Each point on this line asks the same question the cards above ask, but using only the trading days up to that date. Early on there is not much evidence either way. As days accumulate the evidence either builds or it does not, and the line is that process.
Green means the reading stood at or above the 95% threshold on that date. Gray means it stood below. The color comes from what the number actually was that day, not from what it is today, so a strong current reading never repaints a stretch of history that had not earned it.
The line begins at the 60th trading day. Before that the standard error rests on too few observations to mean much, and plotting it would be false precision rather than more information.
The line is a monitor, not the claim. A statistic checked every day will touch a fixed threshold more often than a single reading would, which is why a first crossing is never treated as a finding and why the label downgrades the day the reading falls back. The record crossed the 95% line in February 2026, fell below it several times through April, and has held above it since May; that path is shown as it happened. The claim we make rests on the reading at the weekly report date, and the interval on the excess-return card above is the number that says how much the magnitude is still uncertain.
Technically: expanding-window t-statistic of the mean daily excess return versus RSP, with Newey-West HAC q=4 standard errors. The dashed line is the 95% threshold (t = 1.96), the dotted line is 99% (t = 2.58). The full history is shown, including every period below threshold, because a metric that only appears when it is favorable is not a metric, it is marketing.
Every trading day, we compare the strategy's return against RSP, the equal-weight S&P 500 ETF, the benchmark we committed to before trading began. The average daily edge is then measured against how noisy those days were. The result is a t-statistic, a standard measure any analyst can recompute from the same data. A t-statistic of 2 roughly means that if the strategy truly had no edge, a record this strong would appear less than 5 times in 100 by chance.
The current reading holds under five independent tests, including tests that make no assumptions about how markets behave. Because this is a live, prespecified record, a single strategy against a single benchmark chosen in advance, the discount a reader applies to data-mined backtests does not apply here. The discount for a short sample still does. Fourteen months is long enough to distinguish the edge from zero; it is not long enough to say how large it is or whether it persists across regimes the record has not yet seen, and we do not claim either.
Three things this does not mean: it does not predict future returns, it does not guarantee any level of outperformance, and it is not a third-party verification. It is our live record, tested honestly, updated daily. Most figures on this page are gross of fees. The primary test is also run net of the 2% model advisory fee and both readings are shown side by side in the test battery below, along with net cumulative and annualised return in the metrics table.
Three t-statistics, three questions. A reader moving across FLIGHTDECK meets three significance readings and they are not the same test repeated. The primary reading on this page asks whether the strategy earned excess return over RSP, its prespecified benchmark, across every trading day. The PB/AO Analytics tab on the AC Market Monitor asks a narrower question: on Alpha Opportunity days alone, is the strategy's own return distinguishable from zero? That test uses fewer days and a different null, so its t-statistic is lower even when the same record is strong. Factor Attribution asks the hardest question: after removing what market, size, value and momentum exposures would have delivered, is there an intercept left, and does it survive the 2% model fee? Its t-statistic is lower again because each factor absorbs part of the excess return, and the net version subtracts a fixed drag from the mean without narrowing the standard error, as described for the HAC rows below. A record can clear the first bar, clear the second, and fall short of the third, and that pattern is what a short live record with a real but modest edge looks like. Each page states which question it is answering beside its number; none of the three is a restatement of another. The PB/AO regime method behind the second reading is set out in a peer-reviewed paper forthcoming in the Journal of Investment Management.
| Specification | What it assumes | Reading | Cadence |
|---|---|---|---|
| Newey-West HAC q=4 (primary) | Allows serial correlation in daily returns. Asks whether the selection process produced excess return. | — | Daily |
| Newey-West HAC q=4, net of fees | Same test and same series, less a 2% model advisory fee accrued daily. Asks whether a client keeps excess return after paying us. | — | Daily |
| Independence (IID) t-test | Textbook assumptions | — | Daily |
| Jensen alpha regression (HAC) | Controls for beta rather than assuming beta of one | — | Daily |
| Stationary block bootstrap | No normality assumption; resamples the actual record in 10-day blocks | — | Daily |
| Closed-period sign test | Distribution-free; every completed cycle vs benchmark. Descriptive until ten or more periods have closed | 5 of 5 positive, p = 0.031 | Weekly report |
| Metric | Value | Metric | Value |
|---|---|---|---|
| loading… | |||
These tests run on the same daily record shown across FLIGHTDECK. The next step in the chain is the live evidence itself.