This plot is a heatmap of the Bitcoin distribution in BitMEX accounts on the Bitcoin ledger. The X-axis is the height on the Bitcoin blockchain grouped into batches of 144 blocks (~1 day) and the Y-axis contains accounts that are group by their on-chain balance using a logarithmic batching range where the range of account balances in a bucket gets wider the higher it goes. Each cell represents the fraction of all Bitcoin on the exchange that are attributed to accounts of that tier at that point in time. Bright yellow colors represent higher fractions of ownership, dark blue colors represent less.
This plot is a heatmap of the net-internal account flows between BitMEX accounts on the Bitcoin ledger. The X-axis is the height on the Bitcoin blockchain grouped into batches of 144 blocks (~1 day) and the Y-axis contains accounts that are grouped by their on-chain balance using a logarithmic batching range where the range of account balances in a bucket gets wider the higher it goes. Each cell represents the net-internal (to and from other BitMEX accounts only) flows between accounts where dark green represents accounts that received significant coins and dark red cells represent accounts that lost significant coins. Internal flows are interesting because with only a few notable exceptions, the exchange appears to eventually reconcile profits and losses through trading onto the Bitcoin ledger. This occurs when customers request to extract coins from the exchange, the funds seem to be sourced from accounts that have sustained trading losses and change is made into accounts that have profits.
This plot is a heatmap of the deposits made to BitMEX accounts from external (non-BitMEX) accounts on the Bitcoin ledger. The X-axis is the height on the Bitcoin blockchain grouped into batches of 144 blocks (~1 day) and the Y-axis contains accounts that are grouped by their on-chain balance using a logarithmic batching range where the range of account balances in a bucket gets wider the higher it goes. Each cell represents the total number of Bitcoins that were sent to accounts of that tier at that point in time. Bright yellow represents a signficiant inflow of Bitcoins and dark blue repesents an insignificant infow of coins.
This plot is a heatmap of the withdrawals made from BitMEX accounts to external (non-BitMEX) accounts on the Bitcoin ledger. The X-axis is the height on the Bitcoin blockchain grouped into batches of 144 blocks (~1 day) and the Y-axis contains accounts that are grouped by their on-chain balance using a logarithmic batching range where the range of account balances in a bucket gets wider the higher it goes. Each cell represents the total number of Bitcoins that were extracted from accounts of that tier at that point in time. Bright yellow represents a signficiant inflow of Bitcoins and dark blue repesents an insignificant infow of coins. It is important to note here that the relationships around accounts extracting funds is weak since the BitMEX exchange owns the private keys of these customer accounts and has full autonomy over how they choose to process customer withdrawals. For example, if a customer requests to have their coins extracted, the exchange could source those funds from a different customer account, or possibly several accounts.
This plot shows the volume of flows-(inflows, outflows, internal) on the Bitcoin ledger involving BitMEX accounts over time. This plot was constructed by recording all on-chain flows and then batching them into groups of 432 blocks (~3 days) and weighting them by their size.
This plot shows the number of flows-(inflows, outflows, internal) on the Bitcoin ledger involving BitMEX accounts over time. This plot was constructed by recording all on-chain flows and then batching them into groups of 432 blocks (~3 days).
This plot shows the number new and reutrning BitMEX accounts that have inflow activity on the Bitcoin over time. This plot was constructed by recording all on-chain account activity and then batching them into groups of 432 blocks (~3 days).
This plot shows the total Bitcoins held by BitMEX accounts (AUM) over time. This plot was constructed by aggregating the on-chain value of BitMEX accounts and batching them in 432 block (~3 days) groups. Additionally this plot also shows the Open Interest (OI) and Open Value (OV) figures for the XBTUSD instrument over a period of time.
This plot shows the ratio between the total Bitcoins held by BitMEX accounts (AUM) and Open Value (OV) of the XBTUSD instrument over time. This plot was constructed by aggregating the on-chain value of BitMEX accounts and batching them in 432 block (~3 days) groups to obtain the AUM. Then this plot normalizes by the Open Value (OV) of the XBTUSD instrument.
The plot above is a history of over 200 instruments that have traded on BitMEX over time (indexes not shown). These plots are annotated with their average daily trading volume (zoom in) and the bar for each instrument represents the time period over which it traded. Reoccuring instruments such as settled futures are represented with a distinct entry each time the instrument is re-issued. Each instrument is additionally batched into a broad category and assigned a color to match.
The plot above shows the trading volume of BitMEX over time normalized into US dollars. Over 200 instruments are normalized into USD and then batched together according to their cateogry to produce the resulting plot.
This plot shows the history of all liquidations on all BitMEX products over time. The plot is constructed by aggregating liquidations in 1-day blocks and taking a 7-day moving average over the resulting distribution. Notice that during March of 2020, our system briefly failed to capture liquidations for about 10 days during the substantial decline in prices. Additionally, we are missing liquidation data from November 2020 to January 2021.
This plot shows the number of liquidations that have occured on the XBTUSD perpetual futures instrument grouped by the size of the liquidated position into batches of 50,000 contracts. Note that the Y-axis uses a log scale.
This plot shows the cumulative volume of liquidations that have occured on the XBTUSD perpetual futures instrument grouped by the size of the liquidated position into batches of 50,000 contracts. At each point, the Y-value is the total volume of contracts from positions of similar size or smaller than the X-value which have been liquidated.
This plot shows the cumulative volume of liquidations that have occured on the ETHUSD perpetual futures instrument grouped by the size of the liquidated position into batches of 50,000 contracts. At each point, the Y-value is the total volume of contracts from positions of similar size or smaller than the X-value which have been liquidated.
This plot shows a linear regression fit to the relationship between liquidated LONG contracts on the XBTUSD instrument and the difference between the minimum and maximum traded price in a 168 hour (1 week) period.
This plot shows a linear regression fit to the relationship between liquidated SHORT contracts on the XBTUSD instrument and the difference between the minimum and maximum traded price in a 168 hour (1 week) period.
This plot shows a linear regression fit to the relationship between liquidated LONG contracts on the XBTUSD instrument and the difference between the minimum and maximum traded price in a 24 hour (1 day) period.
This plot shows a linear regression fit to the relationship between liquidated SHORT contracts on the XBTUSD instrument and the difference between the minimum and maximum traded price in a 24 hour (1 day) period.
This plot shows the number of comments in the trollbox since its inception using a 7-day moving average. The comments are additionally categoried by the language-specific channel that they were left in. At over 2,500 messages per hour in 2018, the trollbox was more prolific than many online forums for cryptocurrency discussion such as Reddit.
This plot shows the number of comments in the trollbox since its inception grouped by the UTC hour of the day that the message was left at. It is seen that different languages have different relative popularity based on the time of day as expected, but compared to recreational activities such as video games, this engagement is relatively time-invariant.
This plot shows the classification accuracy of a 5-class CNN that was trained using automatically generated sentiment labels of trollbox messages. The experiment assumes that the bullish/bearish sentiment of the trollbox is related to the price momentum of the Bitcoin market. RSI time parameters that are 'well-aligned' with the sentiment of the trollbox will contain richer sentiment signals and produce a better performing classifier than those which are poorly aligned.
These are a few examples of BitMEX accounts that do not appear to belong to typical traders but are likely instead a part of either the insurance fund or some sort of internal treasury. The distinct feature of these accounts is that they typically have no deposits from external accounts which is how a normal trader funds their account, instead all of their coins come internally from other BitMEX accounts. In some cases they have a small amount of external coins sent from dusting services. There are around 20 of these accounts in total which were each funded with 1,000-1,200 BTC around block 596,364 (September 2019) and they have remained relatively untouched for a while. Recently however it appears that these addresses have been used as a source of funds for processing withdrawls on behalf of different users.
This is a curious external Bitcoin address that has a rich history of interactions with BitMEX. This address has been labeled by Whale Alert as belonging to the Binance Exchange but we are unable to confirm this association. What we can confirm is that this address has been used to fund over 85 distinct BitMEX accounts, for thousands of Bitcoins total while extracting over 70,000 BTC from the exchange for an apparent net profit of 60,000+ BTC. This address is correlated with some extreme fluctuations in OI/OV on the XBTUSD instrument. In particular, this address would fund a few new BitMEX accounts for a total of 300 BTC or so. A few hours later there would be a flurry of buying activity and the OV of the XBTUSD contract would increase by 3,000-5,000 BTC before retracing a few days later at which point profits were extracted back into this address. The biggest action of this account comes from March of 2020 where it appears to have put on a significant short / hedge position that paid off tens of thousands of BTC when prices collapsed.
Spoofy is a term used to refer to a BitMEX trader that would spoof abnormally large positions on the XBTUSD market. BitMEX limits single positions to 10,000,000 contracts and 'Spoofy' would post orders for over 10x that, so it was clear that spoofy must have been using multiple accounts and scripts to create these trades. One possible explanation for who 'Spoofy' is comes from these BitMEX accounts that were created and funded with 3 BTC each, allowing them to apply 50 turns of leverage for a total position size of 150 BTC on each. There are a few instances where dozens of even hundreds of these accounts were created, and they roughly correspond to times where spoofy was most active.
This account currently holds the title for having the most on-chain coins in a single account on the platform at the moment, but most interesting is the interactions that it has had with BitMEX in the past. In March of 2020 when the price of Bitcoin collapsed to under $4,000/BTC on BitMEX's XBTUSD instrument, BitMEX liquidated a tremendous number of contracts. At the very bottom of the market BitMEX went offline and during this time, this particular account shipped in around 8,000 BTC, presumably to either to add margin to their own account, capitalize on the liqudiations, or ensure the solvency of the exchange.
Exchange hot wallets and services such as those for coin mixing are typically characterized by serving as a depositor to a very large number of distinct BitMEX accounts. Additionally, these accounts never recieve any funds from BitMEX other than occasionally on accident, so their flows are nearly entirely 1-direction. The account listed above is believed to be a hot-wallet assoicated with the Bitfinex exchange.
There are a number of accounts like the ones provided that have features consistent with running some sort of HFT or market making strategy. These accounts typically have a tremendous amount of on-chain interactions and in many cases the instantaneous balance is never too high, suggesting that its not so much a huge trader as it is an algorithm that is constantly reconciling BitMEX accounts with each other and with accounts on other exchanges. These accounts however differ from things like hot-wallets and services because they both send and recieve coins from the exchange where as hot wallets only send.
© 2021 Carnegie Mellon University. All rights reserved.
THE DATA AND INFORMATION MADE AVAILABLE THROUGH THIS SITE (THE “DATA”) IS MADE AVAILABLE ON AN “AS-IS” “AS AVAILABLE” BASIS SOLELY FOR NON-COMMERCIAL RESEARCH AND/OR ACADEMIC PURPOSES. TO THE MAXIMUM EXTENT ALLOWED UNDER LAW, CARNEGIE MELLON UNIVERSITY IS NOT RESPONSIBLE FOR ANY CLAIMS, DAMAGES OR OTHER LIABILITY ARISING OUT OF THE USE OF THE DATA, AND YOU ARE RESPONSIBLE FOR YOUR USE OF IT. THE DATA IS NOT PROVIDED TO (AND MAY NOT BE USED BY) ANY PERSON OR ENTITY IN ANY JURISDICTION IN VIOLATION OF APPLICABLE LAWS, RULES OR REGULATIONS. BY DOWNLOADING, COPYING AND/OR USING THE DATA, YOU AGREE THAT YOU HAVE READ AND AGREE TO THE PROVISIONS IN THIS PARAGRAPH AND THE TERMS OF USE BELOW. IF YOU DO NOT AGREE, YOU MAY NOT ACCESS OR USE THE DATA.
bc1qmexgxvszgv65mf8xsl77wmarfyk35m842cu0ztcrrka5nh0tz69q0zhd3g