Five Signals, Eight Dealbreakers: How I Trade Small-Cap Tech

I spend my days in corporate finance and my evenings researching small-cap technology stocks, and the single most useful decision I have made as a trader was writing down what counts as evidence. Here is the whole system in one sentence: I buy small-cap tech companies ($300M to $5B) showing at least one of five statistically backed signals, I refuse any name that trips one of eight structural dealbreakers, and I take profits mechanically at +100%. No chart patterns. No trendlines. If a signal cannot point to peer-reviewed research or filings, it is not in the system.

This post is the reference for how I trade. The small-cap trading framework below is the whole of it. The individual rules get their own write-ups, and my live results are public on the portfolio tracker, wins and losses both.

Why the rules had to come first

Small-cap trading content is mostly astrology with candlesticks. I have a finance degree and a master’s in it, and I still nearly got pulled into drawing lines on charts, because lines on charts feel like analysis. They produce a feeling of insight with no testable claim underneath.

So I set a bar: a signal gets into the system only if there is published research or hard filing data behind it. That bar turns out to be brutally selective. Most of what retail traders stare at all day does not clear it. What survives is a short list, and a short list is a system you can actually follow on a Tuesday night after work.

The five signals

1. Upward earnings estimate revisions. Analysts raising numbers in the last 30 days. Not ratings, revisions. The post-revision drift is one of the oldest documented anomalies in equity research: estimates move, prices follow slower than they should.

2. Relative volume at 2x or more. Volume running at least twice the 20-day average says something changed and the market noticed. I use it as a confirmation that a catalyst is real, never as a reason on its own.

3. Short interest above 15% with days-to-cover above 5. Crowded shorts in an illiquid name are stored fuel. Both legs matter. High short interest with one day to cover unwinds quietly.

4. Insider buying clusters. Three or more insiders buying in the open market within 90 days. Lakonishok and Lee (2001) and Jeng, Metrick and Zeckhauser (2003) both found insider purchases predict returns, and clusters carry the strongest signal. The fine print does the work here: open-market purchases count, option exercises and compensation grants do not. A CEO accepting shares he was going to get anyway has told you nothing.

5. Price momentum, both legs. Top 20% of 6-month returns with positive 1-month acceleration, the factor documented by Jegadeesh and Titman in 1993 and confirmed in dozens of markets since. A strong 6-month return that is currently rolling over fails the test. That is not a momentum stock, it is a momentum stock that stopped.

What never counts

Chart-pattern technical analysis is out entirely: trendlines, Fibonacci retracements, Elliott Wave, head and shoulders, support and resistance. Not because I find it aesthetically annoying (I do) but because I could not find credible evidence it predicts anything after costs.

Two subtler impostors are worth naming, because they wear quantitative costumes.

Analyst ratings are not revisions. Last July my screener served up a company at “Strong Buy” that my own notes had already hard-eliminated for going-concern language. If a metric cannot tell that company apart from a real candidate, it is not a filter. Ratings are opinions with sticky careers behind them. Revisions are numbers that just changed.

A price hitting your alert is not a signal. Price triggers tell you to look. They are never, on their own, a reason to buy.

The eight dealbreakers

Signals argue for a stock. Dealbreakers overrule them, automatically, no matter how good the revenue print looks:

The order of operations matters more than I expected. I once built a full promotion case for a compound semiconductor name on a monster quarter and violent estimate revisions, then pulled the Form 4 data last. Fifteen-plus insider sales in a month, including a CEO who exercised options around $5 and sold about $22M of stock at $112. Thirty minutes of insider data dismantled an afternoon of work. Now the dealbreaker check runs first, always. A filter you run after falling in love is not a filter, it is a formality.

The hardest part is applying this to names you want to own. I have passed on companies with 90% revenue growth and nine-figure cash piles because insiders were selling in clusters, and one of them nearly doubled after I passed. The pass was still correct. Waiving a rule for a winner prices the rule at zero.

Position sizing and the cash rule

Standard positions are 5 to 8% of the portfolio, 10% for rare high-conviction setups, 5% max for anything speculative or under $300M. Deployed capital targets 30 to 40%, which means most of the account sits in cash most of the time.

That is by design. Cash is a position. When nothing clears the bar, holding 60% cash is the system working, not the system failing. An empty pipeline is a screening problem. It is never a reason to lower the bar.

Exits: the free ride

Exits are where most retail systems quietly don’t exist, so mine is mechanical: sell 50% of the original position at +100%, which takes the entire principal off the table, sell another 25% at +200%, and let the final quarter run. I wrote up the full method and the trade that proved it, in which a thesis of mine broke completely and the round trip still closed up 38%.

What it looks like in practice

The framework is mostly a machine for saying no. A 43-name medical device screen last spring eliminated 88% of the list on structural grounds before a single deep dive. A typical month of screening produces zero entries, and the months that produce one tend to pay for the quiet ones.

Dilution deserves its own post because it is the dealbreaker that hides best; the written test I use, with the red flags and the cases that validated them, is here. And if you want to see the research style applied to a single messy name at full depth, my Datavault AI deep dive is the format: filings first, story second.

Questions I get about this

Do you really use no technical analysis at all? No chart patterns, ever. I do use price and volume data, since momentum and relative volume are computed from them. The line I draw is evidence: momentum has decades of peer-reviewed support, trendlines have none.

Why only electronic technology? Because edge comes from repetition in one beat. Semiconductors, SaaS, defense electronics and their supply chains rhyme with each other. Adding biotech would mean betting on trial data I have no advantage reading.

What broker data do you trust? The broker’s own quotes. Search results served me stale, pre-split prices on the same ticker four separate times, so every fact a thesis leans on gets re-verified against primary sources at every review.

Can I copy this system? You can copy the rules in an afternoon. The hard part is the part that cannot be copied: following them when a name you love fails a check.

Disclosure: I currently hold positions opened under this framework, including CLMB and DCTH, visible on the live tracker. This is research and ideation, not financial advice. All trades carry risk of loss.