Directional accuracy measures one thing: when the model says "up," does price go up? It ignores magnitude and focuses on the coin-flip that matters most to a trader. A model with 50% directional accuracy has no edge; consistently beating that threshold, after costs, is where profit comes from.
Accuracy must be measured honestly to mean anything. That means out-of-sample testing (no peeking at the future), accounting for the base rate of up-days, and tracking results over enough trades that the number is not luck. A 78% figure on ten trades is noise; on thousands it is a track record.
Directional accuracy is not the whole story — a model can be right often but lose money if the wins are small and the losses large. Pair it with payoff ratio and position sizing to judge real-world performance.
Sintinel publishes signal performance over time so accuracy is an accountable, evolving metric rather than a marketing claim frozen on a landing page.