HOW WE KNOW
WHAT WE CLAIM.
Working notes from building the platform, written from our own data rather than from theory. Three strands: how a strategy earns its place, what our own risk controls cost you, and why the desk is built to argue with your analysis instead of replacing it. Including the parts that make us look bad.
What our own risk guards actually cost you
Every entry the guards refuse is replayed at the stop the engine would have used, so protection can be priced. Three guards saved money last month. One cost money — and we changed it.
The desk · designA trading desk that argues with you, not for you
The agent on our charts is forbidden from telling you whether to trade. Why that restriction is the feature, how a chart becomes a conversation, and what we are not claiming.
MeasurementWhy a 94% win rate tells you almost nothing
Four of our seven strategies win over nine trades in ten, and every one of them loses about eight times as much on a loser as it makes on a winner. What to ask instead.
MeasurementWhat drawdown should you actually expect?
The recovery maths that makes drawdown the number that matters — and the month our own model priced a halving at one-in-450 while the real path drew down 51%.
BenchmarkWhat a realistic profit factor actually looks like
All seven of ours, from 1.14 to 2.82 — and the month we published four of them against the wrong strategies. What the number needs beside it before it means anything at all.
Post-mortemHow one line of code turned 1.98 into 0.77
Two strategies we shipped had been validated against results that could never have happened. The exact bugs, why out-of-sample testing did not catch them, and what we changed.
Most trading content is written to rank, by people with no numbers of their own to show. We have seven strategies with measured profit factors, real drawdowns and four public failures — and, more unusually, a running scorecard of what our own safety features cost the people they protect. The useful thing we can write is what that data taught us.
Where a piece here is a judgement rather than a measurement, it says so at the top. Nothing here is advice, and nothing here is a forecast. Our evidence standard →