#2 — JPY exposure bug fix

2026-04-01

Verdict: positive · ✅ critical production fix

A currency-exposure calculation that never converted JPY notionals to USD overstated exposure 100-160x, blocking all 12 pairs from trading; the fix raised open positions from 13 to 32.

Hypothesis — what it tests

Whether the system-wide trade blockage was caused by a currency-exposure math bug rather than by the strategies or the risk limits themselves.

Description

All trading in the portfolio manager was silently blocked: JPY-quoted pairs computed notional as units x entry_price without converting yen to dollars, so with USD/JPY near 150-160 the JPY exposure read 29,000%+ instead of the real ~190%, tripping the currency-exposure cap for every pair. The fix added a USD/JPY price tracker (5-minute refresh) and divided JPY notionals by it, and raised the per-currency risk limit from 40% to 200%. Positions immediately rose from 13 to 32 and blocked pairs fell from 14 to 9. The audit table files it under 'Measurement — the apparatus itself can lie': a units-arithmetic bug in the risk layer had the same silencing power as a bad signal.

Key result

JPY exposure misread 29,000%+ vs ~190% real (notionals 100-160x overstated); positions 13 → 32 (+146%); blocked pairs 14 → 9; risk limit 40% → 200%

Algorithms

currency-exposure aggregation (risk arithmetic)

Code

services/portfolio_mgr/main.py

Interactive version (search, filters, figures): the experiment explorer.

← #1 Project genesis + architecture #3 ASI-MC V3 efficiency-ratio gate →

This page is one experiment from the audited record behind the book The Edge That Wasn't — get the book · about the project.