Monday’s Analyst Calls: Buy Ratings Abound—But Show Me the Backtested Alpha
# Monday’s Analyst Calls: A Systematic Traders Guide to Ignoring Most of Them
The morning ritual is familiar: a digest of sell-side research calls on Block, Boise Cascade, Equifax, FirstEnergy, Lockheed Martin, Lumentum, NetApp, Roper, Spotify, and others, each arriving with a rating and a price target that some analyst can explain in four bullet points and a DCF tab nobody will ever audit. For anyone running a systematic book, the correct response is not to trade these calls. It’s to treat them as data — a sparse, noisy, and surprisingly exploitable one if you handle it with the same rigor you’d apply to any other alternative signal.
The academic literature on analyst revisions is well established and consistently replicated: revisions tend to cluster, and post-revision drift — price continuing in the direction of the change for days to weeks — is one of the more robust post-earnings announcement style anomalies. If you want to build something real in Python, this is where to start. `pandas` for the event study, `statsmodels` for regression on abnormal returns against control factors, and a clean `yfinance` or paid data feed for the price series. Aggregate the calls, code upgrades as +1, downgrades as -1, compute cumulative abnormal returns around the event date, and test whether the drift survives transaction costs. Spoiler: at retail-sized spreads on liquid names like Equifax or Lockheed, it usually doesn’t. On mid-caps, occasionally it might.
What actually moves on these mornings is rarely the rating. Price targets are theater — the median analyst target on a given name is a lagging function of where the stock already traded. The informative component is often the *change* in the target relative to the distribution of prior targets, and whether the analyst moved against recent price momentum. That’s a contrarian-conditioned signal, and like everything contrarian, it dies a little every time someone writes about it.
Names on this list illustrate the spectrum. Block and Spotify trade on narrative intensity — enormous sentiment surface area, meaning analyst calls function more as sentiment shocks than information events. FirstEnergy and Lockheed Martin are the opposite: low narrative beta, high macro and policy sensitivity, where an individual analyst’s opinion carries roughly the statistical weight of a coin flip. NetApp and Lumentum sit in the honest middle ground where revisions correlate reasonably well with the guidance cycle — that’s the segment where a revision-momentum factor has historically earned its keep, at least in backtests you should independently verify because mine are mine.
The risk framework matters more than the signal. Single-analyst events have fat-tailed noise; if you’re allocating capital on one call, position sizing is doing all the work and the model is doing none of it. Kelly constraints, volatility scaling, and a hard rule against trading anything inside your signal’s lookback window will save you from the most common failure mode: mistaking one day’s reaction for regime information.
Sell-side research is best understood as a slow, crowded, partially informative feature. It’s not alpha on its own — it never was — but as one input among a dozen in a systematic pipeline, it’s free raw material sitting in an RSS feed. Quantifying the crowd is the job. Reading the crowd is what everyone else does.