Prospective Bitcoin transaction sample, October 2024 - August 2026, with raw API responses

456,292 Bitcoin transactions sampled from 18,400 blocks over 23 months (October 2024 to August 2026), together with the 24,526 raw Esplora API responses they were built from. The blocks were mined during that period and therefore existed before the accompanying study's analysis code was frozen on 16 September 2026; what the freeze establishes is that the sample had not been collected or examined when the analysis plan was written and that the block heights and page offsets were drawn from seeds fixed in that plan. It is an out-of-sample test set for models fitted on 2022–2023 data. Two-stage design: 800 block heights drawn without replacement within each month, then one page of up to 25 consecutive transactions per drawn block, starting at a position drawn uniformly within the block. Every planned page was obtained. Each transaction carries a design weight (inverse inclusion probability) which must be used for monthly estimates; unweighted means over-represent transactions from small blocks. Accompanies: Korde, S.D.; Shekokar, N.M. ZSH: Rank-Weighted Profiling of Bitcoin Transactions and a Pre-Specified Evaluation of What the Profiles Capture. Code, analysis plan, deviations log and all result files: https://github.com/sagarkorde/ZSH (branch v2). Public blockchain records only. No personal data were collected and no attempt was made to identify any person. See README.md for the file list, the sampling details and SHA-256 checksums.