$100,000 invested in 2015:
the JPI methods vs the S&P 500
20 stocks picked each year out of ~1,500, using only the information known before the buy date, sold 12 months later. 11 years in a row, nothing retouched afterwards. → Understand the method
⚡ JPI Risk +
$982,343
+23.1 % a year
+11.6 points a year above the index
Best year+76 %
Worst year-10 %
Beats index9 / 11
🎯 JPI Fair
$608,349
+17.8 % a year
+6.3 points a year above the index
Best year+43 %
Worst year-11 %
Beats index8 / 11
📉 S&P 500 — do nothing
$332,028
+11.5 % a year
the benchmark
Best year+29 %
Worst year-19 %
Beats index—
Same stake, same period, same market: $100,000 becomes $608,349 with JPI Fair and $982,343 with JPI Risk +, vs $332,028 simply tracking the index.
📊 Results by method
| Method | Return /yr i | Risk i | Ratio* i | 100k$ → compounded i |
|---|
*risk/return ratio (Sharpe) — higher is more efficient per unit of risk. A single period, excluding fees and taxes. Since 2026-07-27 the candidate universe is the real index composition at the buy date (no more stocks bought before they joined). A quality signal, not a guarantee. The current year (⏳) is not counted in the total (unfinished cohort).
⚙️
Universe
Buy month (start of month)
We count the analyst "right" when…
📈 The 11 years, one by one
The two methods and the S&P 500 side by side, year by year. Hover a bar to see which is which. Nothing is removed: losing years are shown like the rest.
JPI Risk +JPI FairS&P 500
🔎 The stocks picked, year by year
📅 See a dedicated page per year (2015→2025) with the 2 methods and the real result →
🔴 The 2026 portfolio in progress, tracked live →
What this backtest does not say
⚠️ 11 consecutive years, but a single era and a single market. 2015-2025 on a historically strong US market, before fees and taxes. Companies that left the indices have no analyst history on our side: the method could never pick them, and we can't measure what that changes. A quality signal, not a promise of returns.
Why the figures dropped on 27 July 2026
The backtest drew from the current index composition. So in January 2023 it "bought" Coinbase and AppLovin, which only joined the S&P 500 in 2025 — precisely because they had soared. We now rebuild the real composition month by month since 2015. JPI Fair went from +32.6% to +17.8 % a year. Less spectacular, and verifiable.
Why the current year is not counted
A selection is judged over 12 full months. The 2026 one only has a few: including it would inflate or drag the average depending on when you look. It is visible in the tabs, but excluded from every total.
Many of these stocks had already fallen
💡 Many of these picks had already fallen over 12 months at buy time — normal, the method targets potential (beaten-down stocks = big gap to target). Over 2021-2025, ~half the picks were fallen stocks, and they rebounded +38%/yr on average the next year (vs +33% for those already rising) — each return is over 12 months, not a multi-year hold. ⚠️ But this rebound is very uneven by year: it ranged from −18% in 2024 (the fallen stocks kept dropping) to +95% in 2023 (a huge rebound). A beaten-down stock can keep falling for another year before recovering.
A loss is capped at −100%. A gain is not.
💡 A loss is capped at −100%. A gain is not.
That's the whole logic of the method: we let winners run. A stock cut too early at +150% could have done +743% (TSLA 2020). We tested it over 11 years: taking profits at +150% costs −5 pts/yr, and a tight stop-loss costs too (stocks often rebound). Returns are made on the few winners you hold, not the losers you cut.