Launching soon — TradeOnyx is still in preparation; public sign-up isn't open yet.

Onyx Patterns

Mistake trend

Knowing what your most expensive mistakes ARE is half the battle. Knowing whether they're getting better or worse over time is the other half — that's what tells you whether the journal rules you wrote three months ago are still working.

What it is

Mistake trend is a per-tag rolling-window comparison: how often does each mistake-tag appear in the last 30 days vs the preceding 60 days? Two directions:

  • Drift — current monthly rate is materially HIGHER than the 60-day baseline. Discipline is decaying.
  • Improvement — current monthly rate is materially LOWER than baseline. The rule is sticking.

Rates are normalised to per-month equivalents so the windows can have different lengths without distorting the read. Eligibility: a tag needs ≥ 3 baseline occurrences (else it's too rare to claim a trend) and a drift verdict additionally needs ≥ 3 recent occurrences (improvement doesn't — zero recent is the strongest possible improvement signal).

Formula
Approach (textbook): for each closed trade with a mistake-tag, decide whether its close-date falls in recent (today-30d → today) or baseline (today-90d → today-30d). Per tag count occurrences in each window; compute monthly-rate equivalents (count / window_days × 30); compute percent change ((recent − baseline) / baseline × 100). Filter by eligibility floors and threshold.
 
TradeOnyx-internal: the minimum baseline / recent occurrence floors and the change-percent threshold are calibrated empirically and not published.
Example

Across the analysis window: 'no stop' baseline 4 occurrences (rate 2.0/month), recent 8 occurrences (rate 8.0/month) → change +300%. 'panic exit' baseline 6 (rate 3.0/month), recent 1 (rate 1.0/month) → change -67%.

ResultCard emits with 'no stop' as the strongest drift (+300% — discipline decay) and 'panic exit' as the strongest improvement (-67% — rule sticking). Two reads at once, two journal-rule decisions: re-tighten the no-stop rule, keep the panic-exit rule running unchanged.
How to read it

The two directions:

  • Drift. Re-read the journal rule that USED to suppress this mistake. Either it stopped being honoured or the trigger has shifted to a new context the rule doesn't cover. Tighten the rule, not the count.
  • Improvement. Don't relax. Rules compound only as long as you keep policing them. The improvement signals tell you which rules ARE working — those are the ones to NOT change.

Both directions surface in the same card because both are actionable in different ways. Drift = problem to attack. Improvement = working pattern to protect.

Why monthly-rate equivalents. A trader with 50 trades in the last 30 days and 80 in the preceding 60 days isn't doubling down on discipline if the mistake count is 6 vs 8 — they're trading more, so the relative frequency is what matters. Normalising to per-month equivalents catches this.

Tier: Pro. Wave 6 (Mistake Analysis), pairs with TRA-235 (Top expensive — what's costing you most), TRA-237 (Pair co-occurrence — which mistakes cluster), and TRA-238 (Streak + recovery — sustained discipline).

Where TradeOnyx uses it

How to read the card: Hero shows the most-shifted tag with its percent change. Table breaks down the rest with direction labels. Re-look monthly — the rolling windows make weekly inspection too noisy.

Tier: Pro.

Related reading