Is miner sell-off the reason for Bitcoin's decline? An analysis of exchange flow data

This article was published in Crypto Valley Live, author: Karim Helmy, translation: Edward.

Miners are often blamed for the decline in Bitcoin prices. These accusations are often unfounded—worse, they are sometimes based on incorrect metrics that conflate mining pool payouts with miner expenditures, ultimately misleading users.

Accurately assessing the extent and impact of miner sell-offs is crucial for understanding the market. In "Following Flows: A Look at Miners' On-Chain Payments," Coin Metrics released a new method for estimating miner activity. This method takes into account the structure of pool wallets, allowing users to distinguish between pool and miner activities.

In this article, we will refine our estimates of miner activity through data on the relationship between miners and exchanges, enabling us to determine when and where miners sell coins. All metrics used in this article will be available in our upcoming version 4.9 of Network Data Pro.

According to traditional experience, we find that miners tend to prefer Huobi and Binance over other exchanges. We also found that the flow from mining addresses accounts for a small proportion of the total inflow to exchanges, approximately 5.5% at the time of writing, and is not a major source of market volatility.

Mining Market Metrics

Coin Metrics' miner-exchange flow is based on our existing estimates of miner and exchange activity. Existing inflow, outflow, and supply metrics provide useful context regarding market sentiment, exchange health, and network decentralization. Due to the differing structures of the issues, exchange flow and miner flow are built on two different clustering techniques, each with its own drawbacks.

Exchange flow uses common inputs and wallet ownership to estimate, assuming that addresses sharing the same transaction input belong to a single owner. This technique is precise but requires each exchange to have at least one seed address, limiting coverage to a predetermined universe of exchanges.

Centralized exchange transactions typically require users to deposit coins into the exchange, which are then held by the exchange operator. Mining generally works in a similar way: miners share resources with each other to increase the chances of finding blocks, coordinating with centralized operators through mining pools. Mining pool operators typically receive newly mined coins to addresses they control before distributing them to miners through payout transactions.

Coin Metrics' miner flow illustrates this architecture by clustering based on the hop distance from addresses to coinbase transactions. Addresses that have received coinbase rewards, or 0-hop addresses, are assumed to belong to mining pools. 1-hop addresses that have received payments from 0-hop addresses are marked as belonging to miners. This analytical approach is less precise than the common input-ownership analysis but roughly reflects the structure of pool wallets and provides better coverage.

Neither clustering can detect the use of fresh addresses, and the results of each clustering should be viewed as rough estimates. Since exchange tagging is more precise, we tend to favor this analytical approach when resolving conflicts: addresses tagged as belonging to both miners and exchanges are specifically treated as belonging to exchanges.

By combining these two techniques, we can assess where miners are storing coins, which roughly corresponds to where they are selling coins.

The flow of funds from miners to exchanges is generally similar to the overall inflow to exchanges, but there are several key differences. Binance and Huobi are currently the best-performing exchanges, with Huobi's share being significantly higher compared to its overall inflow proportion.

These results seem intuitive, as Huobi and Binance are the only two exchanges in the coverage area that also operate mining pools. These exchanges also have close relationships with miners and hold significant influence in Asia, where most miners are located.

The outflow of funds from exchanges to miners is also primarily from Binance and Huobi. This indicates that miners are also buying on these exchanges.

Like the inflow of funds, Huobi has a significantly higher share of outflow compared to exchanges that do not operate mining pools. Binance's share is roughly equivalent in both distributions.

In the future, miner-exchange flow can serve as a representation of the geographical distribution of mining rights. This will depend on the assumption that miners use exchanges in their respective regions, and it will also require more comprehensive exchange coverage to paint a representative picture.

Measuring Miner-Exchange Flow

Our method can be used to assess the composition and overall scale of miner-exchange flows. Overall, mining pools are net depositors to exchanges, although the amount of funds transferred daily is small.

Like their overall flow, the 1-hop miner-exchange flow is an order of magnitude larger than the 0-hop flow. Although these flows, measured in BTC, have declined since their peak in 2016, they have gradually increased since 2018. In contrast, the flow measured in USD has surpassed previous highs, reflecting the rise in Bitcoin prices.

Paradoxically, according to our assumptions, miners appear to be net buyers. This is likely due to methodological issues caused by multiple factors.

One factor may be that most miners sell off their coins over-the-counter, meaning that funds do not immediately flow to exchanges. The accumulation of early miners may also be a contributing factor, which can be addressed by filtering out outflows from addresses that have not received funds from 0-hop addresses recently. Finally, payments from exchange-affiliated mining pools may be conflated with exchange withdrawals.

Despite these results, the total flow appears to be in the right direction; when combined with flows from specific exchanges, their granularity is also sufficiently useful. However, since the discrepancies between these inflows and outflows are always unexpected, the net flow calculated from these time series may be of limited value.

Content of the Flow

Understanding miner-exchange flow is best done in the context of its constituent flows. To understand the scale of miner-exchange flows, it is helpful to compare them with total miner flows and total exchange flows.

Funds flowing into exchanges typically account for a high single-digit percentage of miner flows, although this percentage has varied historically. Deposits to exchanges also account for a significant percentage of pool flows.

The proportion of funds withdrawn from exchanges in miner inflows is higher, which may be due to exchange-affiliated mining pools and over-the-counter sales. Funds withdrawn from 0-hop addresses generally account for a much smaller proportion of total flow, although this has recently begun to increase.

In theory, miner deposits to exchanges should correlate with sales, but over the past few years, miner deposits have typically only accounted for a single-digit percentage of exchange inflows. At the time of writing this report, miner deposits accounted for about 5.5% of exchange inflows. While most miners sell off their coins over-the-counter, this number may be upward biased due to the broad net cast by the 1 analysis method, meaning the actual selling pressure from miners within exchanges may be lower.

While miner-exchange flow itself is volatile, its fluctuations are small and not significant in the context of overall exchange flow; the flow between 0-hop addresses and exchange addresses can be ignored. Therefore, we see no reason to attribute downward fluctuations in Bitcoin prices to miner sell-offs.

Miner withdrawals account for a large proportion of the total outflow from exchanges, generally remaining above 10% last year. The difference between this figure and the proportion of inflows from miner deposits may relate to the role of exchange affiliates, which may maintain close ties with their parent companies and may receive payments from overlapping addresses.

The lack of relationship between prices and miner activity is confirmed by the low correlation between prices and deposits. This lack of correlation also exists when trained only on days when prices fall.

By comparing the changes in prices and miner deposits, the lack of relationship between prices and miner expenditures can also be visually confirmed. These values rarely move in sync, indicating a low correlation. Although the distribution of daily percentage changes is skewed compared to the distribution of fundamental variables, they are suitable for informal visual analysis.

Conclusion

The impact of Bitcoin miners on the broader market has yet to be thoroughly studied. However, by developing more refined methods to assess the scale of miner activity, we can begin to better understand the role miners play. Given the relatively small scale of miner deposits compared to total trading flow and the lack of correlation between miner trading flow and prices, we find little reason to believe that miners are the cause of Bitcoin price declines.

These metrics have significant room for improvement, particularly in increasing exchange coverage and filtering out addresses that have not participated in mining activities recently. Nevertheless, the flow at the exchange level aligns with conventional wisdom, that Binance and Huobi are the preferred exchanges for miners, informally validating the effectiveness of these metrics.