Simulations

Play with the actual mechanism

6 interactive simulations of the mechanisms behind the most-argued-about ideas in technical analysis. Each one is anchored to a concept in our research corpus, states that concept's support level exactly as the literature does — including the one where the honest answer is no measurable edge — and lists its verified citations. Drag the parameters, watch the signals fire, and switch the underlying effect off to see what the mechanism does when there is nothing there.

MIXED

The 1992 moving-average rule

Brock, Lakonishok & LeBaron's own statistic — mean return on buy days minus mean return on sell days — rebuilt on synthetic data you control, next to the two things that later erased most of the result: transaction costs, and how many variants you tried first.

You control: short and long lookbacks, the 1% band filter, variable- vs fixed-length signal handling, one-way cost in basis points, and how much real trend the synthetic market has at all.

SUPPORTED

Winners minus losers, J × K

The best-supported idea technical trading points at, run as its original experiment: rank a synthetic cross-section on its trailing J-month return, hold K months, and read the whole J × K grid at once — including the rare, violent drawdowns that come with it.

You control: formation and holding periods, portfolio width, the skip month, cost, how much cross-sectional predictability exists at all, and an injected momentum-crash episode.

PARTLY SUPPORTED

Why RSI 30/70 is arbitrary

Short-horizon reversal is a documented return regularity — the statistical soil oscillators grow in. The specific 30/70 levels are Wilder's round numbers. Sweep the whole entry × exit threshold surface and watch where the best cell actually lands.

You control: RSI lookback, both thresholds, the time stop, one-way cost, and the strength of the short-horizon reversal the synthetic market contains.

NO SUPPORT

Bootstrap a candlestick pattern

The cleanest negative result in the indicator literature, as a test you run yourself. Detect a real pattern on synthetic bars, measure the forward return, then compare it against the same number of randomly chosen entry dates. The control that makes it honest: inject a real edge and the same test lights up.

You control: the pattern, its geometric tolerances, holding period, bootstrap replicates, and an injected post-pattern drift that proves the test has power.

PARTLY SUPPORTED

Where the liquidity actually rests

Take-profit orders cluster ON round numbers and stop-losses just BEYOND them — a measured order-flow asymmetry that produces both reversals at the level and faster moves after a break. Build the book, run price into it, and compare round levels against non-round controls.

You control: take-profit and stop-loss mass, the stop offset beyond the level, cascade feedback, volatility, and a clustering dial that removes the round-number effect entirely.

MIXED

The best of N rules

Search a universe of technical rules on a random walk with zero predictability, find the winner, and watch it look excellent. Then bootstrap the distribution of the BEST rule rather than of one rule, and read the two p-values side by side.

You control: universe size, sample length, bootstrap replicates, cost, and an injected true edge that makes both p-values agree again.

How to read these

These are simulations on randomly generated synthetic data — not market data, not live signals, and not a record of anyone's trading. Nothing in them is an account balance or a projection of what any strategy would earn. Their purpose is the opposite of a backtest screenshot: every one carries a control that switches the underlying effect off, so you can verify the mechanism reports nothing when there is nothing to find. Support levels and findings are quoted from 22 verified sources catalogued on our indicator research and smart money research pages — we cite what each paper actually found and avoid overstating it. Educational information only — not investment advice. Hypothetical backtested results; past performance does not guarantee future results. Trading involves risk of loss.