Calibration
The study behind every reading, and how the bands have done lately
What the Calibration page is
Every number Optionomics shows a trader—the priced move, the odds of reaching it, the cost of the trade, the earnings and volatility reads—appears with the backtest it rests on. The Calibration page at optionomics.ai/calibration, titled The study behind every reading, publishes those backtests in one place: what was measured, on what sample, how, when, and what each reading does not claim.
It also keeps score in public: how the one-session expected-move bands have actually done, month by month and over the last week.
Availability: Public. Anyone can read it, with or without an account.
None of these studies is about direction. Every study of direction in this data measured no better than chance, so no reading on Optionomics is built on one. The readings are about size, odds, and cost.
Touch odds on the expected move
The first section, Touch odds on the expected move, answers a practical question: how often does the next session reach the edge of the range the options market priced, and how often does it close beyond it?
It was measured on 4.0 million single-name sessions from 2016 to 2026, point-in-time—each session judged against its own previous day’s chain, never with hindsight.
| Session | Touch 68% edge | Close beyond 68% | Touch 90% edge | Close beyond 90% |
|---|---|---|---|---|
| Single name | 59.6% | 30.5% | 19.2% | 9.3% |
| Single name, earnings in the session | 55.6% | 32.0% | 20.0% | 10.3% |
| Fund (ETF) | 55.8% | 32.6% | 17.4% | 9.5% |
Figures as published on the page when this guide was written; the page is the source of truth.
The takeaway: an edge is touched about twice as often as it is closed beyond (1.95 times at the 68% edge, 2.05 times at the 90% edge), and that ratio stayed between 1.86 and 2.19 in every year, every liquidity and implied-volatility group, and on both edges. Earnings sessions and funds touch less often for the same close, because more of their move arrives overnight. These are the rates shown on the stock page’s Odds card.
The page notes that the sample is a survivor panel—stocks that are still listed today.
How the one-session bands did, by month
The second section, How the one-session bands did, by month, checks the expected move against what actually happened, continuously. Each session’s range from the night before’s chain is compared with the session that followed.
A well-calibrated band should be:
- 68% band: closed beyond about 32% of the time, and touched about twice that.
- 90% band: closed beyond about 10% of the time, and touched about 20%.
The section opens with a Last week line—for single names and for funds, how many sessions were measured, how many closed beyond the 68% band, and how many touched it. Below it, a table covers the last twelve months:
| Column | Meaning |
|---|---|
| Month | The month the outcomes fall in (marked as “to date” while the month is still running) |
| Group | Single names or Funds (ETF) |
| Sessions | How many stock-sessions were measured |
| Close beyond 68% / Touch 68% edge | How often the session closed outside, or reached, the 68% band |
| Close beyond 90% / Touch 90% edge | The same for the 90% band |
Sessions are left out where the chain’s price is more than 1% away from that day’s close, or where the next session has no price bar. Each rate is a frequency, not a direction. The figures are updated each Saturday, after Friday’s session.
A month quieter or busier than the options market priced will sit away from the targets—that is what calibration looks like in real time. Over many months the rates should hover around 32% and 10%.
Every reading’s record
The last section, Every reading’s record, holds the same records printed under each card of the stock page’s Session Brief. Each record has:
- What was measured — the headline statistic.
- How — the test design: walk-forward, held-out years, point-in-time.
- What it does not claim — the caveat, which is part of the claim.
- The date it was measured and the sample it was measured on.
The records cover:
| Record | What it backs |
|---|---|
| Priced move (the expected-move ladder) | The Priced to move card, the Exp Move chart overlay, Today’s list, and the expected_move MCP tool |
| Touch and close odds | The next-session Odds card |
| Touch decay (the live odds) | The live Odds to the close card for single names during the session |
| The live 0DTE read (SPY and SPX) | The 0DTE tab and the brief’s cards on SPY and SPX |
| Cost to be in the trade | The Cost card and the contract_chooser MCP tool |
| Earnings move and IV crush | The Earnings card and the earnings_move MCP tool |
| Volatility premium and IV path | The IV path card and the vol_premium MCP tool |
Each record also says plainly where a reading stops. For example: the expected move is a size, not a direction; the cost card shows a quote, not a fill; the earnings forecast is not a straddle trade, because the quoted spread is many times the whole earnings premium.
Why it matters
A reading you cannot check is a reading you have to take on trust. The Calibration page lets you see the study behind each number before you act on it, and the monthly table lets you see whether the bands are still doing what the study says they do.
Related:
