What the market actually does after every sweep — independent of today's trading strategy. Reversal/continuation are measured relative to the engine's called (predicted-reversal) direction — positive displacement means the call was right, negative means the pre-sweep trend continued instead. No trade outcome is assumed.
All distance metrics (MFE, MAE, displacement) are shown in pips, normalized by each symbol's pip size — raw price-unit values are still available for audit on hover in the By Symbol tab.
Research explorer
Every table, chart, and hypothesis test on this page updates to reflect the active filters. Filters compose (e.g. Symbol=GBP/USD + Session=london narrows to exactly that intersection); leave a filter on "All" to keep that dimension unrestricted.
Showing 0 sweeps matching current filters
Symbol health score
Composite 0-100 score (Data Volume + Directional Dominance + Statistical Significance + Outcome Stability, 25 points each) saying how much weight to put on a symbol's research numbers at all. Not a trading signal — a symbol can score high and have a bad edge, or vice versa.
scoped to filters above
Daily stability trend
Edge Score (continuation% − reversal%) per symbol, per day. A flat line means a consistent day-to-day edge; a swinging line means it's still settling or genuinely regime-dependent. Will read as noisy with only a few days of history — that's expected, not a finding yet.
Symbol behavioral fingerprint
Radar comparison of each symbol's continuation bias, adverse dominance, entropy, upside probability, whipsaw%, and mean-reversion% (all normalized 0-100). Different shapes are different market "personalities," not different quality — descriptive, not a ranking.
Relationship views
One point per symbol. Exploratory relationship views, not fitted models — look for clusters and outliers, not a line to draw conclusions from with this few points.
Continuation % vs. Adverse Dominance
Entropy vs. Economic Edge
Hypothesis tests
H₀: this variable has no relationship to outcome at all (chi-square test of independence against the full outcome distribution, within the current filtered slice). p < 0.05 conventionally means "probably a real effect," not proof; a non-significant result on a small filtered slice may just mean insufficient data, not "no effect."
scoped to filters above
Pullback quality test
Does a smaller initial pullback (relative to the eventual move) predict a bigger, faster continuation? Splits continuation-outcome sweeps into low/high pullback-ratio halves (median split) and tests via Mann-Whitney U whether their final distance and time-to-peak distributions actually differ.
scoped to filters above
Displacement over time
Average signed displacement (relative to the called direction) at each horizon, per symbol. A line rising and staying positive means that symbol's sweeps tend to confirm the called direction; a line falling and staying negative means the pre-sweep trend typically continues instead.
Overall
0 sweeps with a completed 5-minute outcome window — the top-line view before any breakdown below.
scoped to filters above
Breakdowns
Same filtered sweep universe as Overall above, sliced four different ways. % columns show a 95% confidence interval underneath (Wilson score) — a rate from a small n carries much more uncertainty than the headline number suggests. On the Symbol tab: Edge Score = continuation% − reversal% (frequency only); Adverse Dominance = avg MAE ÷ avg MFE (>1 = the adverse side typically moves farther); Upside/Downside % = absolute market direction, independent of what was called.
scoped to filters above
UTC hour buckets: asian 00–07, london 07–12, overlap 12–16, ny 16–21, late 21–24. Approximate, not exchange-precise.
Conditional distributions, by symbol
For sweeps that ended up continuation or reversal specifically: how far did the move actually go (median/p75/p90 of the excursion in that direction), how long did it take to reach its peak, and how big was the opposite-direction pullback before that peak — "is there enough movement to be worth anything," which percentages alone don't answer.