Why we publish the screen together with the scoring rules
Most stock-pitch content has a built-in cushion: it shows you only today's call and never circles back to how the call from three months ago actually played out. This article does the opposite. It is the first piece in our "Public Review" column, and we are posting, exactly as they stood on 2026-06-12, the opportunity list our whole library produced that day, the screening definitions, and the snapshot price of every name. We also write down the tests we will hold ourselves to before any of it resolves: at post-mortem we answer each one, append the conclusion at the foot of this article, and change not a single word of the body above. If the calls were right, the methodology earns credit. If they were wrong, we say which axis broke and how it gets fixed.
The headline for this cut, up front: of the 760 reports in the library, 15 are simultaneously high-score and low-valuation, and exactly one already trades below its report's fair buy price. That name is CATL.
How we screened: two yardsticks that move independently
Every report, once it lands in the library, runs through the "Baillie Framework: Ten Questions on Growth Investing" scorecard. Ten dimensions, each scored 0–10, for a total of 0–100, measuring how strong the company is along the axes of long-run growth (the full board sits on the Baillie leaderboard). That is the first yardstick: quality.
The second yardstick is valuation position (vp from here on): stretch the report's range from the floor of the bear case to the ceiling of the bull case, then read where today's price falls on that scale. A value of 0 means the price is pressed against the bear-case floor; 1 means it has reached the bull-case ceiling; below 0 means it has already broken through the report's most bearish scenario.
The premise that makes this work is that the two yardsticks are independent. As of this cut, the average total across every valuation band in the library lands inside the narrow 40.1–43.2 range, so high-score companies are not systematically more expensive. The axes being uncorrelated is what gives a two-dimensional screen meaning; otherwise "high-score, low-valuation" would resolve to an empty set or the whole set.
This cut's definitions: a base pool of 760 reports (matching the count in the Baillie leaderboard footer); an inclusion line of total ≥50 (the top 10.4% of the library, 79 reports) and vp ≤0.33 (the conservative third of the valuation band). Snapshot prices are 2026-06-12 closing data, each name in its own currency, and every future return comparison runs in the local currency.
The opportunity quadrant: high-score × low-valuation, 15 names
| Company | Total | vp | Q9+Q10 | Report rating | Snapshot price | Fair buy price |
|---|---|---|---|---|---|---|
| CATL 300750.SHE | 59 | 0.25 | 8 | Cautious Buy | 388.50 CNY | 400 |
| Tencent 0700.HK | 56 | 0.23 | 6 | Cautious Buy | 465.60 HKD | 430 |
| Sea SE.US | 55 | 0.27 | 6 | Watch | 82.44 USD | 78 |
| Microsoft MSFT.US | 55 | 0.32 | 6 | Watch | 397.36 USD | — |
| Kaspi.kz KSPI.US | 53 | 0.17 | 6 | Watch | 78.58 USD | — |
| Workday WDAY.US | 52 | 0.26 | 7 | Watch | 137.47 USD | — |
| Corpay CPAY.US | 52 | 0.32 | 6 | Watch | 348.99 USD | 290 |
| ResMed RMD.US | 51 | 0.21 | 6 | Watch | 193.57 USD | 190 |
| Luckin LKNCY.US | 51 | 0.29 | 9 | Hold | 30.21 USD | 20 |
| Intuit INTU.US | 50 | 0.05 | 7 | Watch | 284.22 USD | — |
| Airbnb ABNB.US | 50 | 0.22 | 5 | Watch | 129.10 USD | — |
| Uber UBER.US | 50 | 0.23 | 6 | Watch | 68.61 USD | — |
| Salesforce CRM.US | 50 | 0.24 | 6 | Watch | 170.92 USD | — |
| Nintendo 7974.TSE ☆ | 50 | 0.26 | 7 | Buy | (snapshot basis) | 7500 JPY |
| Hexagon HEXA-B.ST | 50 | 0.27 | 6 | Watch | 80.66 SEK | 80 |
☆ = no live price that day (filtered out by the data-consistency gate); vp is computed off the report's snapshot price, to be listed separately or dropped at post-mortem. Q9+Q10 is the combined score, out of 20, of two of the ten questions: "what the price implies" and "why the market hasn't noticed." It gauges how heavily a name turns on pricing perception.
"High-score, low-valuation" answers only "good quality, low position." It still can't tell you "should I buy it now." So we add one more gate.
The entry gate: the only hit is CATL
A report gives a "fair buy price" (the top of the buy range). Trading below that line, with a total in the high-score band, leaves exactly one name this cut:
- CATL: 388.50 ≤ 400. The report rates it Cautious Buy, with the ideal entry lower still (somewhere below 340).
Three names sit a breath short on total (<50) yet already stand under the gate, as a second tier to watch: Gartner (154.91 vs gate 155), Lululemon (118.93 vs 120), and FIS (38.97 vs 45).
Why they look cheap: a read on the top three
Cheap comes in two flavors: cheap with a grievance, and cheap for good reason. Telling them apart rests on the body of the answers to Q9 (what the current price implies) and Q10 (why the market hasn't noticed). Here is the condensed read on the top three names:
- CATL (score 59, the only one through the gate): the quality is already priced to a fair degree, with 28 brokers at strong buy and the P/E repaired from 14x to 24x. The genuine perception gap sits in two "can't-see-far" spots. Under the push against destructive price wars, storage winning bids have climbed off the 0.442 yuan/Wh low back to 0.38–0.43, and lithium carbonate has returned to 194,000 yuan; the lift in the earnings center is coming through, and the market doubts it lasts. The depth of the geopolitical policy ceiling has not entered the price either. The report puts the implied neutral return at roughly 10–15% annualized, positioning it as steady compounding plus an anti-price-war option.
- Tencent (score 56, 8% from the gate): half the discount is contempt. The memory of regulatory tightening sits on top of major holder Prosus's steady selling, putting the stock at a 14.4x P/E against Meta's 21x. The other half is myopia. AI already carries around 30% of ad placements (AIM+), yet that commercial payoff has barely entered the valuation. The report puts today's 14x as pricing in almost no AI premium; a re-rating to 18–20x alone would deliver a thirty to forty percent return, with the turn coming as AI's efficiency gains become visible margin over the next 4–6 quarters.
- Sea (score 55, 6% from the gate): the market already sees part of it, with 30 analysts at Strong Buy and target prices 66% above the current level. The discount is really "yet to be proven": first-quarter EPS of 0.67 came in below the 0.77 consensus, on top of the complexity of running e-commerce, finance, and gaming as three engines at once. The second- and third-quarter prints are the catalysts for a re-rating.
Two traps that are easy to step in
Trap one: breaking through the bear range ≠ a bigger opportunity. 8 reports have vp below 0 (price already under the report's most bearish scenario), and three payment processors cluster among them: Fiserv (53.28, gate 60), PayPal (40.70), and Global Payments (62.45, gate 78). In the face of an industry-wide de-rating, this cut's call is "cheap for good reason" first: the four names already through the gate but with a mediocre total get read, by discipline, as value traps first, then checked against the answers at post-mortem. The highest-scoring of these 8, and the closest to the inclusion line, is Adobe (score 48, vp −0.09, snapshot price 233.38), tracked on its own.
Trap two: high-score, high-valuation, where acting now is chasing. The names at the top on total but already high on vp: NVIDIA (score 66, vp 0.86), Palantir (score 61, vp 1.22), Meta (score 58, 0.68), Broadcom (score 57, 0.68; report rating Avoid, snapshot price 372.10 against a fair buy price of 173), and TSMC (score 56, 0.81). Among them Palantir, Innolight, Oracle, and Eli Lilly already carry a vp above 1, with the price punched through the top of the report's bull range. This basket is the methodology's mirror test: if it ends up sharply outrunning the low-valuation basket, that says the "position" yardstick punishes too hard in a momentum market, and the axis itself needs fixing.
A sample where the gate fights perception: Luckin's Q9+Q10 hits 9 (the highest in the library), the densest pricing-perception read of the lot, yet at 30.21 the price is 1.5x the fair buy price of 20, so this cut turns it down on the gate. Another edge case is Coupang (total 46, vp 0.17, Q9+Q10=8, snapshot price 15.12 USD), 4 points short of the inclusion line. These two names are set aside to test, at post-mortem, whether "the gate is stricter than perception" is the right discipline.
The scorecard: how we plan to mark our own answers
The post-mortem is set to run from the fourth quarter of 2026 (ideal window 6–12 months), answering the six questions below one by one and appending the conclusions to the "Post-mortem record" at the foot:
- The primary verdict on whether the methodology worked: the local-currency return of the 15-name equal-weight opportunity basket against the high-score high-valuation basket. The low-valuation basket wins ⇔ the methodology holds.
- The quality of the entry signal: CATL's absolute return from 388.50; whether it tagged the report's ideal entry of 340 along the way, and whether it climbed back above 400.
- The opportunity cost of "waiting for the gate": whether the four names within ≤8% (Tencent 430 / Sea 78 / ResMed 190 / Hexagon 80) touched their gates, and how the ones that did performed from the gate price onward.
- Whether the trap calls were right: the three payment names (Fiserv 53.28 / PayPal 40.70 / Global Payments 62.45) keep bleeding or going sideways = "cheap for good reason" called correctly; a V-shaped reversal that beats the market = a wrong cut and a missed read, with a check back on whether we ignored evidence at the time. Adobe 233.38 reconciled separately.
- The predictive power of pricing perception: how Luckin (perception 9 yet turned down by the gate) and Coupang (4 points short of the line) trade afterward, testing whether "the gate is stricter than perception" holds.
- The mirror test: how the high-score high-valuation basket fares from snapshot price, especially the Avoid-rated Broadcom (372.10 against a gate of 173) and the four names with vp>1. If they outrun sharply, the "position" axis needs a rethink.
One internal discipline goes public alongside this: the board is alive. A given company may be taken over by a newer report, and scores may be re-marked as calibration shifts, so at post-mortem we will record "score at the time vs score then" side by side, and the score drift is itself a review item for the calibration system.
Post-mortem record
(To be appended, from the fourth quarter of 2026.)
Disclaimer: this article is a quantitative read on signals from our in-library reports; every rating and buy price is drawn from an individual report and carries the timeliness of when that report was written, and snapshot prices are 2026-06-12 data. The above is for research reference only and does not constitute investment advice; act on it at your own risk.