barometer@atheron:/method$
// Method
Every number, explained.
Sources, steps and limits.
A measurement is only useful if you know how it was made. This page sets out where each feed's data comes from, how it is calculated, and what it does not tell you.
// Expected Volatility
A HAR-RV forecast
The heterogeneous autoregressive model of realised volatility, from Corsi (2009), a standard method in research on volatility forecasting.
Inputs
One-minute prices from our own history, turned into five-minute returns to keep out the noise of single trades.
Three horizons
Realised variance over the last day, the last five trading days and the last 22 trading days.
The fit
The weights on those three are fitted by least squares on our history, refitted every month, and checked on both halves of the year.
The output
Forecast variance for the next 30 days, turned into an annual volatility in per cent, beside its realised value and percentile.
// Stablecoin Supply and Flows
Read from the chains
Supply is what each token contract reports as its total supply. Flows are the change in that number.
Inputs
Total supply from the USDT and USDC contracts on Ethereum, Tron and Solana, through our own connections to each chain.
Flows
Net flow over 24 hours and 7 days is today's supply minus the supply at that time before.
Bridged tokens
Each chain is reported on its own. Totals add the chains' own contracts and are labelled as such.
One source
Supply comes only from the contracts. No exchange or third-party figure is mixed in.
// Limits
What the numbers do not say
Read these before you build on either feed.
Not a direction
Volatility says how much prices may move, never which way.
Not a promise
A forecast can be wrong, and in sudden markets it often is. Past readings do not predict future ones.
Not advice
Atheron Barometer is market information. It is not a recommendation to buy, sell or hold anything.
// Versions
How changes are made
So that a number you stored last month still means the same thing.
- 1
Versioned feed
Every reading names the version of the feed and of the method that made it.
- 2
Published weights
The model's current weights travel inside each volatility reading.
- 3
Notice before change
A change to the method gets a new version and a note in the docs before it goes live.
Questions
- Why not use option prices?
- An index from option prices needs the venue's permission to publish values derived from its data. We use only data we own or may publish, so today the forecast comes from price history alone.
- Where does the price history come from?
- From the public market data of the exchanges we read, collected and stored by our own systems.
- Where can I read the reference?
- Corsi, F. (2009), A Simple Approximate Long-Memory Model of Realized Volatility, Journal of Financial Econometrics 7(2), 174 to 196.
Read the full reference
The developer docs list every field in every reading, with examples you can run.