Why We Publish Our Methodology

If you've spent any time looking at trading indicators, you know how they're usually sold: a screenshot of a chart where every marker lands perfectly, a claimed win rate with no test behind it, and a methodology described — if at all — as "proprietary." You're asked to trust the tool because the vendor says it works.

We think that's backwards. Oisigma publishes its full methodology — the construction, the assumptions, the statistical tests, the figures, and the failures — in a working paper anyone can download and check. This article explains why we do that, what exactly we publish, and how you can use the same standard to evaluate any indicator, including ours.

An industry that mostly asks for trust

Search any indicator marketplace and you'll find thousands of tools making implicit statistical claims: this band contains price, this signal marks reversals, this level acts as support. What you'll rarely find is published evidence for those claims — a documented test, on stated data, with the misses included. That's not an accusation about any particular vendor; it's something you can verify yourself in a few minutes of looking. Ask, for any tool you're considering: where is the published test?

The absence isn't surprising. Publishing a methodology is work, it invites criticism, and it constrains marketing — you can no longer claim more than your own tables support. A vendor who publishes nothing can promise anything. A vendor who publishes everything can only promise what the data showed.

We can't tell you why others make the choices they make, and we won't claim their tools fail tests we haven't run. What we can tell you is which side of that line we chose, and let you inspect the result.

A band makes a claim you can check

Our indicator draws an expected range around price — an estimate of how wide "normal" currently is, measured from recent behavior. That kind of tool makes an unusually concrete claim: price should stay inside the range a stated percentage of the time. Claims like that are testable — you count, through history, how often the next close actually landed inside. We've written a full guide to comparing band indicators on that containment metric, and an explainer on what "calibrated" means for an expected range.

The point here is simpler: because the claim is checkable, refusing to check it — or checking it and not publishing the result — is a choice. We chose to run the test at full scale and publish everything it produced.

What we actually publish

The working paper — AlEssa (2026), "How Well Does a Rolling-Volatility Band Calibrate? Evidence Across Asset Classes and Market Regimes" — documents the model's construction step by step: the rolling window of returns, the expected-price anchor, and how the inner and outer bands are placed. There is no proprietary secret sauce held back; the How It Works page walks through the same logic in plain English.

It then reports the calibration record in full. Historically, across 97 years of daily S&P 500 data and 40 instruments in five asset classes, the inner band contained the next close about 71% of the time and the outer band about 94% — decade by decade, with confidence intervals, on the Proof page and in the paper. Past behavior is not a guarantee of future results, and the paper says so as plainly as we do here.

Just as important, it documents the checks that a marketing page would normally skip: whether the result is a statistical circularity (it was tested against simulated markets), whether it survives out-of-sample splits and instruments the model was never tuned on, and how a deliberately simple construction compares with the volatility models institutions use.

We publish the misses too

A methodology document that only contains successes is an advertisement. Ours has a section most vendors would leave out, and the same limits are printed on the Proof page itself:

The model's calibration is an average property, not a moment-by-moment one. In the first days of a fast crisis, the range runs too narrow, because volatility spikes faster than a rolling window can register. The outer band is slightly optimistic in the deep tails. The center line carries almost no information about direction — it anchors the range; it does not forecast where price will go. And the model will not make you money on its own: it's descriptive context, not a strategy, and the paper validates the range's calibration, not the profitability of any use of it.

Publishing that list costs us some marketing shine. It buys something we value more: when we say the band held historically, you have reason to believe we'd also tell you where it didn't.

Why transparency is the whole point

There's a practical reason a tool like ours can afford to be open. An expected range is descriptive — it tells you where price has tended to trade relative to its own recent behavior, not what to buy or when. A descriptive tool doesn't need mystique to function; its value is the accuracy of its description. Publishing the methodology doesn't erode that value. Hiding it would.

There's also an honesty reason. A published methodology is falsifiable: anyone with the data and the patience can attempt to reproduce our tables, and the paper invites exactly that — it's a working paper, complete and citable but not yet peer-reviewed, and comments are open. We'd rather be checked than merely believed.

The standard you should hold every tool to — including ours

You don't need to take a position on any vendor's intentions. Just ask the same four questions of every indicator you evaluate. Is the construction documented, or "proprietary"? Is there a published test of the tool's central claim, on stated data and settings? Are the failures and limits documented alongside the successes? And can you, in principle, check the numbers yourself?

Tools that clear that bar give you something to reason about. Tools that don't are asking for faith. We built Oisigma to clear it — not because our model is perfect (the paper is explicit that it isn't), but because a range you can verify is worth more than a promise you can't.

If you'd like to apply that standard to us, start with the working paper and the Proof page, then watch the band recalculate on the markets you actually follow. The indicator is available on TradingView with a free 30-day trial — start your trial here and judge the description against your own charts.