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.
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.
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.
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.