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

Onyx Patterns

Pyramid analysis

Scaling INTO a position with additional entries — pyramiding — is one of the sharpest dividing lines in trading: the textbook 'add to winners' is widely edge-positive, the textbook 'add to losers' is the single most common account-killer. The detector splits your behaviour cleanly between the two without you having to confess.

What it is

Pyramiding is opening additional same-direction entries on the same instrument while a previous entry is still open. Two textbook variants:

  • Adding to winners — the second entry is at a price MORE favourable than the first (buy higher / sell lower). The market has confirmed the original thesis; capital follows confirmation. Documented as edge-positive when the underlying setup is intact.
  • Adding to losers — the second entry is at a price LESS favourable than the first (buy lower into red / sell higher into red). Classic averaging-down. Documented as the single most common path to ruin: the trader fights market direction with capital.

The detector groups overlapping same-symbol same-direction trades into pyramid clusters, classifies each by the price progression of its entries (favourable to direction = winner, unfavourable = loser), then compares the cluster aggregates to solo (single-entry) trades on the same instruments.

A cluster classifies cleanly only when ALL additional entries point one way. Mixed clusters (some entries above, some below the first) drop into a 'mixed' bucket and don't drive the verdict.

Formula
Approach (textbook): for each pair of trades on the same symbol + same direction whose time windows overlap, group them into one pyramid cluster (extend the chain whenever a new trade opens before the running latest close-time). For each cluster, compute the per-entry price delta against the first entry of the cluster. Direction-relative: for a long, more-favourable = price rose; for a short, more-favourable = price fell. A cluster of `n ≥ 2` entries where every additional entry is favourable is add-to-winner; every additional entry unfavourable is add-to-loser; otherwise mixed.
 
TradeOnyx-internal: the minimum number of clusters before emission, the minimum number of add-to-loser clusters before flagging, and the price-noise dead-zone band are calibrated empirically and not published.
Example

Across the analysis window: 4 pyramid clusters detected on EURUSD. Three of them are add-to-loser (each opened a long, then opened a second long at a lower price), aggregate net-P&L of those three: -€135 average per cluster. Solo EURUSD trades (no pyramid): 22 trades, +€18 average net-P&L.

ResultCard emits with verdict **Edge drag**: every time the trader pyramids by adding to losers, they give back roughly seven solo trades' worth of profit. The actionable read: a hard rule — never add to a position that hasn't already shown a profit.
How to read it

The two verdicts:

  • Edge drag (adding to losers). The pattern that closes accounts. Reasoning the trader gives themselves: 'the trade is even better at this price.' What the math says: the original entry's stop got triggered emotionally, not technically; the second entry is a re-bet on the same broken thesis with less room to recover.
  • Edge boost (adding to winners). The pattern that builds accounts. Reasoning the trader gives themselves: 'the trade is working, I'll add to it.' What the math says: capital is concentrated where price action confirms the thesis, with the original entry's profit cushioning the risk on the new entry.

The Onyx-Engine reports whichever pattern is dominant in your data; mixed-or-neutral pyramid behaviour is not surfaced (the card stays silent).

The fix for edge drag is mechanical, not motivational. A simple rule: 'I add to a position only if the original entry already shows a profit.' That single sentence eliminates the entire add-to-loser bucket without removing any winner-pyramid behaviour.

Tier: Pro. Pairs naturally with the Risk-of-Ruin Monte Carlo — pyramid-into-loser behaviour is the single largest distortion of the per-trade-risk assumption that drives the Monte Carlo. If both fire on the same account, fix the pyramid behaviour before re-reading the ruin number.

Where TradeOnyx uses it

How to read the card:

1. Hero (left) — verdict word + the dominant pyramid kind. Single-glance read. 2. Four MicroStats — add-to-winner cluster count + avg P&L (green if positive), add-to-loser cluster count + avg P&L (typically red), pyramid-aggregate avg, solo-baseline count + avg. 3. Hint line — the rule that addresses the dominant verdict.

Re-look frequency: weekly. Pyramid behaviour changes slowly because it's a position-management habit; daily noise isn't useful here.

Tier: Pro.

Related reading