Can a market fragility score pick better stocks? We measured it and built nothing
A retrospective study of historical market data, 1975 to 2026. tickerseer has published live analysis since May 2026; nothing below is a call we made at the time. It is not advice, and it is not a strategy we sell. It is the record of a feature we designed, measured, and did not build.
The short answer
No. We tried to make a market fragility score change which stocks our rankings favor, and it could not be calibrated on any data available to us. The gauge most people reach for, valuation measured against a market's own history, reads the 99.6th percentile through a calm stretch that produced no bear market, and the 99.7th percentile through the run-up to the dotcom crash. Those are the same reading. A rule that fires on the second fires on the first.
The 2008 run-up is the harder problem. It sat at the 50th percentile of that gauge, so the deepest bear in the sample never registered on the measure people would reach for to anticipate it.
What we were trying to build
Our per-stock score blends six factors with fixed weights: valuation against a company's own normal price-to-earnings ratio, forecast annual return, a quality grade, the PEG ratio, analyst accuracy, and a debt-quality adjustment. The plan was to move those weights with market conditions. In a fragile backdrop, lean toward valuation and balance-sheet strength and away from forecast optimism, on the grounds that forecasts are the input that breaks when conditions are fragile.
The reasoning holds up. Our earlier study of fourteen market shocks found that valuation, leverage and credit deterioration together marked the bears that did lasting damage. Turning that into a threshold is where it fell apart.
Why no threshold worked
The false alarm reads the same as the real thing. Median valuation percentile is 99.6 through 1996 to 1998, a stretch with no bear market, and 99.7 through January 1999 to March 2000. Set the cutoff above the 75th percentile and the calm window is active more often than the real one.
The deepest event does not register. The 2005 to 2008 valuation peak reached the 82.3rd percentile, against an all-history median of 82.4. No cutoff above the median can ever fire for 2008.
The gauge that did flag 2008 is not available to us. High-yield credit spreads led the 2007 equity peak by months. The free public history for that series runs three trailing years, capped by its licensor in April 2026. Reaching the 2007 run-up would need about twenty.
What we publish instead
The Market Fragility page carries the score as a reading on conditions. It says how much damage a shock could do, based on how similar conditions resolved before, and it says nothing about when. The wording does not change with the number: the same sentence appears at a low reading and a high one, with no band, tier or adjective attached.
Our per-stock rankings keep fixed weights. The one place those weights move is the Market Dashboard, where the reader sets them.
What this study cannot show
Three events is a small sample, and the study before this one said so first. A threshold tuned until 2000, 2008 and 2022 lined up would be fitted to three points. We did not tune one.
The measurements run on a monthly series computed back to 1975 from external data. That is a retrospective computation, not a record of readings we published at the time.
A different gauge might work. An investment-grade credit spread with forty years of history separates the 1996 to 1998 false alarm from the 1999 to 2000 real one, then reads a calm 2013 to 2017 stretch at the same level as the 2008 run-up. That moves the problem rather than solving it.
Frequently asked
Can a market fragility score tell you which stocks to buy?
Not on the evidence we have. A fragility score describes market conditions, and the gauge inside it that carries the most history cannot tell a false alarm from a real run-up: the 99.6th percentile through a calm 1996 to 1998, the 99.7th through the approach to the dotcom crash. We tested whether it could tilt our per-stock rankings, could not calibrate it, and did not ship it.
Does tickerseer weight its stock score by market conditions?
No. The six factors carry fixed weights on every page except the Market Dashboard, where the reader moves them. We designed a version that varied with a market fragility reading, measured it against fifty years of monthly data, and did not build it.
Why does the 2008 crash not show up in valuation data?
Because valuation was unremarkable going in. The 2005 to 2008 peak on our monthly series reached the 82.3rd percentile against an all-history median of 82.4. The gauge that did move early was high-yield credit, which widened months before the equity peak. Valuation flagged 2000 and 2022 and missed 2008.
What is the difference between measuring fragility and predicting a crash?
Fragility is a statement about severity: if a shock arrives now, how much damage could it do, judging by how similar conditions resolved before. Prediction is a statement about timing. Fifty years of evidence supports the first and not the second, and the same gauges that identify fragile conditions have stayed elevated for years with no bear market following.
Why publish a feature you decided not to ship?
Because the measurement is the useful part. A score that tilted stock rankings would look plausible on a page while having been fitted to three events. Publishing why this one failed is what lets a reader judge the ones we did ship.