Swiss Brix Swiss Brix

🎯 Score accuracy

We test our own Buy Score against historical appreciation — and publish the result unpolished. As of 2026-07-02.

The limits of this analysis, first
  1. Historical appreciation is backward-looking; the Buy Score is forward-looking. Agreement is evidence of calibration, not proof of predictive power.
  2. Our own daily price measurements only started in May 2026. A true prediction test follows once 6+ months of data exist (around November 2026).
  3. Two factors appear on both sides of the comparison (momentum ↔ 12-month growth, spread ↔ lifetime return). We therefore also show corrected values — the more honest numbers.
  4. 12-month growth data only exists for already-retired sets.

Buy Score vs. appreciation since release (annualized)

+0.78
Rank correlation (raw) · n=198
+0.44
corrected for spread circularity

By score decile: appreciation per year

Sets sorted into 10 equal groups by Buy Score. If the score works, the top groups must outperform.

Decile Score range Mean %/yr Median %/yr
117.0–24.0-2.3-2.5
224.0–31.0-2.1-2.1
331.0–33.0-2.0-1.7
434.0–38.0-0.3-0.4
538.0–42.0+0.6-0.8
642.0–47.0+1.9+2.8
747.0–51.0+5.5+5.9
851.0–55.0+6.8+6.0
955.0–64.0+8.8+8.3
1065.0–73.0+10.7+9.3

Top decile − bottom decile: +13.0 percentage points per year.

Buy Score vs. 12-month growth

+0.12
Rank correlation (raw) · n=88
-0.00
corrected for momentum circularity

Honest verdict: over the short term (12 months) the score currently shows no demonstrable discriminating power.

Which factors carry signal?

Rank correlation of each individual factor with appreciation since release. Circular pairings are flagged and don't count as evidence.

Factor Weight Correlation Finding
discount12constant — no variance
retirement18+0.17no signal
spread18(+0.93)circular — not evidence
forecast22+0.52moderate
momentum10+0.49moderate
theme20+0.13no signal

All numbers are recomputed daily from the same raw data that powers the set pages. Raw data: calibration.json