Assets / How this works

How this works, and where it can mislead you

What the numbers compute, what has been checked against real data, what is a judgement call, and one check that didn't hold up.

How one price is assembled

Every exchange reports. Each stroke is one exchange's price for Bitcoin and how much traded there in 24 hours. Recorded on a fixed schedule.

Some aren't counted. CoinMarketCap flags some exchanges as outliers or excludes them. They're drawn grey. We follow its choice and show it.

The middle becomes the price. Among the rest, the volume-weighted median is the agreed price. Big exchanges pull harder; no single one decides unless it holds most of the volume.

What the trust score weighs

40 · How spread across exchanges the trading is. One exchange holding everything scores zero. The biggest weight, because a price one exchange can move alone is the core risk.

30 · How many exchanges quote the same price. The share of counted volume quoting within 0.5% of the agreed price.

15 · How recently exchanges updated. Quotes more than 10 minutes old count against it.

15 · How much volume CoinMarketCap itself keeps. Heavy exclusion lowers the score a little and is always shown next to it.

The weights are a stated judgement call, not a fitted or backtested model.

80 and above reads Reliable, 60–79 Watch, below 60 Unreliable.

What puts something on the watchlist

50%+of an asset's volume on one exchange

1%+ offfor a counted exchange with at least $1M traded daily

−15 ptsfall in trust score

0.3%+between decentralised exchanges and the rest

What we checked, and what we didn't

Tested in code, independent of the marketUnit detection, the concentration, agreement and freshness maths, and tokenised scoring all run against recorded API responses in 105 tests: a zero-volume exchange, a single-exchange asset, negative funding, a dead pool.

Held across every reading so farBCH's concentration on Deepcoin, SunX quoting off-market on most assets, and the same Kraken and DigiFinex markets returned twice with conflicting prices have held in every reading since recording began.

Checked once, didn't hold up. Kept here on purposeEarly on, Deepcoin's reported total derivatives volume looked smaller than its volume for one BCH market. A later re-check found the opposite. Exchange-reported volume moves too much for a one-off comparison to count as evidence, so it isn't claimed as a finding.

Not checked, and out of scopeWhether CoinMarketCap's published price actually uses the flagged rows. Whether any exchange's volume is real or wash-traded. Whether the weights hold over a longer history.

Known limits

38 crypto assets and around 100 tokenised assets, sized to the CoinMarketCap API tier this project runs on, not the whole market.
Decentralised-exchange prices cover only BTC, ETH and LINK, on Uniswap v3 on Ethereum.
Recorded on a fixed schedule. Nothing here is real-time.
Tokenised assets are compared issuer against issuer, not against the real stock price, which the API doesn't provide.

Full source, the raw API evidence behind every claim, and the 105 tests: github.com/ketutezraugm/consensus-cmc