Built in the open, phase by phase.

Canli Capital is a company you can audit. The engine is 12 phases deep, each tested before the next began. This page tracks what is proven, what is not, and what is next.

12 phases shipped 2,820+ automated tests mypy --strict continuous integration

Scroll

01 / The build

The integrity came first.

The order of the work is the thesis. The data integrity, the cost authority, the validation gauntlet, and the crash-safe execution loop were built before any serious effort went into chasing return, because a good backtest only means something on top of a process honest enough to trust. The engine is statically typed under mypy strict, every phase is covered by tests that must pass in continuous integration before the next phase starts, and research and live code share one path. Nothing here is a prototype dressed up as a product.

12 phases / 2,820+ tests, green / mypy --strict / CI on every change

03 / Edge status

What the build has and has not proven.

Twelve phases produced a system that is hard to fool: leak-proof data, real costs, and tests most strategies never survive. The engineering is proven. A standalone crypto edge is not yet proven; the live record is too young to claim one. A test that comes back empty is still a test that ran.

Proven The engineering and the integrity. Leak-proof, point-in-time data. One cost authority across research and paper. A validation gauntlet built to catch us out. A crash-safe loop that resumes exactly where it stopped.

Proven, modest Four algorithms, live in paper. AlphaMax in US equity momentum (standalone Sharpe 0.91) and AlphaForge in crypto funding carry (0.68), near-uncorrelated, combined at equal risk into ALPHAC. AlphaTrend (managed-futures trend) joined 2026-06-27 as the fourth: a modest 0.33 Sharpe but the first new sleeve to clear deflation (DSR 0.83). Modest, and the cross-asset book still awaits its live forward record.

Not yet The standalone crypto edge. Tested rigorously on crypto perpetual futures alone, the signal is real enough to be interesting and not yet strong enough, after deflation and cost, to stake capital on. We say so plainly.

// A platform that never reports a strategy failing its own bar is not running the bar.

Daily bars, equity lake: 24M+
US stocks, survivorship-free: 8,4361997 to 2026
Point-in-time fundamentals: 392K+
Crypto instrument universe (live + delisted): 94

// Running 24/7 on paper across the live and delisted book. No real capital in play. The build in the open.

04 / Roadmap

The way forward is breadth.

A single asset class is one bet on one market staying inefficient one way. The answer to a thin standalone edge is more lowly correlated sources of return through the same machine. The equity momentum sleeve is already live in the book. The next breadth would be managed-futures trend, but that needs futures data we have not yet invested in, so for now it stays a research question, not a commitment. The engine was built multi-asset from the first line: one data contract, one cost authority, one validation library, one execution loop. A new strategy inherits that discipline and clears the same gauntlet before it earns a dollar.