Mikronomy Research Blog
Notes from the Mikronomy research desk
Short, honest write-ups from the team building the neural-engine autopilot. We publish on model performance, market structure and what we learn running €96M of daily turnover.
Research · 8 min read
Why we rebuilt our signal-weighting layer this autumn
Three months ago the market regime shifted. Our weighting layer, tuned for a lower-volatility environment, started under-performing its back-test. Here is what changed in the data, how we detected the drift, and the 53-layer rework that replaced it.
Infrastructure · 5 min read
Co-locating in Frankfurt: a before-and-after latency diary
Moving part of our matching stack into Equinix FR2 shaved mean execution latency from 0.41ms to 0.26ms. A small chart tells most of the story, but the interesting parts are in the network path and the retry budget — both of which changed more than we expected.
Product · 4 min read
Three risk flavours, one engine — what each one actually does
Steady, Balanced and Aggressive are not three different models. They are three sizing policies on top of the same engine. We explain how each policy translates to real position sizes, how they behave in a drawdown, and why we did not add a fourth.
Opinion · 6 min read
Transparent back-tests or none at all
We publish our modelled daily rates. We also publish the periods where those rates broke. If you cannot see the losing months, you cannot judge the engine. Here is the full data file and a quick explainer on how to read it without being misled.
New posts arrive roughly every two weeks. Subscribe from the footer of the Mikronomy homepage to get them in your inbox.