Monolith Research
The year was a building year. The desk did not begin the year with a settled view of which strategies it would prosecute; it ended the year with a settled view of which ones were not worth prosecuting, and with a small, deliberate list of candidates that had earned the next round of observation. The shape of the file is the principal achievement. The live book remains modest by design.
The paths that were closed during the year are the larger part of the file. A broad transfer of a working intraday index model into spot foreign exchange was tested and closed; the model did not survive once the implicit friction of a different liquidity venue was made honest. A daily macro-sentiment pipeline built around tone-classified news flow was tested and closed; the classifier could not separate the next session at better than coin-flip and was retained only as a possible future risk-warning layer. A handful of decorative additions to the existing intraday work, more aggressive frequency variants and optimistic overlay filters, were tested and closed for the same kind of reason. Each gave back more in risk shape than it added in return.
The paths that were opened are short and deliberate. A cross-asset diversification scan was begun, framed by the question of what the principal book had been quietly borrowing return from, and what would actually reduce that borrowing. A promotion pipeline was written down so that future research candidates have a named route to live size rather than an ad-hoc one. A deterministic risk shell was written down as the next operational change, in front of any signal-layer change.
The observation worth recording for the year is structural. Three independent passes produced the same conclusion in different forms: the marginal cost of complexity in this stack has, every time it has been measured, been higher than the marginal return. That is not an argument against new research. It is an argument for the order in which research is allowed to move into the book. The shell goes in front of the signal. Walk-forward goes in front of paper. Paper goes in front of shadow. Shadow goes in front of size. The year closes with that order intact, and the next year is being planned inside it.