Late-entry replay: every later listing of a signal scored against its first report (#115)

[yanivil]

Summary

The question behind #115: the report shows a confirmed row again on every following session (HAL for a week), but every replay fills a signal once, at its first report. This PR makes the backtester score the later listings too, and carries the eleven-year finding.

Code

• walk_forward(..., repeats=[]) records every later CONFIRMED listing of an already-seen signal (first_day, listed_day = sessions since the first report), filled at the next open under the same rules through the shared _fill_outcome; the first-seen rows are unchanged (asserted by a test).
• late_entry_table summarises each day on the list and pairs every traded listing with its own first report (mean difference and its month-block interval); rendered per window as Late entry, in the JSON as late and repeats.
• The backtest workflow keeps backtest.json as an artifact (backtest-<run id>, actions/upload-artifact SHA-pinned) so several runs pool exactly.
• Tests (140 pass), wiki 03 ("Late entry, tested") and 04, changelog.

Finding (runs 34323013559 to 34323039560, tuned, point-in-time membership, bars=500, horizon=60; 1692 first-seen signals, 4915 later listings; full tables on #115)

Day on the list	Listed	Traded	Hit rate	Mean R	95 % CI	Pairs	Same signals on day 1	On day N	Difference	95 % CI
1	1692	1632	35 %	+0.22	[+0.09, +0.36]
2	1140	1111	35 %	+0.28	[+0.13, +0.45]	1090	+0.28	+0.27	−0.01	[−0.07, +0.07]
3	995	952	35 %	+0.22	[+0.08, +0.36]	928	+0.31	+0.22	−0.09	[−0.15, −0.02]
4-5	1541	1479	36 %	+0.19	[+0.05, +0.32]	1440	+0.31	+0.18	−0.12	[−0.18, −0.06]
6-9	1230	1169	36 %	+0.14	[−0.04, +0.32]	1135	+0.26	+0.14	−0.12	[−0.22, −0.01]

• A row on its second day is as good as new; from the third day the same signal pays about 0.1 R less than on day 1, days 4-9 worse in 10 of 11 years; from day 7 no edge remains (+0.12, +0.05, −0.00 on days 7 to 9); a Wolfe on days 6-9 lost money (−0.45 R on 44 traded).
• The quoted entry drifts only +0.3 % to +1.2 %: the cost is the missed first sessions, not the price paid.
• No rule changes. What follows for the report and the site (the day on the list next to the age, the entry line for days 1-2 only) is on Late entry: buying a signal on its Nth day on the list versus at its first report #115 for the owner's decision.

Test plan

• ruff --select E,F,W --line-length 120 clean, 140 tests pass locally
• the workflow's smoke run on push (34323013378) and the eleven yearly runs succeeded, artifact step included
• tests workflow on this PR

🤖 Generated with Claude Code