Chart explorer

Nine days of MtGox in the autumn of 2011, and eleven months of 2014 and 2015 across six venues.

Pick an era, pick venues, and the page pulls one static JSON file per series out of the release tree on bitcoinpricedata.com. Nothing is computed while you wait. Nothing here is live either, so the newest bucket on this page is exactly as old as the newest release on the dataset site.

MtGox failed in February 2014, taking the exchange that priced most of the world's bitcoin for four years with it. BTC-e was seized by the FBI in 2017, BTCChina closed its exchange the same year, and OKCoin's yuan book died with the People's Bank of China ban. Those books are on this chart because someone was writing them down at the time, four seconds apart in 2011 and two seconds apart in 2014, and the dumps survived in an archive.

Plot it

Era
Plot
Quote currency
Resolution
Venues

Each point is the price the bucket closed on, in the venue's own quote currency. In the 2014-2015 series that is the last trade. In the 2011 MtGox series it is the midpoint between best bid and best ask, because that bot logged the book and never the tape, so there is no trade price in it to plot. The two yuan venues are never drawn on the same axis as the dollar venues: no exchange rate was recorded beside these rows, and inventing one would put a number on the chart that the venue never quoted.

Drawn from the historical series as published on bitcoinpricedata.com. Read the methodology, or download the release and check the arithmetic yourself.

Read it as a table
VenueQuotePointsFromToFirstLastLowHigh

One row per plotted series, over the whole span the file covers. The full rows at full cadence are in the release, not here.

Why the eras are never joined

This project has measured order book depth three times, three different ways. The 2011 trading bot summed resting volume within fifty cents of the best price and weighted it by distance, then stored the log of the ratio between the two sides. The 2014 collector took the top sixty levels a side, whatever price range those levels happened to cover, and stored a ratio running from -1 to +1. The collector running now sums everything inside two percent of the best price and is not charted here at all. Same idea, three windows.

A line that ran them together would be a series nobody ever collected. So the era selector takes one era at a time, the boundary between them is a boundary rather than a joint, and the imbalance plot switches itself off for an era whose depth numbers are not that ratio. The methodology page works through what changes across it and what a downloader has to carry into the analysis.