← Research library
Research · 研究 · 40 · Foundations

Do stock signals actually work?

18 Jun 20266 min readFoundationsShishin Research

This article is educational. It is not personalised investment advice, not a claim that any signal service is profitable for any individual, and not a recommendation to buy or sell any security.

“Do stock signals work?” The honest answer is: some do, most don’t, and the difference is rarely what the marketing points at. A signal “works” only if it clears three bars at once — positive expectancy net of costs, a result that reproduces out of sample, and a human disciplined enough to actually follow it. Plenty of services clear the first in a backtest and fail the other two. Here is how to think about it.

What “work” has to mean

A signal that works is not one that was right last week, or that produced a screenshot of a 300% winner. It is one with a positive expectancy — average outcome per signal, across the whole population, after fees and slippage. That distinction matters because a service can show you its winners and quietly bury a longer tail of losers; the only number that survives that game is the average over every signal, winners and losers together.

Why most don’t

Two failure points, and most services hit at least one.

  • The signal has no real edge. It looks brilliant on history because the history was curated — tuned to its own backtest, tested only on the names that survived, or never checked out of sample. Those traps are the subject of why backtests lie and survivorship bias.
  • The human doesn’t follow it. Even a genuine edge fails the person who overrides it, skips the losing stretch, or sizes erratically. The classic finding that the most active individual investors underperform the most (Barber & Odean, 2000) is mostly a discipline problem, not a signal problem.

What a signal that works actually looks like

Strip away the marketing and the checklist is short. A signal worth following is one whose edge is positive after costs, reproducible (the backtest lands on the same number when re-run, and survives losing its best trades — see leave out the winners), survivorship-free, and transparent about its losers, not just its winners. And critically, one with a payoff shape and a cadence you can actually stick to — because the best signal you abandon mid-drawdown returns nothing.

Where Shishin stands

We are a research publisher, not an adviser, so “does it work” is a question we try to let you answer rather than assert. The full five-year backtest is survivorship-free and reproducible to the dollar, and every trade — winners and losers — is published, not a highlight reel; the figures live on the track record. What you do with the research is your decision. A signal only ever improves the odds; it never removes the risk.

Sources & further reading

  • Barber, B. M. & Odean, T. (2000). “Trading Is Hazardous to Your Wealth: The Common Stock Investment Performance of Individual Investors.” Journal of Finance, 55(2), 773–806.
  • Jegadeesh, N. & Titman, S. (1993). “Returns to Buying Winners and Selling Losers.” Journal of Finance, 48(1), 65–91.
Frequently asked

Do stock signals actually work?

Some do, most don't. A signal 'works' only if it has positive expectancy net of costs, reproduces out of sample, and is actually followed with discipline. Many services clear the first in a curated backtest and fail the other two.

Why do most stock signal services fail?

Two failure points: the signal has no real edge (curve-fit, survivorship-biased, or never tested out of sample), or the human doesn't follow it consistently. Even a genuine edge fails the person who overrides it or abandons it mid-drawdown.

How can you tell if a signal is worth following?

Check that the edge is positive after costs, reproducible, survivorship-free, and transparent about its losers — not just its winners — and that its cadence is one you can actually stick to.