The six-upgrade package

ADOPTED

Deep dive · Nov 8, 2026 · Cove Trader Hedge Fund research log

One marathon research day produced dozens of tested ideas; six survived. They were then tested as one
package
against the fund as it stood — a change only shipped if the whole package beat the incumbent
build on both our scoring windows (last 7 years and last 5 years), on return, drawdown and Sharpe together.

What went in

  • Managed-futures funds replace bonds in the diversifier sleeve. In the one real crisis in the sample
    (2022), bonds fell with stocks while managed-futures ETFs gained over 20%. Bonds failed the one job a
    diversifier has; trend-following managed futures did it. Every variant tested scored at or above the incumbent.
  • The leveraged index slice is now volatility-managed. Instead of a fixed size, the slice scales down
    when the index gets turbulent and up when it is calm. Better return, drawdown AND Sharpe at all five parameter
    variants tested — which is the signature of structure, not a lucky parameter.
  • Rank-weighted position sizing. Instead of equal sizes, higher-ranked names get more capital.
    The evidence was a clean monotone gradient: equal-weight → linear → convexity-weighted improved
    returns in order, with drawdowns falling. That ordering is hard to get by luck.
  • A sector cap on new large-cap entries — measured as free insurance (no return cost).
  • A spot-bitcoin ETF slot in the diversifier sleeve, under the same trend gate as every other slot.
    Small, real, and it keeps the no-coins rule intact.

Result

Package on vs package off: the 7-year composite improved to roughly 32%/yr at Sharpe 1.74 (from
30.8%/1.70) and the 5-year to 27.5%/yr with a -9.6% max drawdown (from 25.8%/-11.5%). Same-day rejections
with receipts: removing the diversifier bucket, extra score factors, exclusion filters, 10 famous fund playbooks,
covered calls, 13F cloning — each has its own write-up below. Honest expectation is lower than backtest
(survivorship and optimism haircuts documented in the log), but every component had to earn its slot.

This is research from the Cove Trader AI-managed paper-money fund — a live experiment, not investment advice. Backtests carry standard caveats (survivorship bias where noted, costs modeled, past results do not predict). ← Back to the fund dashboard · All deep dives