Our 'Top Conviction' Leaderboard Beats the S&P by 4 Points a Year. It's Not an Edge.
June 29, 2026 · FindataFox
Our own "By Conviction" leaderboard — the highest-conviction institutional names — returned +19%/yr against the S&P 500's +15%. That looks like a 4-point edge. It isn't. Here's the difference between beating the market and having an edge — and why we won't sell you the first one dressed up as the second.
TL;DR
On our home page there's a leaderboard ranking stocks by conviction — the sum of every institution's portfolio weight in a name. The most-conviction stocks are the ones the "smart money" bets biggest on. So we backtested the obvious strategy: each quarter, buy the top-10 highest-conviction names, hold, repeat.
It returned +19.1%/yr vs the S&P's +14.8%. Four points of outperformance, 13 years running. Every hedge-fund-tracking site would slap that on a sales page.
Then we ran one control that killed it: a basket of just the 10 biggest stocks — zero analysis, no conviction data — beat the S&P by the same margin (more, out-of-sample). The conviction signal added nothing. The "edge" was never about what institutions know. It was "own the biggest stocks," during a decade when the biggest stocks happened to win.
Beating the S&P is not the same as having an edge. This is the single most important — and most counterintuitive — idea in honest investing research, so let's actually walk through it.
Where "beating the market" actually comes from
When any basket beats the S&P 500, the extra return comes from one of two places:
- A cheap factor/style tilt. You took a different shape of bet than the index — more concentrated, bigger-cap, more momentum — and that shape happened to do well. Anyone can buy this exposure in an ETF for three basis points. It is not skill.
- Skill. You picked winners that a same-shaped bet wouldn't have.
Only #2 is an edge. The whole job of a factor model (we use the standard four: market, size, value, momentum) is to subtract out #1 and see whether any #2 is left.
For our conviction leaderboard, after stripping the cheap factor exposures, the residual "alpha" is +8.1%/yr — but with a t-statistic of 1.58, meaning roughly an 11% chance it's pure luck. The bar for "statistically real" is t ≈ 2 (≤5% chance). It doesn't clear it. We literally cannot distinguish it from noise.
The control that settles it
The decisive test isn't the factor model — it's a basket a monkey could build. We ranked stocks by size ($ held) instead of conviction and bought the top 10:
| Top-10 basket, held one quarter | Return | vs S&P | Factor-adj. alpha (t) |
|---|---|---|---|
| By Conviction (the "smart money" signal) | +19.1%/yr | +14.8% | +8.1% (t=1.58) |
| By Size (just the biggest stocks) | +16.3%/yr | +14.8% | +6.2% (t=1.11) |
Same outperformance. Out-of-sample, the size basket was actually stronger (+20.1% vs +18.7%). So the conviction analysis added nothing you couldn't get by mechanically buying the largest companies. The outperformance is the mega-cap boom of 2012–2025 — AAPL, NVDA, MSFT, AMZN, GOOGL — not a signal about what institutions know.
And it's monotonic in the tell-tale way: the alpha shrinks as you widen the basket (top-10 +8.1% → top-100 +4.2%), exactly because the top is where the mega-caps concentrate. Dilute the mega-caps, dilute the "edge."
Why "+4 vs +6" is the wrong thing to stare at
The size of the margin doesn't matter — the source does. A bigger margin from the same contaminated source is still not skill. Three concrete reasons the raw "+4 over the S&P" doesn't bank:
- It's a tailwind, not a faster runner. Picture two runners: the S&P on flat ground, the conviction basket running downhill (the mega-cap tailwind). It finishes faster — but it didn't run better; it had a favorable slope. And the slope reverses. In 2000–2010, equal-weight and small-caps beat the mega-caps. The tilt that won last decade is a regime, not a durable edge.
- You took far more risk to get it. Ten concentrated stocks versus five hundred — higher beta, brutal drawdown potential. Beating the index by 4 points while running ~1.5× the risk isn't a free lunch. (That's why the risk-adjusted number, not the raw one, is what counts — and the risk-adjusted number isn't significant.)
- The real part is free. Whatever genuine factor tilt is in there — momentum, size — you can buy in an ETF for pennies, with no bet that your specific picks were the good ones. Paying for "smart-money skill" that's actually just cheap factor exposure is the con this whole site exists to call out.
What would have been an edge
A basket whose factor-adjusted alpha is positive, statistically significant (t ≳ 2), survives out-of-sample, AND beats the cheap size-matched control. That's "real skill the market hasn't priced." Our conviction leaderboard fails every one: not significant, doesn't beat the size control, regime-driven. We also tested changes in conviction — the names institutions were adding to or trimming most — and those were null too (no horizon, no basket size cleared significance).
The uncomfortable part: this is our own leaderboard
That "By Conviction" table is a real feature on our home page. People look at it. The honest thing to do with it is exactly what we just did: test whether it's tradeable, find that it's mega-cap size in a costume, and tell you. It's a fine descriptive view — "here's where institutional conviction is concentrated" is a true and interesting thing to see. It is not a buy list, and we won't pretend it is, even though the raw number is sitting right there begging to be a headline.
The honest caveats
- We excluded broad-market ETFs (SPY/IVV/VOO) from the leaderboard — they sit at the top of the conviction list, and "buy the S&P to beat the S&P" is not a stock-picking test. With them in, the numbers are lower.
- Returns model the 45-day filing lag, transaction costs, and booked delistings (survivorship). Factor-adjusted, vs the S&P and a size control, in- and out-of-sample.
- The point is not "mega-caps are a bad bet." It's narrower and more useful: the conviction signal added nothing beyond owning big stocks, and the result is regime-dependent and not statistically significant.
- Research and education, not investment advice. Past performance does not predict future results — which, again, is the entire finding.
Companion studies: we backtested the most popular ways to trade 13F filings — none beat the market · institutional skill doesn't persist · the famous stock-pickers are closet indexers. Method: Alpha Methodology. NOT investment advice.