Model Proof

Prove whether high scores worked.

Start with top-score outcomes versus the S&P 500 baseline, then inspect the distributions and raw archived rows behind the proof.

Snapshot filters

Choose archive inputs

Forward performance is calculated from cached daily bars, not from live API calls.

Back to ranking Open factor study

Model Proof

2026-07-20 · 10D

twelvedata 0 checked / 500 total

Model Proof

Do high Cursor_18.25 scores perform better?

Compare top score groups against the same-day S&P 500 baseline using archived forward returns.

2026-07-20 · 10D 0 checked / 500 total 500 pending outcomes

Forward outcomes are still pending.

Choose an older snapshot or a shorter horizon to see checked model proof metrics.

Model Proof uses archived market data to evaluate research quality. It is not a trading result, prediction, or financial advice.

Avg forward return n/a

Median forward return n/a

Hit rate n/a

Avg max runup n/a

Avg drawdown n/a

Pending 500

Main score distribution

X axis shows score ranges: 50 → 80. Bar-top labels show count.

5D momentum v2 distribution

X axis shows score ranges: 0 → 35. Bar-top labels show count.

Score with momentum v2 distribution

X axis shows score ranges: 60 → 90. Bar-top labels show count.

Dollar volume distribution

X axis shows bucket midpoint: 10^7.8 → 10^9.6. Bar-top labels show count.

Risk level distribution

Elevated

Medium

103

Low

High

10D forward return distribution

No chart data yet.

Live comparison

Which chart looks strongest? · 2026-07-20 · 10D

How to read Corr, R², and n

Each scatter chart shows sample size, correlation, and explanatory power in the top-right box.

Solid line = linear trend Orange line = bucket average Top-right box = Corr / R² / n

Sample size (n) How many checked outcomes are included in the chart.

Correlation (Corr) Whether higher scores align with higher forward returns.

Explanatory power (R²) How much return variation the score linearly explains.

n (sample size)

Level	Approx.
Best	400+ (near full checked universe)
Usable	200+
Caution	<100 (Corr / R² can swing easily)

A small n can look good by chance even when Corr looks attractive.

Corr (correlation)

The model goal is higher score → higher forward return, so positive Corr is preferred.

Corr	Meaning
+0.15 ~ +0.30	Decent for noisy short-term stock returns
+0.30+	Strong; check whether it repeats across dates
~0	Little linear relationship
< 0	Direction reversed for this horizon

R² (explanatory power, shown as %)

R² will usually stay low. Do not expect 50% or 80%.

R²	Meaning
<2%	Very weak; score barely explains return
2% ~ 8%	Normal and usable for short-term finance research
8% ~ 15%	Relatively strong; worth a closer look
>15%	Uncommon; check sample size and sparse buckets

Compare across the five charts: prefer the chart with positive Corr and the highest R².