Our own audit deleted
104.2% of our reported profit.
We are leading with that.

Most systematic strategies are a backtest in a notebook and a broker API key. What is expensive, slow, and rarely built is everything between: the ingestion and validation layer, the reconciliation of intended state against actual state, and above all the evidence layer that decides which records are admissible as proof of anything.

That layer is the reason this page reports −$27.17 rather than +$642.51. It is a database-enforced integrity apparatus that separates opening from closing legs, links every close to its originating entry, quarantines synthetic and diagnostic records, and fences records whose fills cannot be shown to be clean.

It is also the part a competitor cannot copy from a paper, and the part an allocator cannot verify by looking at a return series. It has to be built and it has to be trusted, which is a slower thing to earn.

All figures in this section counted directly from the repository and production database on 20 August 2026.

The strategy does not try to predict direction more often than not. It tries to be paid more when right than it gives back when wrong. Expected value is then a two-parameter question — win rate p and payoff ratio b:

E[R] = p·b − (1 − p)

Breakeven falls at p* = 1/(1+b). At the observed b = 2.992 that is 25.0% — the system can be wrong three times in four and remain solvent, provided b holds. Across the clean cohort, it has.

A barbell only pays if large upside moves actually occur, so the entry gate is a distributional test rather than a view. Fitting a Generalised Pareto Distribution to threshold exceedances yields a shape parameter ξ; when ξ > 0.25 the upper tail is heavy enough that outsized moves are plausible rather than assumed. The same family underpins flood modelling and catastrophe reinsurance, where the question is identical: how far out does the tail reach?

Position size follows a Hill-adaptive Kelly fraction, scaled well below full Kelly because parameter uncertainty at this sample size makes full Kelly reckless.

Any pitch to Nigerian capital that quotes 28% inflation and a 31% hurdle is working from stale data, and a Lagos allocator will know it in the first minute. Headline CPI printed 15.91% in June 2026, down from 15.93% in May, after the CPI rebasing.

The binding constraint is not inflation. It is opportunity cost. The CBN held the Monetary Policy Rate at 26.5% on 21 July 2026, its third hold of the year. Naira sovereign instruments therefore pay a high nominal yield — roughly 10.6 points of real return — at sovereign risk. That is the genuine alternative, and it is a demanding one.

We will also state the part that cuts against us. The naira has appreciated about 4.3% against the dollar so far in 2026 on the back of CBN FX reform. Dollar exposure has been a drag this year, not a windfall. Any pitch built on "the naira always falls" is describing 2023, not today.

So we do not claim to beat naira T-bills, and this is not a substitute for them. The honest case is narrower and, we think, more durable: an uncorrelated, USD-denominated, systematic satellite position against a book that is otherwise concentrated in a single sovereign and a single currency — sized as a satellite, not a core.

A seed investment here is not an allocation to a strategy — if it were, the 26.5% sovereign yield would win today and we would say so. It is equity in infrastructure whose primary revenue path is licensing, not performance fees.

Nigerian asset managers, pension fund administrators and family offices run large books with very little systematic infrastructure. What they lack is not a view on the market; it is exactly what has been built here — validated data ingestion, regime-aware forecasting, mechanical risk sizing, and an audit layer that can survive a regulator asking how a number was produced.

That last capability is the wedge. The restatement documented above is the demonstration: a system that fences its own profit when the fills are not clean is precisely what a compliance function needs and cannot buy off the shelf in this market.

We are a software vendor and nothing else. We do not hold client money, do not execute on a client's behalf, and do not take a share of any licensee's trading result. Licence fees are flat and recurring, priced on the software. The licensed institution holds the accounts, executes, and carries the regulatory obligation. That boundary is deliberate and we will not cross it.

The proprietary book continues in parallel, on the founder's own capital, because it is the reference implementation and the source of the evidence. It is not the business model.