How This Works

Signals come in wildly different units — a percentage, a dollar amount, a count of insider trades — so before they can be combined, each one is converted onto a common scale from -2 to +2. A signal is scored one of two ways depending on what it measures:

• Relative to other stocks today (for signals specific to one company, like insider buying): the stock is ranked against every other stock in the universe on that signal, and that ranking becomes the score.
• Relative to its own history (for broader signals like TSMC revenue or capex guidance): today's reading is compared to that signal's own recent past to see whether it's unusually strong or weak right now.

In both cases, a positive score means the signal is bullish; negative means bearish; a score near zero means the signal is neutral or unremarkable right now.

Each stock's overall score is a weighted average of every signal that currently has data for it. Signals aren't all treated as equally important — each has a weight reflecting how much it should matter to the overall picture (see the caveat below on where these weights come from). A stock missing some signals is still scored fairly using whichever signals it does have.

• Composite score: typically falls somewhere between -2 and +2. Positive means bullish signals currently outweigh bearish ones for that stock; negative means the opposite; a score close to zero means the signals are mixed or not saying much right now.
• Rank and percentile: often more useful for quick comparison than the raw composite number — they show exactly where a stock stands relative to the rest of today's universe, from strongest (#1) to weakest.

New data flows in daily from each signal source, and every stock's score is fully recomputed each night to reflect the latest information. A stock's rank can move night to night as new signals come in — a snapshot of "today," not a static rating.

Tracks whether company insiders (executives, directors) are net buyers or net sellers of their own stock on the open market over the past 90 days. Sustained insider buying can signal confidence from the people with the best view of the business; heavy selling can be a caution sign (though routine selling for diversification is common and not always meaningful on its own).

Tracks whether short sellers are increasing or reducing their bets against a stock, relative to that stock's own recent history. Rapid short-covering (bears closing positions) often coincides with improving sentiment; a fast build-up in new short positions signals growing bearish conviction.

Tracks how much the big cloud/AI spenders (e.g. Microsoft, Meta, Google, Amazon) are raising or lowering their infrastructure spending guidance each quarter. This spending flows downstream into chips, equipment, and memory demand, typically with a lag of one to two quarters.

Tracks TSMC's own monthly revenue growth. As the leading chip manufacturer for most major AI hardware companies, TSMC's revenue trend is an early read on real demand across the whole chip industry.

Tracks how fast semiconductor equipment makers' sales are growing year-over-year. This reflects how much new manufacturing capacity the industry is currently building — a read on demand one step further upstream than chip sales themselves.

Tracks how fast memory chip (DRAM) contract prices are rising. Rising contract prices generally reflect suppliers gaining pricing power as demand outstrips supply — directly relevant to memory-chip makers' margins.