Structural context · Ratings
Analyst ratings, with the receipts attached
Every platform shows you analyst upgrades and downgrades. Almost none tell you whether the firm issuing them has historically been right. We backtested 5,600+ graded calls from 2012 to 2026 against realized returns, and we publish the per-firm record next to every rating. The headline: the average 5-day hit rate is 56.5% — barely above a coin flip.
The measured record, not the reputation
| Firm | 5-day hit rate | Graded calls |
|---|---|---|
| Susquehanna | 68.8% | 32 |
| Telsey Advisory Group | 67.7% | 31 |
| Jefferies | 67.2% | 58 |
| BMO Capital | 66.7% | 39 |
| — average, all qualifying firms — | 56.5% | 2,508 calls |
| DA Davidson (lowest qualifying) | 42.3% | 52 |
Real backtest output as of August 2026, firms with ≥30 graded calls at the 5-day horizon. Scores are recomputed as new calls mature; the dashboard always shows the current table. A historical hit rate describes past calls — it is not a forecast of the next one.
What you get
Market-wide ratings feed
The latest upgrades, downgrades, initiations and price-target changes across the market, not just one watchlist.
Per-firm accuracy score
Each firm’s measured hit rate at 5 and 21 trading days, with the sample size shown — small samples are visible, not hidden.
Ratings in your trade context
Analyst actions on a ticker appear in the Context Journal as part of the catalyst layer frozen with your trades on that name.
Both horizons, both directions
Upgrades and downgrades are graded symmetrically; a downgrade that precedes a decline counts as a hit for the firm.
HOW THIS IS CALCULATED
Direction is taken from the rating action itself. Return is measured from the close of the publication day to 5 (and 21) trading sessions later; a matching sign is a hit. Calls too recent to have matured are not graded. Firms enter the accuracy table at 30 graded calls. Data source: Financial Modeling Prep grades and price-target feeds; prices from market close. None of this constitutes a rating of our own — we measure other people's calls, we do not make them.
Frequently asked questions
How do you measure whether an analyst call was "right"?
We take each rating action with a clear direction (upgrades, downgrades, initiations with a directional grade), then measure the underlying stock’s return from the close of the call’s publication day to N trading sessions later (5 and 21 days). A call counts as a hit if the realized sign matches the call’s direction. Only calls old enough to have matured are graded — recent calls are excluded until their window completes.
What did the backtest actually find?
Across 5,600+ graded calls spanning 2012–2026, the average 5-day hit rate for firms with at least 30 graded calls is 56.5% — above a coin flip, but far below what the confidence of a typical price target implies. Individual firms in our sample range from roughly 42% to 69% at 5 days. The spread between firms is the useful information, not the average.
Why publish this if it makes analyst ratings look weak?
Because it is true, and because context beats authority. Our own directional research reached the same conclusion about short-horizon prediction in general, so we treat measured accuracy as the honest way to present ratings: every rating in our feed can be read next to the issuing firm’s track record instead of on faith.
Does a high historical hit rate mean I should follow that firm?
No. A historical hit rate is a description of past calls, not a prediction about the next one, and samples per firm are small enough that ranks move. The tracker exists so you can weigh ratings with evidence rather than reputation — it does not turn any rating into a recommendation, and neither do we.
Where does the ratings data come from?
Rating actions come from Financial Modeling Prep’s grades and price-target feeds; returns are computed against market close prices. The feed shows the most recent actions market-wide, each linked to the firm’s measured record. Accuracy scores are recomputed as new calls mature.
Read the next rating with its track record attached
The full per-firm accuracy table and the market-wide feed live in the dashboard.
Not financial advice. Trading involves risk. TradeAI News provides structural market context, not directional recommendations.