Three names two sigma below, one worth a look

Three names cleared the liquidity and gap filters this morning and printed a z-score below −2 against their own 50-day mean. Working through them in the order the screen returned them, because the skips are more instructive than the keep.

The two I’m skipping

Name one. Two sigma below, and the 50-day mean has rolled over hard in the same window. This is failure mode number one in its natural habitat: the gap is closing, but from the top. When the anchor is falling as fast as the price, you are not buying a stretched rubber band. You are buying a slow decline with a technical justification stapled to it. Skip.

Name two. Clean anchor, decent balance sheet, and it fails on correlation. It is the third name from the same industry group to clear the screen in ten days, and I already have exposure there. Six oversold names in one sector is one trade. If I want more of that trade I should size the existing position up deliberately, not accumulate it accidentally through a screen that keeps handing me the same bet with different tickers on it.

The one I’m watching

The third has the profile the screen is actually built to find: the decline was a grind rather than a gap, the 50-day mean is roughly flat over the last quarter, forward estimates have not been cut, and free cash flow has been positive in three of the last four quarters.

That is not a reason to buy it this morning. It is a reason to do the work:

• Read the last call, specifically the guidance language rather than the numbers.
• Find out what the market thinks it now knows that it did not know six weeks ago, and decide whether that is a change in the anchor or a change in mood.
• Write the exit before the entry — the mean-based invalidation, the time stop set against the estimated half-life, and the stop distance that sets the size.

If I take it, the entry and the reasoning go up here first, and the review goes up when it closes, whichever way it goes.

The general point

A screen that returns three names and produces one candidate is working. A screen that returns thirty and produces thirty is a bad-news detector with a threshold set to always produce work.

The tickers listed on this post are the ones the screen surfaced, not recommendations, and the figures described are illustrative of the process rather than a published record of prices. Do your own work.