bibdle/scripts/average-guesses.ts

import Database from 'bun:sqlite';

// Database path - adjust if your database is located elsewhere
const dbPath = process.env.DATABASE_URL || './local.db';
console.log(`Connecting to database: ${dbPath}`);
const db = new Database(dbPath);

// Query all rows from daily_completions
const query = db.query(`
    SELECT date, guess_count
    FROM daily_completions
    ORDER BY date
`);
const rows = query.all() as { date: string; guess_count: number }[];

if (rows.length === 0) {
	console.log('No completions found in the database.');
	db.close();
	process.exit(0);
}

// Group by date and calculate average guesses
const dateStats = new Map<string, { total: number; count: number }>();

for (const row of rows) {
	const existing = dateStats.get(row.date) || { total: 0, count: 0 };
	existing.total += row.guess_count;
	existing.count += 1;
	dateStats.set(row.date, existing);
}

// Display results
console.log('\n=== Average Guesses Per Day ===\n');
console.log('Date          | Avg Guesses | Total Completions');
console.log('--------------|-------------|-------------------');

for (const [date, stats] of dateStats) {
	const avg = (stats.total / stats.count).toFixed(2);
	console.log(`${date.padEnd(14)}| ${avg.padStart(11)}| ${stats.count.toString().padStart(19)}`);
}

// Calculate overall average
const totalGuesses = Array.from(dateStats.values()).reduce((sum, s) => sum + s.total, 0);
const totalCompletions = Array.from(dateStats.values()).reduce((sum, s) => sum + s.count, 0);
const overallAvg = (totalGuesses / totalCompletions).toFixed(2);

console.log('--------------|-------------|-------------------');
console.log(`Overall Average: ${overallAvg} guesses across ${totalCompletions} completions`);

// Calculate correlation between avg_guesses and completions
function calculateCorrelation(data: { avgGuesses: number; completions: number }[]): number {
	const n = data.length;
	if (n < 2) return 0;

	const avgX = data.reduce((sum, d) => sum + d.avgGuesses, 0) / n;
	const avgY = data.reduce((sum, d) => sum + d.completions, 0) / n;

	let numerator = 0;
	let sumXSquared = 0;
	let sumYSquared = 0;

	for (const d of data) {
		const xDiff = d.avgGuesses - avgX;
		const yDiff = d.completions - avgY;
		numerator += xDiff * yDiff;
		sumXSquared += xDiff * xDiff;
		sumYSquared += yDiff * yDiff;
	}

	const denominator = Math.sqrt(sumXSquared * sumYSquared);
	return denominator === 0 ? 0 : numerator / denominator;
}

// Prepare data for correlation analysis
const allData = Array.from(dateStats.entries()).map(([date, stats]) => ({
	date,
	avgGuesses: stats.total / stats.count,
	completions: stats.count
}));

// Split into pre and post marketing periods
const marketingStartDate = '2026-01-08';
const preMarketing = allData.filter(d => d.date < marketingStartDate);
const postMarketing = allData.filter(d => d.date >= marketingStartDate);

console.log('\n=== Correlation Analysis ===\n');

const allCorrelation = calculateCorrelation(allData);
console.log(`Overall correlation (avg_guesses vs completions): ${allCorrelation.toFixed(3)}`);

if (preMarketing.length >= 2) {
	const preCorrelation = calculateCorrelation(preMarketing);
	console.log(`Pre-marketing correlation (before ${marketingStartDate}): ${preCorrelation.toFixed(3)} (n=${preMarketing.length} days)`);
}

if (postMarketing.length >= 2) {
	const postCorrelation = calculateCorrelation(postMarketing);
	console.log(`Post-marketing correlation (${marketingStartDate} onward): ${postCorrelation.toFixed(3)} (n=${postMarketing.length} days)`);
}

console.log('\nInterpretation:');
console.log('  r close to -1: Strong negative correlation (easier verses → more completions)');
console.log('  r close to 0:  No correlation');
console.log('  r close to +1: Strong positive correlation (harder verses → more completions)');

db.close();