot projected backward.
5. A trading calendar marking NYSE sessions, which drives the 20-day rebalance cadence, the sleeve phases, and the monthly TQQQ schedule.

4. Results

Net of costs. The hybrid is the rotation at 1.0x equity plus the 0.5x TQQQ margin sleeve.

The margin account never breached in the full window or in any of the 151 rolling windows; the lowest daily equity ratio in the whole sample was 48.9%. Annual turnover is about 569% of equity. The average largest single-name weight per sleeve is 63.4%.

5. Stress tests

All start from a fresh $1,000 account at the stated rebalance date.

Dec 2021 peak start: the account lost 44.2% in year one, suffered no margin breach, compounded at 79.9% annualized through September 2026, ended at 16.7x starting equity, and reclaimed its drawdown peak 9 months after the trough.

Feb 2020 shock start: the account gained 780.3% in year one and compounded at 122.4% annualized to September 2026, ending at 191.8x starting equity. Its -74.4% drawdown is the same 2022 episode as the full-window worst, not the 2020 crash itself. No margin breach.

Corrupted picks: for 36 months from the Dec 2021 start, each pick was given the price path of one of the worst-ranked eligible names. Full-horizon CAGR was 99.4% against 79.9% clean, with no breach. The corrupted-minus-clean gap ranges from -5.1 to +26.4 points across the four standalone phases.

Cost sensitivity: the full-window hybrid at 2, 5, and 10 bps per side returns 68.4%, 68.0%, and 67.5% annualized, with no breaches at any level.

References

Index membership history: unliftedq/index-constitution. Prices: Yahoo Finance daily OHLCV, cached through 2026-09-18. Share counts: Nasdaq quote API snapshot. K. Daniel and T. J. Moskowitz, "Momentum crashes," Journal of Financial Economics, 2016, cited for reversal mechanics.