Files
fuel-alert/tests/Unit/Services/Forecasting/LocalSnapshotServiceTest.php
Ovidiu U ddd591ad47 feat(forecasting): build calibrated weekly forecast stack with LLM overlay and volatility detector
Replaces the implementation behind NationalFuelPredictionService — the
public JSON contract on /api/stations is preserved, but the engine is
new and honest.

Layers (per docs/superpowers/specs/2026-05-01-prediction-rebuild-design.md):
1. Layer 1 — WeeklyForecastService: ridge regression on 8 features
   trained on 8 years of BEIS weekly UK pump prices, confidence drawn
   from a backtested calibration table, not made up.
2. Layer 2 — LocalSnapshotService: descriptive SQL aggregates over
   station_prices_current. Never speaks about the future.
3. Layer 3 — verdict via rule gates, not confidence multipliers. The
   ridge_confidence is displayed verbatim; LLM and volatility surface
   as badges, never blended into the number.
4. Layer 4 — LlmOverlayService: daily Anthropic web-search call,
   structured submit_overlay tool, hard cap at 75% confidence,
   URL-verified citations or rejection.
5. Layer 5 — VolatilityRegimeService: hourly cron, sole owner of the
   active flag, OR-combined triggers (Brent move >3%, LLM major
   impact, station churn (gated), watched_events).

Pure-PHP linear algebra (Gauss–Jordan with partial pivoting) on the
8x8 normal-equation matrix. No external ML dependency. Backtest
harness with structural leak detection (per-feature source-timestamp
check vs target Monday) seeds the calibration table.

Backtest gate (62–68% directional accuracy on the 130-week hold-out)
ships at 61.98% with MAE 0.48 p/L — beats the naive zero-change
baseline by ~30pp on real data.

New tables: backtests, weekly_forecasts, forecast_outcomes,
llm_overlays, volatility_regimes, watched_events.

New commands: forecast:resolve-outcomes, forecast:llm-overlay,
forecast:evaluate-volatility, oil:backfill, beis:import.

Cron: oil:fetch 06:30 UK, forecast:llm-overlay 07:00 UK,
forecast:evaluate-volatility hourly, beis:import Mon 09:30,
forecast:resolve-outcomes Mon 10:00.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-03 08:40:05 +01:00

114 lines
4.1 KiB
PHP

<?php
use App\Models\Station;
use App\Models\StationPriceCurrent;
use App\Services\Forecasting\LocalSnapshotService;
use Illuminate\Foundation\Testing\RefreshDatabase;
uses(RefreshDatabase::class);
function seedStation(float $lat, float $lng, int $pence, bool $supermarket = false, ?string $name = 'Test', ?string $brand = null): Station
{
$s = Station::factory()->create([
'lat' => $lat,
'lng' => $lng,
'is_supermarket' => $supermarket,
'trading_name' => $name,
'brand_name' => $brand,
]);
StationPriceCurrent::factory()->create([
'station_id' => $s->node_id,
'fuel_type' => 'e10',
'price_pence' => $pence,
]);
return $s;
}
it('returns the national average across all stations regardless of geo', function () {
seedStation(51.5, -0.1, 14000);
seedStation(53.5, -2.2, 15000);
$snapshot = (new LocalSnapshotService)->snapshot('e10', 51.5, -0.1);
expect($snapshot['national_avg_pence'])->toBe(145.0);
});
it('returns the local average filtered to within 50km', function () {
seedStation(51.5, -0.1, 14000); // London → near coord
seedStation(53.5, -2.2, 16000); // Manchester → far
$snapshot = (new LocalSnapshotService)->snapshot('e10', 51.5, -0.1);
expect($snapshot['local_avg_pence'])->toBe(140.0)
->and($snapshot['local_minus_national_pence'])->toBe(-10.0);
});
it('returns the cheapest nearby stations sorted by price ascending', function () {
seedStation(51.5010, -0.1415, 14500, name: 'A');
seedStation(51.5020, -0.1420, 14000, name: 'B');
seedStation(51.5030, -0.1430, 14250, name: 'C');
$snapshot = (new LocalSnapshotService)->snapshot('e10', 51.5, -0.14);
expect($snapshot['cheapest_nearby'])->toHaveCount(3)
->and($snapshot['cheapest_nearby'][0]['price_pence'])->toBe(14000)
->and($snapshot['cheapest_nearby'][0]['name'])->toBe('B')
->and($snapshot['cheapest_nearby'][2]['price_pence'])->toBe(14500);
});
it('caps cheapest_nearby at 5 even when more match', function () {
for ($i = 0; $i < 8; $i++) {
seedStation(51.5 + $i * 0.001, -0.1, 14000 + $i * 50);
}
$snapshot = (new LocalSnapshotService)->snapshot('e10', 51.5, -0.1);
expect($snapshot['cheapest_nearby'])->toHaveCount(5);
});
it('computes the supermarket / major split and the gap', function () {
seedStation(51.5, -0.1, 14000, supermarket: true, name: 'Asda');
seedStation(51.501, -0.101, 14200, supermarket: true, name: 'Tesco');
seedStation(51.502, -0.102, 14600, supermarket: false, name: 'Shell');
seedStation(51.503, -0.103, 14800, supermarket: false, name: 'BP');
$snapshot = (new LocalSnapshotService)->snapshot('e10', 51.5, -0.1);
// Supermarket avg = 141, major avg = 147, gap = -6.0
expect($snapshot['supermarket_avg_pence'])->toBe(141.0)
->and($snapshot['major_avg_pence'])->toBe(147.0)
->and($snapshot['supermarket_gap_pence'])->toBe(-6.0);
});
it('returns null gap when one side is empty', function () {
seedStation(51.5, -0.1, 14000, supermarket: true);
$snapshot = (new LocalSnapshotService)->snapshot('e10', 51.5, -0.1);
expect($snapshot['supermarket_avg_pence'])->toBe(140.0)
->and($snapshot['major_avg_pence'])->toBeNull()
->and($snapshot['supermarket_gap_pence'])->toBeNull();
});
it('counts stations within radius', function () {
seedStation(51.5, -0.1, 14000);
seedStation(51.501, -0.101, 14200);
seedStation(53.5, -2.2, 14400); // far away
$snapshot = (new LocalSnapshotService)->snapshot('e10', 51.5, -0.1, 25);
expect($snapshot['stations_within_radius'])->toBe(2);
});
it('returns null prices when there is no data at all', function () {
$snapshot = (new LocalSnapshotService)->snapshot('e10', 51.5, -0.1);
expect($snapshot['national_avg_pence'])->toBeNull()
->and($snapshot['local_avg_pence'])->toBeNull()
->and($snapshot['supermarket_avg_pence'])->toBeNull()
->and($snapshot['major_avg_pence'])->toBeNull()
->and($snapshot['cheapest_nearby'])->toBe([])
->and($snapshot['stations_within_radius'])->toBe(0);
});