{
  "date": "2026-09-14",
  "explanation": "This forecast uses current MLS standings, each team’s season-long scoring and defending record, the remaining schedule, and an assumed home-field advantage. It simulates every remaining match involving a Western Conference team 50,000 times, then counts how often Austin finishes in each position. Teams share the same result when they play each other. The percentages are estimates from a simple statistical model—not guarantees or calibrated betting odds.",
  "fixes": [
    "Replaced 29 independent placeholder team appearances with the complete 80-match Western run-in.",
    "Corrected Vancouver's GP and restored its tenth remaining fixture.",
    "Used seed 20260914 and published executable code and input data for reproducibility.",
    "Removed subjective club strength overrides; baseline derives attack/defense rates from season goals.",
    "Replaced fixed 25% draws with draws arising from simulated scorelines.",
    "Replaced strength-based tiebreaking with points, wins, GD, GF in official priority order; explicitly bounded remaining ties.",
    "Placed playoff cutoff below 9th and labeled top-9 probability correctly.",
    "Headline gives median points and median Western position; identifies them as separate marginal medians.",
    "Removed unsupported form and home-game claims; computes remaining home counts directly from fixtures.",
    "Shows model sensitivity and identifies selected parameter choices as uncalibrated.",
    "Preserved model.js, data.json, report.json and update instructions as deployable public source.",
    "Corrected reach-42/45 versus qualifying probability distinction.",
    "v2.2: added a Dixon-Coles low-score correction calibrated to the season draw rate; independent Poissons produced ~85 draws where the 2026 season had 95 in 372 matches.",
    "v2.3: added every club’s finishing-position heatmap, GD/GF and remaining expected points in the standings, a weekly rooting guide (conditional odds for every fixture in the first week of remaining play), the points curve, and deterministic magic/tragic numbers. fixed results now accept any fixture from the home side’s view."
  ],
  "limitations": [
    "Not backtested or calibrated. Not a trained AI predictor, bookmaker model, or xG model.",
    "Eight pseudo-games of league-average scoring shrink extreme rates; this and the log home effect 0.13 are assumed parameters.",
    "Poisson goal counts are corrected only for the low-score draw deficit (Dixon-Coles rho, refitted each ingest); they still omit within-match correlation beyond that, lineups, injuries, travel, schedule-adjusted historical strength, and roster changes.",
    "Strengths stay fixed during simulations. East-vs-East fixtures cannot alter them.",
    "After ties on points, wins, GD and GF, the model uses a seeded random order instead of historical head-to-head, disciplinary and venue tiebreakers. Austin boundary bounds are published.",
    "Sensitivity runs change one assumption at a time and are not confidence intervals.",
    "Data is an explicitly dated snapshot, not a live feed.",
    "Scheduled site updates are configured separately; unattended publishing has not yet been demonstrated."
  ],
  "tests": [
    "Identical seeds reproduce every result",
    "All tables obey probability totals",
    "One shared score conserves goals and points",
    "Wins outrank GD; GD outranks GF",
    "Missing and duplicate fixtures fail validation",
    "42/45-point records sum correctly",
    "All Austin wins/losses and home-away conditional result mapping",
    "All 15 position distributions sum to 100%"
  ],
  "tieAudit": {
    "austinAnyTiePct": 0.034,
    "top9LowerPct": 10.58,
    "top9UpperPct": 10.58
  },
  "version": "2.3.0",
  "latestRefresh": {
    "asOf": "2026-10-01",
    "iterations": 50000,
    "seed": 20261001,
    "validation": "passed",
    "tieAudit": {
      "austinAnyTiePct": 0.042,
      "top9LowerPct": 7.362,
      "top9UpperPct": 7.364
    }
  }
}