🪞 Francois Rochon · Giverny
Quality compounders.
Hit rate is measured, not predicted: across 2,030 investor-quarters of walk-forward holdout, the name this model ranked first was the investor’s actual next new buy 1.7% of the time, against 0.62% for a name drawn at random from the same board. That is 2.77x chance. The model ranks the whole board; AUC and the rest of the scoring detail are on the how it works page.
Not separated from chance. Across 7 quarters this ranking scored 0.505 against 0.500, two-sided p = 0.9196. An edge above 0.596 would have registered on this board, so the test could have resolved it and did not. A figure under 0.500 here is the sampling noise around a coin flip, not a signal pointing the wrong way. Pooling every scored row instead of averaging the quarters, which uses the big quarters and the thin ones at their real weight, the same board reads 0.524 over 4,736 rows (two-sided p = 0.611).
Build a basket from the top picks →
🔮 Predicted next buys: quarter ending 2026-09-30
The model orders Francois Rochon · Giverny’s candidates for the next 13F filing. No public confirmation yet. The rate beside each name is the model’s own score rewritten as a measured quantity: how often a name in that position turned out to be the next buy across 2,030 investor-quarters of holdout. Picking at random from this board lands 0.62%.
| # | Ticker | Company | Measured hit rate | vs base |
|---|---|---|---|---|
| 1 | WSO | Watsco Inc | 1.8% | 2.9x |
| 2 | RRC | Range Resources Corp | 1.7% | 2.8x |
| 3 | VEEV | Veeva Systems Inc-Class A | 1.7% | 2.8x |
📉 Predicted exits, conviction fading
Of Francois Rochon · Giverny’s current disclosed holdings, the names a model flags as most likely to be EXITED or TRIMMED next quarter, before the 13F shows it. Model score (not a calibrated probability), ranked. Holdings as of 2026-06-30.
| Ticker | Company | Weight | Last move | Sell likelihood |
|---|---|---|---|---|
| TSM | Taiwan Semiconductor Mfg Ltd | 0.06% | trimmed last Q | ▲▲ 69 |
| CPRT | Copart Inc | 0.01% | trimmed last Q | ▲▲ 67 |
| ANET | Arista Networks Inc | 1.94% | trimmed last Q | ▲▲ 62 |
⚠️ This predicts the ACTION (whether Francois Rochon · Giverny reduces the position), not a return and not a short. A 13F shows long positions only, so a “sell” here means exiting or cutting a long, never shorting the stock. Walk-forward validated on the decision (AUC ≈ 0.66), but acting on it is not shown to make money. Research, not advice.
📜 Top disclosed holdings: Jun 2026
| Ticker | Company | Weight | Last change |
|---|---|---|---|
| GOOG | Alphabet Inc C | 7.19% | ✊ Held |
| HEI/A | Heico Corp New | 6.38% | ✊ Held |
| META | Meta Platforms Inc | 5.74% | ✊ Held |
| SCHW | Schwab Charles Corp | 5.22% | ✊ Held |
| ADP | Automatic Data Proc Inc | 4.93% | ➕ Added |
Source: SEC Form 13F (public domain), long U.S. equity positions only, disclosed up to 45 days after quarter-end. Top 15 of 51 shown.
💲 Where they bought: accumulation cost
13F shows holdings, not trade prices. Implied average cost is total value ÷ shares; the zone is the price range that quarter, where they accumulated. “vs cost” is today versus that average. This is where they bought, not a future price.
| Ticker | Avg cost | Accumulation zone | Today | vs cost |
|---|---|---|---|---|
| GOOG | $353.33 | $287 to $399 | $350.87 | -0.7% |
| HEI-A | $257.91 | $201 to $260 | $272.21 | +5.5% |
| META | $563.29 | $543 to $688 | $665.75 | +18.2% |
🧬 Factor fingerprint, the style behind the book
77% of Francois Rochon · Giverny’s day-to-day swings are explained by public, buyable factors (Fama-French 5 + momentum), regressed on their current disclosed book (12 of 12 names, ~406 trading days). Market beta 0.96.
⚠️ This is the part the great investors can’t hide: their style is a tilt toward known factors, and every one of these is sold as a cheap index ETF. Buying the factors gets you most of the book for a few basis points. What’s left after the factors (the residual) is noisy on a single concentrated book and is not a reliable edge, read this as transparency about the style you’re copying, not a secret signal. Loadings are in-sample on current static weights, U.S. long positions only. Research, not advice.
📈 This mirror’s track: backtest vs S&P 500
⚠️ Read this honestly. Leak-free walk-forward backtest: top-10 predicted new buys ranked from a broad ~2,000-name universe known at the rebalance date, equal-weight, 1-quarter holds, complete quarters only, gross (before costs). Small samples, per-investor edges are noisy and not statistically significant. The consensus mirror's backtest shows +2.2 pts/quarter net vs the S&P 500 over 35 complete quarters (2017Q3 to 2026Q1), directional, NOT statistically significant (t≈1.3; 95% CI includes zero). Most of that average is one quarter: 2026Q1 posted a +47.5 pt edge, roughly half of it two AI-hardware names (SNDK +258%, AMD +186%) marked at what proved to be a local peak; excluding it the edge is +0.9 pts/quarter. Past performance does not predict future results. Not investment advice.
🔔 Follow this mirror
New filings and predicted-buy refreshes are published to every reader at once, weekly digest (email) · build a basket.
Prior Moves is a research publication: one impersonal model portfolio per investor, identical for every reader. You place any trades yourself at your own broker. No execution, no custody, no individualised advice.