287 of 1,234 Pick6 capture pairs showed a candidate target difference

Question 1: does one market id ever change target?

Sample: every pickable_market_id in pick6_capture_snapshots before the freeze. Numerator: ids seen with more than one target_value.

Result: 0 of 13,880 market ids. A market id binds one target. A consumer watching one market row for a target change will never see one, which is why this study measures moves at the prop level.

Question 2: what do recorded change rows contain?

Sample: every pick6_changes row before the freeze. Classes are exclusive: line, then a same-side multiplier difference, then an is_paused flip, otherwise metadata.

Selection ids: of 2,611 same-side selection pairs compared, 2,279 had a new id with a new multiplier, 0 had a new id with the same multiplier, and 0 had a new multiplier under the same id. Classifier check: the SQL classes matched the application's change mapper on 2,459 of 2,459 rows.

Question 3: how often does the candidate target change?

Sample: single-athlete pickables with a two-sided (More and Less) row in at least two captures before the freeze. Unit: two consecutive captures of one pickable. Numerator: target_value differs.

Population: 996 props had a two-sided row. 584 were seen in one capture only. After eligibility and ambiguity exclusions, 412 props contribute candidate transitions. There were 2,230 prop-capture pairs, including 0 with multiple two-sided rows. Quote shape does not identify the provider primary market. Ambiguous captures are excluded from candidate transitions.

By the live flag on the later capture

287 of 287 moves came with a new market id. 5 pairs had a new market id with the same target. The live split is descriptive: leagues contribute very differently to each column, so it does not isolate an effect of live play.

First observation against last

Question 4: attributed sportsbook mainlines

Sample: every history row on source espn-draftkings before the freeze. Unit: consecutive stored quotes of one event, market and side. Numerators: line differs, American price differs. Source: espn-draftkings, observed through ESPN.

Each side is counted separately, so one spread move usually appears twice. Most quotes have no successor yet, which is why pairs are far fewer than quotes.

Limits

• One retained day of capture. Nothing here describes a season, a typical week or any league's normal behaviour.
• Changes between captures are unobserved. Candidate target switches are not verified primary-line moves, so these counts are neither complete nor a lower bound on primary moves. Captures with more than one two-sided row for a pickable are excluded from candidate transitions.
• Two-sided rows are candidate observations only. The retained model does not preserve the provider defaultPickableMarketId, so switching between these rows cannot establish a primary-line move, a settlement change or an independently observed price change.
• Selected public categories only, not a full board. A pickable's capture frequency depends on the collector's rotation.
• A multiplier is not decimal odds, a probability or an entry payout. No result here implies an edge or predicts an outcome.
• Sportsbook rows are ESPN-attributed DraftKings mainlines, observed through ESPN, not direct DraftKings access.

Disclosure: The tables' structure was inspected before the definition was fixed, because the two-sided row and the market id behaviour had to be understood to define a move. No threshold or exclusion was changed after the final run.

Reproduce it

Every statement is in the SQL file, with the freeze clock as its parameter. Freeze clock for this edition: 2026-11-03T20:00:00.000Z. The JSON above is free. The rows behind it are served by the API, with plans on pricing.