CEO/TFTF Iteration Loop — Weight Variants
Variants: 17 | Horizons: [-90, -180, -365] | As-of: 2026-02-14 | Time: 0.3s
Top findings:
- moat_focused (avg |ρ| = 0.296): trail 90d ρ=+0.342 trail 180d ρ=+0.292 trail 365d ρ=+0.255
- founder_plus_moat (avg |ρ| = 0.261): trail 90d ρ=+0.309 trail 180d ρ=+0.261 trail 365d ρ=+0.212
- tftf_plus_founder_rating (avg |ρ| = 0.247): trail 90d ρ=+0.274 trail 180d ρ=+0.249 trail 365d ρ=+0.219
Variant Rankings (by avg |Spearman ρ|)
| # | Variant | Strategy | Avg |ρ| | Best ρ | Best H | ρ trail 90d | ρ trail 180d | ρ trail 365d |
|---|
| 1 |
moat_focused |
thematic |
0.296 |
+0.342 |
trail 90d |
+0.342 | +0.292 | +0.255 |
| 2 |
founder_plus_moat |
thematic |
0.261 |
+0.309 |
trail 90d |
+0.309 | +0.261 | +0.212 |
| 3 |
tftf_plus_founder_rating |
composite |
0.247 |
+0.274 |
trail 90d |
+0.274 | +0.249 | +0.219 |
| 4 |
emphasize_moat_depth |
single_emphasis |
0.233 |
+0.268 |
trail 90d |
+0.268 | +0.243 | +0.189 |
| 5 |
value_focused |
thematic |
0.224 |
+0.261 |
trail 90d |
+0.261 | +0.236 | +0.176 |
| 6 |
tftf_only |
subset |
0.205 |
+0.238 |
trail 180d |
+0.213 | +0.238 | +0.163 |
| 7 |
emphasize_founder_intensity |
single_emphasis |
0.204 |
+0.224 |
trail 90d |
+0.224 | +0.224 | +0.164 |
| 8 |
emphasize_talent_magnetism |
single_emphasis |
0.195 |
+0.219 |
trail 90d |
+0.219 | +0.219 | +0.146 |
| 9 |
emphasize_capital_efficiency |
single_emphasis |
0.191 |
+0.213 |
trail 180d |
+0.201 | +0.213 | +0.159 |
| 10 |
talent_plus_velocity |
thematic |
0.184 |
+0.212 |
trail 90d |
+0.212 | +0.212 | +0.127 |
| 11 |
equal_weight |
baseline |
0.170 |
+0.201 |
trail 180d |
+0.176 | +0.201 | +0.134 |
| 12 |
emphasize_velocity |
single_emphasis |
0.170 |
+0.201 |
trail 180d |
+0.176 | +0.201 | +0.134 |
| 13 |
founder_focused |
thematic |
0.130 |
+0.182 |
trail 90d |
+0.182 | +0.134 | +0.073 |
| 14 |
growth_focused |
thematic |
0.126 |
+0.177 |
trail 180d |
+0.128 | +0.177 | +0.073 |
| 15 |
execution_focused |
thematic |
0.119 |
+0.170 |
trail 180d |
+0.109 | +0.170 | +0.079 |
| 16 |
emphasize_compounding |
single_emphasis |
0.103 |
+0.160 |
trail 180d |
+0.086 | +0.160 | +0.061 |
| 17 |
people_only |
subset |
0.096 |
+0.117 |
trail 365d |
+0.105 | +0.068 | +0.117 |
Interpretation
- Positive ρ = higher weighted score → higher excess return (desired)
- Avg |ρ| measures consistency across horizons — variants that predict at multiple horizons are more robust
- With N=10, |ρ| > 0.5 is a strong signal; |ρ| > 0.3 is worth investigating
- Next iteration: take top variants, generate perturbations around them
Companies Used
ALGM (Allegro MicroSystems), INDI (indie Semiconductor), 6920.T (Lasertec), LSCC (Lattice Semiconductor), MTSI (MACOM Technology Solutions), 6963.T (ROHM Semiconductor), SMTC (Semtech), SLAB (Silicon Laboratories), SITM (SiTime), 6526.T (Socionext)