Screening for Compression Without Deceleration
Cheap stocks are usually cheap for a reason. The interesting exceptions are the ones where the price fell but the business didn’t. You can screen for exactly that by crossing two independent measurements: the valuation percentile on one axis, the growth trend on the other.
The two-axis idea
A low multiple by itself is ambiguous. It could mean the market is repricing a deteriorating business (correctly), or repricing an intact business (interestingly). One number cannot distinguish the two, because the multiple only knows about price and trailing fundamentals, not about the direction the fundamentals are moving.
So add the direction as a second, independent axis. Axis one, the valuation percentile: where today’s multiple ranks in the stock’s own five-year distribution (see the percentile guide). Axis two, the growth trend: is trailing revenue growth holding, accelerating, or rolling over? Crossing them yields four quadrants, and each has a distinct meaning:
| Growth intact or accelerating | Growth decelerating | |
|---|---|---|
| Low percentile | Compression without deceleration: the screen’s target. | Possible justified de-rating: the market is repricing a bent trajectory. |
| High percentile | Expensive-and-earning-it: the market is paying up for momentum in the business. | The dangerous quadrant: a rich multiple resting on a fading trajectory. |
The target quadrant (multiple near the bottom of its own range, growth not decelerating) is where price and business have visibly disconnected. That doesn’t make its members good investments. It makes them good research candidates, because the usual explanation for cheapness (the business broke) is absent from the trailing data, so whatever explanation remains is worth finding.
Running it in the screener
- Open the screener and pick the multiple to rank on: P/E for profitable cohorts, P/S or EV/Sales where earnings are thin (see when each multiple matters).
- Set the percentile ceiling to 25, so only stocks in the bottom quartile of their own five-year range pass. Tighten to 10 for the strictest cut.
- Set the growth filter to “not decelerating”: latest TTM revenue growth at or above its level two quarters ago. This is deliberately a loose test: it doesn’t demand acceleration, only the absence of a visible rollover.
- Sort by percentile rank ascending, then scan the per-multiple columns: names ranked low on every multiple are cleaner candidates than ones cheap only on the most flattering metric.
- Narrow by sector or market cap if you want a specific hunting ground, or start from the prebuilt compression-without-deceleration preset, which encodes steps 2–4 in one click.
Then do the part no screen can do: open each candidate’s band chart and check the shape of the disconnect: when it opened, how fast, and what the fundamentals overlay was doing at the time.
Why forced selling produces exactly this signature
Most price moves carry information: someone learned something and traded on it. But some of the largest moves carry none: a leveraged holder hits a margin call, a fund faces redemptions, a mandate change forces liquidation. The seller’s constraint, not the company’s prospects, sets the pace of selling, and everything in the portfolio gets sold regardless of merit.
On a valuation platform, that mechanical selling has a fingerprint: multiples across the affected names compress sharply and near-simultaneously, while the fundamentals, which report on a quarterly clock and don’t care who is selling, are unchanged. Whole cohorts drop into the target quadrant at once. In late July 2026, the rapid unwinding of a large fund concentrated in AI-infrastructure names produced exactly this pattern: valuation percentiles across semiconductor, memory, and data-center stocks fell to bottom-decile readings within days, on no new company-level information, while trailing growth trends were unchanged. The screen is, in part, a machine for noticing episodes like that while they are still on the tape.
The honest caveat: a percentile that collapses on flow can recover on flow’s end, or keep falling if the episode changes what buyers will pay going forward. The screen identifies the disconnect; it doesn’t promise the disconnect closes in your favor.
The false-positive checklist
Every name that passes the screen deserves an attempt to kill it. The recurring escapes:
- The deceleration is announced but not yet printed. Trailing data lags. If management guided next quarter sharply lower, the market’s repricing is information-driven even though the TTM growth trend still looks intact. Check the latest report and guidance first; this is the single most common false positive.
- Acquired growth. A recent acquisition can hold reported TTM revenue growth up while the underlying business decelerates. Organic and reported growth are different animals.
- Cyclical peaks dressed as trends. In commodity-adjacent businesses (memory, shipping, energy), booming trailing growth can mark the top of a price cycle; the market compresses the multiple because it has seen this movie. “Not decelerating yet” is precisely how cycle peaks look.
- Thin or distorted percentile history. Short listing histories and bubble-stretched five-year windows both weaken the rank (the percentile guide covers these).
- Risks outside both axes. Litigation, regulation, customer concentration, financing needs: a real repricing reason that neither the multiple nor the growth trend can see. If a compression has no flow story and no fundamental story, assume there is a story you haven’t found yet.