The details of how we trace miners are explained in the Appendix. We track the largest 20 pools except for four Chinese pools: BTCC Pool, BixIn, Huobi Pool, and OKExPool. These four pools are closely integrated with their corresponding exchanges. In particular, their redistribution addresses are held on these exchanges, which impedes the tracing of individual miners. Of the pools we trace, Bitfury and Lubian are private pools, which we treat as single entities. To the best of our knowledge, this is the rst study that accurately links miners to their mining pools.

Some miners choose to collect their rewards using their private wallets and some send their rewards directly to their accounts with an exchange or on-line wallet services. We call the former type private-wallet miners and the later exchange-wallet miners. We dierentiate between private-wallet and exchange-wallet miners because in the case of private-wallet miners we can more precisely identify the size of a miner since we can assign dierent mining addresses that belong to the same cluster to one miner. For exchange-wallet miners, we cannot group dierent addresses together so we treat each exchange mining address as a separate miner. As a result, we can only provide a lower bound for the size of these exchange-wallet miners since a given entity could control several addresses.

To separate private-wallet miners from exchange-wallet miners we rst check if a miners address belongs to a known exchange or entity. Since our data can miss some exchanges or OTC desks, we treat all miner addresses that belong to suspiciously large clusters as exchange-wallet clusters. These are clusters that (1) consist of many addresses, (2) receive a large number of bitcoins that cannot be traced to mining activity, (3) have many mining addresses as their members. This means we err on the side of being conservative when dening miner size.

In the next step, we screen out entities that receive irregular rewards and that received less than $1000 or fewer than 25 times of reward over their lifetime. Finally, we manually check the largest 150 largest independent-wallet miners by USD rewards to ensure that we are not mistaking re-distribution addresses for miners. After applying these lters, we end up with 105,494 private-wallet clusters and 137,656 exchange-wallet addresses. The exchange-wallet addresses belong to 305 known exchanges and on-line wallets and 284 unknown clusters. Since a miners reward is proportional to its mining capacity we measure each miners capacity as the bitcoins that are sent by pools through distribution transactions.22 In Figure 10 we plot how our algorithm captures the mining capacity in the Bitcoin