A reproducible replacement is now published
The replacement is at 0xinsider.com/research/nba-betting-polymarket-sports: one defined set of Polymarket NBA large trades, one retained row set, and every figure traced to a committed, read-only query. It answers a narrower question than this page tried to, favorite-longshot pricing instead of a general activity census, and it does not restate or rescue any number below.
This page stays published so links that already point at it reach the correction instead of a silently swapped article.
Correction: the previous NBA dataset is withdrawn
The previous page claimed 16,382 large trades while one of its own chart datasets summed to 16,377. No committed provider capture or reproducible query showed which row set, time window, threshold, or pagination result produced either number. We removed the volume, market, trader, timing, grade, consensus, strategy, and profitability claims, and every chart and stat block that depended on them.
This retracts the evidence, not a conclusion about NBA markets. The URL stays live so links to the earlier analysis reach this correction instead of a dead page.
What a reproducible NBA analysis requires
A replacement has to define the provider and market universe, the large-trade threshold, event and market identity rules, the exact UTC window, pagination and retry behavior, duplicate handling, missing-data policy, price and volume units, and the outcome clock. A wallet-level analysis also has to say how wallet identity, portfolio P&L, and market-row diagnostics are joined.
Every headline, table, chart, and FAQ has to come from the same retained row set. Totals have to reconcile before publication, and the article has to separate provider facts from stored data and illustrations.
How to use live trade records
The large-trade feed shows the trades we ingest, filtered by our rules. It is not a full ledger or a performance study. A record is a lead for checking the market, price, size, side, timestamp, and the wallet when the provider names one.
Trade size alone does not show skill, consensus, cause, or a profitable strategy. Save the source response and retrieval time before making a factual claim, and mark provider failures or missing records as unavailable instead of dropping them.
