Investment intelligence & decision support · 2026-07-31 18:01 CT ·
not a signal dashboard, not a backtest dashboard
Given that SPY is my default position, does today's evidence justify doing
anything different?
No.
Every tested entry family underperforms SPY at every horizon with stops and exits
removed entirely, so the entries themselves — not the trade management — are what fail.
No candidate currently clears the frozen hurdle of +0.75pp annualised after-tax return with
drawdown no worse than SPY. Remain invested in SPY.
Default position
SPY
SPY vs 200-day
above
738.93 vs 698.54 · 2026-07-24
Live swing book
flat — no open positions
Research state
FROZEN
2026-07-31 · complete
The frozen decision rule
Maximise after-tax terminal wealth, subject to maximum drawdown no worse than SPY.
Materiality hurdle: a candidate must beat SPY by ≥ 0.75 percentage points of annualised
after-tax return while holding the drawdown constraint. Identical starting capital, SPY total
return with dividends reinvested, fair tax treatment on both sides, real cash yield on idle
capital, realistic costs.
Frozen 2026-07-31, before the next candidate exists. It may not be changed
because a future result happens to win some other metric. Naming the axis in advance is the
entire point — otherwise "materially better off" is unfalsifiable.
The one actionable finding
Weak 12-1 relative strength is avoidance information. The bottom relative-strength quintile
trailed SPY by 2.76% at 60 sessions, consistently, across every signal family tested.
Level 1 — flag, do not auto-block. Strong relative strength is not the
mirror image: the top quintile's excess collapses from a naive t of +4.41 to a month-clustered
t of +1.05, and 47 trades carry nearly all of it. Avoidance is supported; selection is not.
Names currently in Grimes "Underway"
Ticker
Side
Stop
Status
CBRL
long
55.08
descriptive only — not an entry trigger
Underway is a lagging description of favourable past movement, not evidence of
favourable future movement. Measured from the transition timestamp it trailed SPY by 3.74% at
60 sessions (t −2.34). Against matched control stocks every comparison was null (|t| < 1.0) —
neither skill nor anti-skill. Do not copy this list. Do not copy the public short list.
Setting Up (4 names) is research candidates only.
What has been closed, and on what evidence
Family
Result
n
Breakout geometry (Qullamaggie scan)
−0.66% vs SPY at 60 sessions, no stops or exits (t −3.3)
−0.106R/trade; capped 218 trades that would have passed +2R
7,478
Diversified 12-1 sleeve (after tax)
80/20 → 7.18% CAGR vs SPY 10.14% tax-deferred
58 months
Grimes public Underway list
−3.74% vs SPY at 60d from the transition (t −2.34)
916
All exploratory, on a survivorship-biased universe — which inflates results,
so these negatives are generous. ~110 cells searched on 2026-07-31.
What would change the answer
An information variable that is not price geometry — earnings surprise, estimate revisions,
institutional accumulation — since everything closed above is chart shape alone.
A materially lower-turnover structure, since 3.8× annual turnover and short-term tax destroyed
the only sleeve that showed pre-tax promise.
Genuinely new data, or an untouched dataset. Not another parameter variation.
Harness controls
Control
Guards against
test_excess_return_signs.py
Flipped benchmark leg on short positions.
Produced an invalid t of +3.57 on 2026-07-31; corrected to +0.80.
test_lookahead_audit.py
Conditioning on future information. Its synthetic
demo shows the bug fabricating t +11.51 from pure random walks with zero edge by construction.