Bitcoin Research Lab Methodology

A brief overview:

• All data is sourced from a local Bitcoin Core Node.
• Data is processed by python and R code.
• Most of the data is analyzed as daily snapshots of the blockchain (chainstates).
• Some data, coinmetrics data being the most important, is analyzed as complete datasets.
• The Bitcoin Research Lab does not cluster UTXOs into entity.  

Limitation and Advantages of the Snapshot Approach
The fact that Bitcoin Research Lab analyses data as daily snapshot has a few implications. First, this implies that UTXOs with age less than one day are not recognized.

Some may consider this a limitation, but there are definitely advantages to this approach. Many intra-day transactions are the result of mixers and internal exchange housekeeping etc. We don't want the rise and fall of mixers to cloud our analysis, do we?

The snapshot approach effectively reduces noise resulting in a clearer picture of economical meaningful activity in the context of market analysis.

Second, due to the snapshot approach some metrics will deviate from those of other onchain providers.  

Most of the time rapid chains of transactions are due to mixers, exchanges etc. Arguably, such transactions do not represent economically meaningful on-chain activity.

The argument for the snapshot approach is therefore the same as above. It provides a cleaner onchain perspective in the context of broad market analysis.

One more note on volume metrics: The Bitcoin Research Lab does not differentiate between a transaction output being sent to a receiver or back to the sender. This is a limitation, but in reality there is no way to be 100% certain of the sender's intent. The analysis will therefore be probabilistic (most of the time).

Address clustering and entities

The Bitcoin Lab is not designed as a forensic tool but to analyze on-chain trends . It is obvious that the ability to track UTXOs to entities, such exchanges, ETFs and even individuals, can give great insight. There exists a few onchain providers (albeit expensive) that offer such data. On the other hand, clustering analysis and associated techniques comes with their own caveats. Briefly, they can be summarized:

• It is computationally very expensive and/or requires manual laborious investigations.
• The analyses gives a probabilistic result, which is inaccurate.
• It is relatively easy to circumvent tracking techniques, if someone really want to.
• As the bitcoin usage changes (e.g. taproot, segwit) over time, so do the efficacy of clustering and tracking techniques. This is potentially dangerous because in market analyses you rely on the metric to be accurate not only for present, but also historical data.

The old RBN app did basic clustering and address monitoring, but this feature is now removed (in fact, the code is still there, just disabled). The reason is these data did not add much value. Instead I found them not clean and susceptible to changes in how the bitcoin network was used.

Is firmly believe history will show that the best onchain data is pure onchain data.