Clarify dataset description, methods and limitations

README.md

@@ -1,6 +1,6 @@
 ---
 license: odc-by
-pretty_name: "What each Bitcoin mining pool is building on, second by second"
+pretty_name: "Bitcoin mining pool Stratum job observations"
 tags:
   - bitcoin
   - mining
@@ -19,44 +19,32 @@ size_categories:
 
 # Bitcoin mining pool templates
 
-What each mining pool is building on, at the moment it was building it.
-
-Pools issue work to their miners describing the block they are extending, the coinbase they will
-claim, and whether previous work should be abandoned. That work is replaced every few seconds.
-Once a block is found, the templates every pool was building are gone.
+Timestamped Stratum job messages collected directly from Bitcoin mining pool endpoints. The data records changes in the work each endpoint sends to miners, including the previous block hash, coinbase data and clean-jobs flag.
 
 ## Contents
 
-| name | one row is |
+| Table | Record |
 |---|---|
-| `e20_stratum_jobs_direct` | one job from one pool: the block it extends, its nTime, coinbase, merkle branch count and clean-jobs flag |
+| `e20_stratum_jobs_direct` | A job received from a pool endpoint, with its observation time, nTime, coinbase, merkle branch count and clean-jobs flag |
+
+## Using the data
+
+Use `pool` to identify the endpoint and `operator` to group endpoints belonging to the same operator. Compare `observed_ts` across endpoints to study when new jobs arrived. The pool's `nTime` field uses its own clock and update policy, so it should not be used as an arrival timestamp.
 
-## Templates
+The initial configuration covers eleven endpoints across ten operators. Endpoint availability varies between collection windows. Regional endpoints operated by the same pool may distribute different jobs.
 
-Pools building on the same block do not agree, and the disagreement is structural rather than
-cosmetic. In one observation across eleven endpoints on a single block, nTime spread across 30
-seconds, coinbase length ranged from 318 to 1,488 characters, and merkle branch counts split
-between 10 and 13, meaning some pools were assembling materially smaller blocks than others at
-the same moment. Two endpoints run by the same operator differed by 10 seconds.
+## Limitations
 
-`pool` is the endpoint connected to rather than an identity inferred from the data. `operator`
-collapses endpoints belonging to the same operator, and rows should usually be grouped on it.
+- Observations come from one collection location and include connection latency.
+- Merkle branches are stored as a count and first entry. The dataset does not contain the full transaction list for each proposed block.
+- The clean-jobs flag instructs miners to discard earlier jobs. Check the previous block hash when identifying a change of chain tip.
+- Endpoint coverage is a sample of mining pool activity. It does not measure each operator's full mining capacity or global response time.
 
-## Before you build on this
+## Files and access
 
-- One vantage point, roughly 120 peers out of tens of thousands reachable. This is a sample of
-  the network, not the network.
-- nTime is the pool's own clock and pools do not all update it on the same cadence. It indicates
-  when a pool rebuilt its template. For arrival ordering use `observed_ts`, which is consistent
-  across pools.
-- Eleven endpoints across ten operators. Pools running regional endpoints may serve different
-  work elsewhere, so this is a sample of what each operator was building, not a census of it.
-- Merkle branches are stored as a count and first entry. The count with the coinbase identifies a
-  distinct template; the full list is large and mostly redundant.
-- `clean_jobs = true` marks a pool switching blocks. Sorting those by `observed_ts` gives the
-  order in which pools reacted.
+Data is stored as Parquet files under `dataset/YYYY/MM/`, with partitions for collection windows. Each measurement table has a fixed 7-day sample beginning at its configured collection start date. The sample windows in this repository span 2026-08-28 to 2026-09-03. Availability within each window depends on successful collection. The public sample dates remain fixed as additional history accumulates privately. Contact DataForge through the discussions tab to enquire about additional history.
 
-Partitions are parquet, one file per collection window, under `dataset/YYYY/MM/`. Every dataset here carries a FIXED 7-day sample WINDOW starting at its own first day of collection, together spanning 2026-08-28 to 2026-09-03, so you can check schema, coverage and quality before asking for more. It does not advance, so there is nothing to gain by re-downloading it. The full history is held privately, available on request.
+### Load a table
 
 ```python
 from huggingface_hub import snapshot_download
@@ -70,11 +58,8 @@ df = pd.concat(map(pd.read_parquet,
 
 ## Coverage
 
-`e0_run_manifest` lists every collection window with its poll counts and failure counts, and is
-published in full rather than windowed. Gaps between windows are real, cannot be filled in
-afterwards, and nothing here is interpolated.
+`e0_run_manifest` records collection windows, poll counts and failures. It is published in full and may cover dates beyond the fixed data sample. Collection gaps are not interpolated. Use the manifest together with table timestamps and error fields to assess coverage.
 
 ## License and contact
 
-ODC-BY: use it freely, credit "DataForge (dataforge-labs)". Questions and requests for the full
-history via the discussions tab.
+The public sample is published under ODC-BY. Attribute it to "DataForge (dataforge-labs)". For questions about the data or access to additional history, open a discussion in this repository.