Strategy Iteration Report — 2026-03-03T09:43:40

All Variants Ranked by Composite Score

#	Name	CV	n_test	Spearman	p-value	Q5-Q1 med	Composite	Detail
1	small-cap+hiq	kfold	20	+0.806	0.0052 ***	+40.4%	1.937	2 folds
2	small-cap+hiq@pca50	kfold	20	+0.806	0.0052 ***	+40.4%	1.937	2 folds
3	small-cap+hiq@pca150	kfold	20	+0.806	0.0052 ***	+40.4%	1.937	2 folds
4	small-cap+hiq@pca200	kfold	20	+0.806	0.0052 ***	+40.4%	1.937	2 folds
5	small-cap+hiq@auto-pca	kfold	20	+0.806	0.0052 ***	+40.4%	1.937	2 folds
6	baseline-no-fund+hiq	kfold	51	+0.248	0.3290	+17.1%	1.739	5 folds
7	baseline-no-fund+hiq@pca50	kfold	51	+0.248	0.3290	+17.1%	1.739	5 folds
8	baseline-no-fund+hiq@pca150	kfold	51	+0.248	0.3290	+17.1%	1.739	5 folds
9	baseline-no-fund+hiq@pca200	kfold	51	+0.248	0.3290	+17.1%	1.739	5 folds
10	baseline-no-fund+hiq@auto-pca	kfold	51	+0.248	0.3290	+17.1%	1.739	5 folds
11	small-cap	kfold	22	+0.568	0.0691 *	+33.5%	1.597	2 folds
12	small-cap+hiq@log-flip	kfold	20	+0.709	0.0223 **	+32.2%	1.570	2 folds
13	small-cap+long	kfold	20	+0.606	0.1926	+32.4%	1.320	2 folds
14	best@90d	kfold	20	+0.370	0.3305	+23.0%	0.933	2 folds
15	baseline-no-fund-log+hiq	kfold	51	+0.172	0.4638	+6.1%	0.623	5 folds
16	baseline-no-fund+hiq@log-flip	kfold	51	+0.172	0.4638	+6.1%	0.623	5 folds
17	baseline-no-fund+value	kfold	28	+0.128	0.7483	+3.8%	0.220	3 folds
18	baseline-no-fund-log+value	kfold	28	+0.114	0.6026	+2.0%	0.119	3 folds
19	baseline-no-fund	kfold	55	+0.215	0.4894	-0.6%	-0.044	5 folds
20	baseline-no-fund-log	kfold	55	+0.115	0.4851	-1.8%	-0.172	5 folds
21	baseline-log	kfold	52	-0.055	0.3392	-2.8%	-0.285	5 folds
22	value-only	kfold	27	-0.178	0.6765	-7.3%	-0.406	3 folds
23	long-only	kfold	47	+0.031	0.6192	-4.6%	-0.429	5 folds
24	baseline-no-fund-log+long	kfold	49	-0.131	0.5614	-5.4%	-0.543	5 folds
25	baseline	kfold	52	-0.118	0.3398	-7.3%	-0.739	5 folds
26	baseline-no-fund+long	kfold	49	-0.010	0.3436	-9.0%	-0.887	5 folds

Name

n_test

Spearman

p-value

Q5-Q1 med

Composite

Detail

small-cap+hiq

kfold

+0.806

0.0052 ***

+40.4%

1.937

2 folds

small-cap+hiq@pca50

kfold

+0.806

0.0052 ***

+40.4%

1.937

2 folds

small-cap+hiq@pca150

kfold

+0.806

0.0052 ***

+40.4%

1.937

2 folds

small-cap+hiq@pca200

kfold

+0.806

0.0052 ***

+40.4%

1.937

2 folds

small-cap+hiq@auto-pca

kfold

+0.806

0.0052 ***

+40.4%

1.937

2 folds

baseline-no-fund+hiq

kfold

+0.248

0.3290

+17.1%

1.739

5 folds

baseline-no-fund+hiq@pca50

kfold

+0.248

0.3290

+17.1%

1.739

5 folds

baseline-no-fund+hiq@pca150

kfold

+0.248

0.3290

+17.1%

1.739

5 folds

baseline-no-fund+hiq@pca200

kfold

+0.248

0.3290

+17.1%

1.739

5 folds

baseline-no-fund+hiq@auto-pca

kfold

+0.248

0.3290

+17.1%

1.739

5 folds

small-cap

kfold

+0.568

0.0691 *

+33.5%

1.597

2 folds

small-cap+hiq@log-flip

kfold

+0.709

0.0223 **

+32.2%

1.570

2 folds

small-cap+long

kfold

+0.606

0.1926

+32.4%

1.320

2 folds

best@90d

kfold

+0.370

0.3305

+23.0%

0.933

2 folds

baseline-no-fund-log+hiq

kfold

+0.172

0.4638

+6.1%

0.623

5 folds

baseline-no-fund+hiq@log-flip

kfold

+0.172

0.4638

+6.1%

0.623

5 folds

baseline-no-fund+value

kfold

+0.128

0.7483

+3.8%

0.220

3 folds

baseline-no-fund-log+value

kfold

+0.114

0.6026

+2.0%

0.119

3 folds

baseline-no-fund

kfold

+0.215

0.4894

-0.6%

-0.044

5 folds

baseline-no-fund-log

kfold

+0.115

0.4851

-1.8%

-0.172

5 folds

baseline-log

kfold

-0.055

0.3392

-2.8%

-0.285

5 folds

value-only

kfold

-0.178

0.6765

-7.3%

-0.406

3 folds

long-only

kfold

+0.031

0.6192

-4.6%

-0.429

5 folds

baseline-no-fund-log+long

kfold

-0.131

0.5614

-5.4%

-0.543

5 folds

baseline

kfold

-0.118

0.3398

-7.3%

-0.739

5 folds

baseline-no-fund+long

kfold

-0.010

0.3436

-9.0%

-0.887

5 folds

Skipped Variants

Name	Status	Reason	n_filtered
short-only	skipped	Too few samples: 5 < 20	5
growth-only	skipped	Too few samples: 14 < 20	14
short-thesis	skipped	Too few samples: 5 < 20	5
special-sit	skipped	Too few samples: 3 < 20	3
mid-cap	skipped	Too few samples: 11 < 20	11
large-cap	skipped	Too few samples: 19 < 20	19
tech-only	skipped	Too few samples: 6 < 20	6
financials-only	skipped	Too few samples: 3 < 20	3
small-cap+value	skipped	Too few samples: 10 < 20	10
best@365d	skipped	Too few samples: 12 < 20	20

Name

Status

Reason

n_filtered

short-only

skipped

Too few samples: 5 < 20

growth-only

skipped

Too few samples: 14 < 20

short-thesis

skipped

Too few samples: 5 < 20

special-sit

skipped

Too few samples: 3 < 20

mid-cap

skipped

Too few samples: 11 < 20

large-cap

skipped

Too few samples: 19 < 20

tech-only

skipped

Too few samples: 6 < 20

financials-only

skipped

Too few samples: 3 < 20

small-cap+value

skipped

Too few samples: 10 < 20

best@365d

skipped

Too few samples: 12 < 20

Rounds

Round 1: Baselines + Sweeps

7/15 variants evaluated successfully

Top 3: small-cap, baseline-no-fund, baseline-no-fund-log

Round 2: Combinations

8/9 variants evaluated successfully

Top 3: small-cap+hiq, baseline-no-fund+hiq, small-cap+long

Round 3: Parameter Tuning

10/10 variants evaluated successfully

Top 3: small-cap+hiq@pca50, small-cap+hiq@pca150, small-cap+hiq@pca200

Round 4: Horizon Check

1/2 variants evaluated successfully

Top 3: best@90d

Best Variant: small-cap+hiq

Fold-by-Fold

Fold	n_train	n_test	Spearman	p-val	Q5-Q1 med	Ridge α	PCA dims	Bootstrap CI
fold_0	10	10	+0.830	0.0029	+60.8%	1.0	9	[+0.423, +0.987]
fold_1	10	10	+0.782	0.0075	+19.9%	1.0	9	[+0.245, +0.963]

Fold

n_train

n_test

Spearman

p-val

Q5-Q1 med

Ridge α

PCA dims

Bootstrap CI

fold_0

+0.830

0.0029

+60.8%

1.0

[+0.423, +0.987]

fold_1

+0.782

0.0075

+19.9%

1.0

[+0.245, +0.963]

Aggregate: Spearman = +0.806 ± 0.024, Q5-Q1 median spread = +40.4%, composite = 1.937

Key Findings

26/36 variants evaluated successfully, 10 skipped.

Best variant: small-cap+hiq — Spearman=+0.806, p=0.0052, composite=1.937

Significant at p<0.05: 6/26. At p<0.10: 7/26.

LONG-only (+0.031) outperforms baseline (-0.118).

WARNING: Small test set (20 samples, kfold CV). Results are noisy — grow the dataset before drawing conclusions.

finance/eval — autonomous iteration loop. Data: iterations/

Horizon	30d
Direction	all
Cap filter	small
Thesis types	all
Industry	all
Fundamentals	yes
Log transform	no
PCA	100