Premier League Predictor: Astro — Chapter 6, Exercise 3
====================================================

TASK
Enter an incorrect result for a fixture with at least two recorded
predictions, confirm both predictions' points_awarded values, then
call PATCH /api/fixtures/{id}/result again with the corrected result
and confirm that every prediction's points_awarded actually changes
to reflect the correction — not just predictions added after the fix.

SOLUTION

Steps to complete the hands-on part:

1. Using fixture 42 from the chapter's own worked example (with the
   user's 2-1, the expert's 1-1, and the ai's 2-0 predictions already
   recorded), first enter a deliberately wrong result:

   PATCH /api/fixtures/42/result
   { "home_score": 1, "away_score": 1 }

   This makes 1-1 the "actual" result, so the response for GET
   /api/fixtures/42 (joined with its predictions) shows:
     - user (predicted 2-1): outcome(2,1) = HOME_WIN, actual outcome
       DRAW -> neither matches -> points_awarded = 0
     - expert (predicted 1-1): exact match -> points_awarded = 40

2. Confirm those two values directly, e.g. via
   SELECT source, predicted_home_score, predicted_away_score,
   points_awarded FROM predictions WHERE fixture_id = 42;
   which should show user at 0 and expert at 40.

3. Now call PATCH /api/fixtures/42/result again with the real,
   corrected result:

   PATCH /api/fixtures/42/result
   { "home_score": 2, "away_score": 1 }

4. Re-run the same SELECT. Both rows should have changed:
     - user (predicted 2-1): exact match against the corrected 2-1 ->
       points_awarded = 40 (up from 0)
     - expert (predicted 1-1): outcome(1,1) = DRAW, corrected actual
       outcome HOME_WIN -> neither matches -> points_awarded = 0
       (down from 40)

Written explanation:

Both rows changed, including the expert's, which was never touched
or re-inserted between the two PATCH calls — it's the exact same row,
scored twice against two different results. This is the direct,
observable consequence of the route's own re-scoring behavior:
because the applyResult transaction always re-reads every prediction
currently on the fixture and recalculates points_awarded for all of
them from scratch on every call, a correction to a wrongly-entered
result automatically corrects every affected prediction's own points
in the same call, rather than leaving the first (wrong) scoring
decision permanently baked into rows that existed before the fix.

WHY THIS WORKS AS AN ANSWER
----------------------------
It walks through a real before-and-after scoring scenario with actual
numbers, shows both predictions' points_awarded values genuinely
flipping (0 to 40, and 40 to 0) between the two PATCH calls, and
correctly attributes the behavior to the route's own full re-scoring
of every existing prediction rather than an incremental update that
would only affect newly added rows.