Dodep Capital is a research firm. The laboratory is a live market, and the market keeps the score. We build systems that form hypotheses about price, test them against recorded history, and then submit them to the only referee that cannot be argued with.
Most research is graded by other researchers. A backtest is a claim about a world that has already happened; a paper is a claim about a claim. We are not interested in either. A strategy either quotes into a real book at real size and survives, or it does not. That constraint removes the need for opinion, and it removes the need for adjectives.
The work is a loop. Read the tape. Propose a mechanism. Replay it over tick-level data with the fills the venue would actually have given. Split the record so that the test set was never seen. Size it small. Watch it for longer than is comfortable. Kill it early, and without ceremony, when the evidence stops arriving. Most of what we write is deleted, and the deletion is the product.
We run this loop on ourselves — the research agent proposes, the market grades, and the score is written down before we are allowed to look at it.
Two disciplines follow from this. The first is measurement: environments faithful enough that a result obtained in simulation is worth acting on, with contamination treated as a defect rather than a nuisance. The second is presence: continuous two-sided quoting, because a firm that must be in the market at all hours learns things a firm that chooses its moments never will.
The first account we ever funded spent its first eleven months underwater. We did not add capital, change the mandate, or stop writing down the score. It recovered. Nothing about that period was interesting except that we kept the record, and the record is why we trust the method.
We publish when there is something to publish, and not before.