Sigma-1 model

# How it counts.

The activity score compares volume and price movement with the asset's own recent history.

## Read the scale

Illustrative

53/100

Elevated activity

Activity is elevated compared with this asset's baseline. At 53, it has not crossed the event threshold of 70.

1. 01 **Compare with its own history.** Each asset is compared with its own last 30 days. Volume and price-move anomalies are combined into a 0 to 100 activity score.
2. 02 **Record the threshold crossing.** A score of 70 opens an event. The starting time and price become the reference for what happens next.
3. 03 **Keep the receipt.** Measure the price at 24 hours, 7 days and 30 days. Rising and falling prices stay in the same record.

## Bands

- **Quiet · 0 to 19**: At or below its usual level
- **Normal · 20 to 39**: Close to normal
- **Elevated · 40 to 59**: More than usual
- **High · 60 to 79**: Well above normal
- **Rare · 80+**: Among the most unusual
- **Scale**: A composite 0 to 100 activity score, not a z-score or a number of standard deviations.

## Contents

## The models

- **The number**: Counted by code under a written rule with a version and a hash. The same code runs in public in the tickerz-open repository, so anyone can count it again and get the same number. A chain count is signed only when two providers return exactly the same value.
- **Estimates**: Where a number is sampled ($GIGS counts a fixed 1 in 16 sample; some launched tickers read short windows), the sample is fixed by the rule and the number carries its 95% range. It is a model too, and a reproducible one: the same sample gives the same answer.
- **The score**: The 0 to 100 Activity score is a statistical model: each reading against its own past, as clamped z-scores. A version 2 (robust z-scores against the same weekday's past, falls counted as much as rises) was backtested on every index's public history on October 7 2026. It caught big moves no better (9 of 21 under 40 for both), so it stays off. Nothing settles on the score.
- **AI**: Used where judgment helps, never on a settled number. Claude drafts a launched ticker's rule from plain words when the matcher cannot, and the draft is checked like any other pick, so it cannot name a source Tickerz does not read. Jev labels event text and checks each X post before it goes out.
- **Holds**: A print that looks wrong (a robust z beyond 6, or under half the usual rows) is held and read once more before it is signed.
- **Calls**: Calls on the next print, in the Arena and on Ticker it, are for display and points. Nothing settles on them.

## Model · Sigma-1

- **Compared with**: Each asset is compared only to itself, so the unit of measurement is departure from baseline, not size. A quiet asset waking up outscores a large asset doing what it always does.
- **Inputs**: 24-hour traded volume and the absolute 24-hour price move. Both are public and checkable.
- **Not inputs**: Stablecoin supply and DeFi TVL are collected as context for future models and do not touch Sigma-1. Funding rates and open interest were collected the same way until October 6 2026.
- **Blend**: The clamped z-scores combine in a fixed weighted blend, scaled so 100 needs both inputs at the clamp at once. Volume counts more than the price move. The weights are proprietary.

## Baselines

- **Window**: Each input is converted to a z-score against that asset's own trailing 30-day distribution. Z-scores, per ticker. That is where the name comes from.
- **Clamp**: Z-scores are clamped at four standard deviations, so one extreme print cannot pin the scale. The score uses the clamped values; the event record keeps the raw z-scores that opened it, in the API and in the stamped open record.
- **Floor**: Z-scores are floored at zero before the clamp: a reading below its own baseline average contributes zero, never a negative number.
- **Minimum**: No asset is scored until 14 days of baseline exist. New listings appear unscored.
- **Regime change**: An asset that becomes permanently busier normalizes within weeks, so the score keeps measuring what is new.
- **Correction · September 22 2026**: The 30-day baseline included the asset's own latest reading, which pulled real anomalies toward zero on close calls. It now leaves out the current day. Inputs and blend weights did not change. Scores and events recorded before this date used the earlier baseline.
- **Source change, October 6 2026**: Prices and volumes come from Alchemy for every asset: coins by symbol, each stock through its xStock (a tokenized share backed 1:1) by contract address. CoinGecko and Alpaca are no longer read. A stock's volume is its xStock's own 24-hour trading, around the clock, so stocks are scored like coins: against the last reading of each prior day. Each asset is compared only with readings from the same source; its first 30 days came from Alchemy's daily history (marked as loaded), so scores did not pause. Stocks with no xStock, or under $50,000 of xStock trading in a day, are off the Terminal. Everything recorded before the change stays as recorded.
- **Correction · equities, September 23 2026**: A stock's volume counts from the opening bell, but its baseline was built from end-of-session readings, so every stock read quiet in the morning: over the prior 30 days the median equity volume z-score was -1.7 before 11:00 New York time and -0.2 after 15:00. Each stock was then compared with its own readings at the same point in the session, in 10-minute steps, over the same 30 days, and the price move from the prior close was matched the same way. On the same history the median volume z-score by hour sat between -0.13 and -0.01. The first reading of each session, taken seconds after the open, was recorded and not scored, because the feed could still carry the prior session's daily bar. Two equity events, RKLB and QQQ at 09:30 on September 22, opened on such a reading. Equity scores through September 22 read low in the morning; they, and the events and receipts they opened, stay as recorded.

## Events

- **Opens**: An activity score at or above 70 opens an event and freezes the asset's price at that instant.
- **Closes**: The score must fall below a cooler band and stay there across consecutive readings, so one surge produces one event.
- **Recorded**: The opening activity score, the peak and the triggering z-scores.
- **Correction · September 22 2026**: The field published as the opening score held the event's peak. The record now publishes both, the score of the reading that opened the event and the peak.

Illustrative

## Receipts

- **Horizons**: Every event resolves at +24h, +7d, and +30d as the forward return from the frozen open price, computed from Tickerz's own stored history so nothing can drift after the fact. Rises and falls alike.
- **Benchmark**: Each receipt also records the benchmark's return over the same window, Bitcoin for crypto and SPY for stocks.
- **Clock**: Receipts resolve on the next Terminal reading at or after each horizon, within 15 minutes, for coins and stocks alike. Before October 6 2026 a stock resolved at the next reading in a regular session.
- **Finding**: The one finding that has held so far: a crypto upside spike while under 3% of coins are unusual lagged Bitcoin by a median of about 5% over the next 7 days, about 3 points worse than an ordinary day, in all 5 years studied. Downside spikes show nothing separable, and the stock study is not published yet.
- **Read**: [Evidence](https://oracle.tickerz.com/evidence) [Attention fades](https://oracle.tickerz.com/research/attention-fades) [Receipts](https://oracle.tickerz.com/receipts) [Sandbox](https://oracle.tickerz.com/sandbox)

## Proofs

- **At open**: Since 20:30 UTC on September 22 2026, every new event is stamped when it opens. Within about an hour of opening, Tickerz hashes the open record (event id, symbol, open time, open price, horizon open, opening score, triggering z-scores) with SHA-256 and submits the digest to public OpenTimestamps calendars, so the record proves the call existed before its outcome.
- **Earlier events**: Events already on the record at that time were stamped later, in hourly batches from 20:30 UTC on September 22 to 05:30 UTC on September 23 2026; a receipt whose open stamp landed more than two hours after open says how long after.
- **Receipts**: Each resolved receipt is stamped the same way (event id, symbol, open time, reference price, horizon, resolved time, return).
- **Bitcoin**: The `.ots` proof is stored with each record. Once calendar aggregation lands in a Bitcoin block, the receipt shows the block number, linked to the block, beside a download link for the `.ots` proof. Until a Bitcoin height is known the stamp is calendar pending, not yet a block anchor.
- **Verify**: 1. Download the `.ots` from `/api/receipts/proof?event_id=&horizon=&format=ots` (horizon is open, 24h, 7d or 30d; omit `format=ots` for JSON metadata). 2. Hash the canonical JSON with SHA-256 and compare it to `digest_sha256`. 3. Check the proof at [opentimestamps.org](https://opentimestamps.org/) or with the `ots` CLI.

## Index prints, signed

- **Rule**: Each index has one written rule with a version and a hash: source, query, unit, rounding, revision policy, and whether the source's terms allow a market to settle on it. `/api/indices/{ticker}/rule`
- **Print**: Each print is a report signed with the Tickerz publisher's key. Chain counts are signed only when two providers return the same number; official numbers only when every read equals the number as first published. `/api/indices/{ticker}/report`
- **Final**: A print is provisional until an independent recount in the public tickerz-open repo counts it again, gets the same number, and co-signs, and 24 hours pass. A correction is a new signed print that names the one it replaces; nothing is edited.
- **Solana**: One account per index holds the newest print and the 16 before it. Only the publisher writes it; any program can read it.
- **Status**: Prints are signed and recounted since October 7 2026. The Solana accounts are written once the program is deployed; until then the signed report is the record.
- **Every feed**: Each index's newest signed number, its status and its Solana account: [the oracle](https://oracle.tickerz.com/).
- **Check it**: Keys, rule hashes and the account layout: [tickerz-signer.json](https://oracle.tickerz.com/.well-known/tickerz-signer.json). Every signed print and where it stands: [Signed numbers](https://oracle.tickerz.com/signed).

## Coverage

- **Crypto**: The top 250 crypto assets by market capitalization that Alchemy prices by symbol refresh every 15 minutes around the clock. The universe widened to the top 250 on September 22 2026; a coin Alchemy cannot tell apart from another is off the Terminal, and its page says why.
- **Left out**: Stablecoins, wrapped and staked tokens, tokenized funds and other pegged assets: no readings, no score. A token that shares a stock's ticker is recorded but kept off the Terminal.
- **Equities**: A curated set of major US stocks and ETFs, each read through its xStock every 15 minutes around the clock on the identical mechanism. Until October 6 2026 they were read every 10 minutes in market hours, against their own readings at the same point in the session.
- **Late feeds**: When a feed goes quiet, the Terminal says how late it is, and a stopped row reads STALE with its date.
- **Access**: The board, asset pages, receipts, proofs and the MCP server are free through the [API](https://oracle.tickerz.com/docs), with no key. Three deeper reads (full history, the board at a past moment, and every receipt with its proof) are built to be paid per call and are closed for now.

## Public and proprietary

- **Public**: The inputs, per-asset standardization, the four-sigma clamp, the 14-day minimum baseline, the event threshold at 70, the existence of hysteresis, and the receipt horizons.
- **Proprietary**: The blend weights, the close band, the confirmation count, and any future calibration parameters.

## Evidence

- **Measure**: [Evidence](https://oracle.tickerz.com/evidence) measures every resolved receipt as excess return: the event asset's return minus its market's return over the same window, Bitcoin for crypto and SPY for stocks, reported whichever way it comes out.
- **Why excess**: Raw return after an event is not evidence: if the whole market rose, every event looks like it worked.
- **Significance**: An exact two sided binomial test against a coin flip on how many events beat the benchmark. It tests direction only, not size. Stock rows are still few, so read them as a count, not a finding.
- **Ordinary day**: On Evidence an ordinary day is every 7 day window for every listed asset since the record began on August 31 2026, event days included. On Attention fades it is a day the score stayed well below the event threshold.
- **Limits**: Nothing there is a forecast or advice, and none of it is annualised.

## Sigma-2

- **Stored now**: Market context beyond Sigma-1's inputs: stablecoin supply and DeFi TVL.
- **Next**: Once the receipt record is long enough, Sigma-2 will replace the fixed blend with a calibrated score and publish the calibration.
- **Sigma-1**: Stays frozen and versioned beside it.

The activity score measures attention. It is not a prediction or financial advice. [Disclaimer](https://oracle.tickerz.com/disclaimer)