H70: course likes held up last

The notebook

In order: what the lab saw, what it asked, what it proposed and why, what the tests said, and what an investigator concluded. Numbers are aggregates on the discovery data unless a step says otherwise.

1. Hypothesised15 Oct, 23:10

   H70: course likes held up last KILLED amber

   because Held up last time out at a course where such runners have beaten the market by 25%+ before. Tracks that favour closers (last year only).

   cp_held_up_last_runs >= 60 AND cp_held_up_last_wins >= 1.25 * cp_held_up_last_exp

   made by import

2. Tested15 Oct, 23:10

   broad: A/E 1.04 on 1,146 bets, by period 1.04. info

   Betfair SP return -5.5%. Every period above 1.00 (spread 0.00). Zone: all. On H70: course likes held up last

   By period

   betsA/E0.71.01.3

   2020-220–

   2023 to a year ago0–

   Last year1,1461.04

3. Tested15 Oct, 23:10

   100+ earlier runners: A/E 1.01 on 632 bets, by period 1.01. info

   Betfair SP return -13.3%. Every period above 1.00 (spread 0.00). Zone: all. On H70: course likes held up last

   By period

   betsA/E0.71.01.3

   2020-220–

   2023 to a year ago0–

   Last year6321.01

4. Tested15 Oct, 23:10

   jumps: A/E 1.01 on 537 bets, by period 1.01. info

   Betfair SP return -21.8%. Every period above 1.00 (spread 0.00). Zone: all. On H70: course likes held up last

   By period

   betsA/E0.71.01.3

   2020-220–

   2023 to a year ago0–

   Last year5371.01

5. Tested15 Oct, 23:10

   Flat: A/E 1.07 on 587 bets, by period 1.07. info

   Betfair SP return +11.3%. Every period above 1.00 (spread 0.00). Zone: all. On H70: course likes held up last

   By period

   betsA/E0.71.01.3

   2020-220–

   2023 to a year ago0–

   Last year5871.07

6. Killed15 Oct, 23:10

   H70: course likes held up last Discovery → KILLEDimport: tier C: rejected, do not test again unless new data changes the idea; best [Flat] AMBER A/E 1.068 on 587 bets (broad RED)