State what was collected—and what was not
No exchange API was queried for this study. Five invented top-of-book records test arithmetic and filtering in a deterministic local generator. Pair name, times, prices and sizes are fabricated; they are not disguised historical measurements.
Download the five invented records and all-row results from Sources & method. Reproducing the calculation verifies those rules, not market conditions. A later observed study needs a separate collection record, permitted data use and an editorial check before publication.
Evidence: Download the five invented book records (JSON) · Download all spread results and exclusions (CSV)
Declare the denominator and units
For an accepted record, spread in quote units is ask minus bid. Midpoint is (ask + bid) / 2. Midpoint-relative spread in basis points is spread / midpoint × 10,000; divide basis points by 100 for percent.
An invented bid of $99 and ask of $101 produce a $2 spread, $100 midpoint and 200 basis points, or 2%. At $99.90 and $100.10, the spread is $0.20 and 20 basis points, or 0.20%. These are distinct units, not alternative labels for the same numerical value.
| Bid | Ask | Midpoint | Spread | Spread (bps) |
|---|---|---|---|---|
| 99.00 | 101.00 | 100.00 | 2.00 | 200.00 |
| 99.90 | 100.10 | 100.00 | 0.20 | 20.00 |
Preserve diagnostic rows instead of forcing a result
The fixture also contains a crossed record (bid above ask), a missing bid, and a locked record (bid equals ask). This teaching protocol requires finite positive quotes and sizes with ask greater than bid for its ordinary continuous-market comparison.
The locked record is excluded by that chosen scope, not because a zero arithmetic difference is impossible. The crossed record is not clamped to zero. The saved CSV retains all five rows, statuses and reasons; two are accepted and three excluded. No mean or venue ranking is inferred from this toy set.
Define a future collection record before requesting data
Coinbase Exchange’s public book documentation distinguishes best-quote level 1 from deeper levels and specifies price/size arrays and sequence/time fields. That documentation was read as a schema reference only. It does not license a claim that these invented rows came from Coinbase.
For any future collection, retain venue and exact product identifier, requested level, market/auction state, source timestamp when supplied, local UTC retrieval time, raw response and parse version. Confirm access and redistribution terms. A sequence value is not a substitute for an observation timestamp.
Evidence: Coinbase Exchange: get product book
One snapshot does not measure execution or a whole day
Even a valid observed best-quote spread would describe a particular pair and moment. It would not establish how much liquidity survives until an order arrives, what fees apply, or what a larger order actually pays.
A time average needs a declared sampling schedule, outage policy, weights and denominator. A cross-venue comparison additionally needs matched products, currencies and observation windows. No such sampling or matching was performed for this study.
Audit the saved fixture before interpreting the chart
Use the linked JSON fixture and CSV to check both accepted formulas and confirm that rejected observations remain visible. Apply the declared positive-quote/size and ask-above-bid rules to each record. Calculate midpoint, spread and midpoint-relative basis points, rounding to six decimal places only for saved display fields.
The current CryptoTrades market table has no bid–ask series. Use its source disclosure as a reminder to separate available data from a proposed study. This article is a synthetic method demonstration, not an assessment of any exchange’s spread quality.
Evidence: CryptoTrades data methodology · Download the five invented book records (JSON) · Download all spread results and exclusions (CSV)
Sources & method
- Coinbase Exchange: get product book
Public schema and depth-level documentation only. No live API request or observed book was collected. Checked 2026-10-29.
- CryptoTrades data methodology
Market snapshot coverage and timing labels; no order-book depth or spread series is exposed. Checked 2026-10-29.
- Download the five invented book records (JSON)
All prices, sizes and timestamps are synthetic. No exchange market data was collected. Checked 2026-10-29.
- Download all spread results and exclusions (CSV)
Two accepted and three excluded records with reasons. No provider ranking, observed spreads or liquidity study. Checked 2026-10-29.
Continue learning
Found an error? Contact the editorial team with the page and supporting source. Never send private account information.