I spent 7 months building a trading system that lost money. Here's all the data.
August 2026 · by the developer of Dojix
Before Dojix was a product, it was a survival tool. I was running five automated forex algorithms on a VPS — institutional liquidity-sweep strategies on MetaTrader 5 — and I needed to know, honestly, whether they worked. So I built a 13-module analytics engine to measure every trade: maximum favorable excursion, R-multiple capture, session and hour edge decomposition, exit-policy counterfactuals.
The answer was no. They didn't work. And the full post-mortem — every number, every bug, every chart, the raw trade data — is now public.
The short version: a broken exit system was manufacturing a 74% win rate out of entries that won 32% of the time when measured honestly. Fixing the bug didn't reveal a hidden edge — it revealed there was never an edge. The pre-registered performance gate I'd set months earlier is what let me quit honestly instead of revising forever.
Read the full post-mortem on GitHub — the article, the analytics architecture, a complete EA source file, the revision logs, and the raw verification trades as CSV. Almost nobody publishes losing results with real numbers. This is what that looks like.
Read it on GitHub →How Dojix was actually used in this project
Dojix wasn't a bystander here — it was part of the decision loop. Every trade the portfolio produced flowed from the VPS into Dojix, and two of its analyses fed directly into the revision decisions you can read in the post-mortem:
- Kill Zone Optimizer — decomposes your P&L by hour and session to find the windows where you make money and the windows that quietly bleed. On this portfolio it flagged hour 10 (London) as an avoid window and hour 17 (NY) as one of the few genuine edge windows — calls that held up in the final verification data, where the three worst hours accounted for more than the entire net loss.
- What to Cut — surfaces the specific conditions (hours, setups, directions, instruments) with persistent negative expectancy. Its exports were on the table for the June evaluation that cut the two worst algorithms from the portfolio.
That's the honest version of a product claim: the app's job is to find where your edge isn't, and on this portfolio it kept being right.
What else it analyzes
The same engine runs 66 analytics across your imported history — an Edge Report that grades every setup you trade, timing analytics (hourly P&L, session heatmaps, day-of-week), honest R-multiple and MFE/MAE capture (the exact analysis that exposed this project's exit bug), behavioral analytics like emotional P&L and checklist compliance, drawdown recovery, and Monte Carlo risk-of-ruin projections. Imports from 9 platforms, everything local on your Mac.
If the charts in the post-mortem look like the truth you wish you had about your own trading: that's the app.
Dojix is built by Aaxis Holdings Group.