foxhunt/ml/tests/ewma_thresholds_test.rs

//! EWMA (Exponentially Weighted Moving Average) Threshold Tests
//!
//! Test Suite for adaptive threshold calculation using EWMA.
//! Tests cover initialization, value updates, span parameter effects,
//! and edge cases like volatility spikes.

use approx::assert_relative_eq;
use ml::features::ewma::{AdaptiveThreshold, EWMACalculator};

#[cfg(test)]
mod ewma_basic_tests {
    use super::*;

    #[test]
    fn test_ewma_initialization() {
        let calculator = EWMACalculator::new(100);

        // Verify alpha calculation: α = 2 / (span + 1)
        assert_relative_eq!(calculator.alpha, 2.0 / 101.0, epsilon = 1e-10);

        // Initial state should be None
        assert!(calculator.ewma.is_none());
        assert!(calculator.current().is_none());
    }

    #[test]
    fn test_ewma_first_value() {
        let mut calculator = EWMACalculator::new(100);

        // First value should initialize EWMA to that value
        let first_value = 100.0;
        let result = calculator.update(first_value);

        assert_relative_eq!(result, first_value, epsilon = 1e-10);
        assert_relative_eq!(calculator.current().unwrap(), first_value, epsilon = 1e-10);
    }

    #[test]
    fn test_ewma_constant_values() {
        let mut calculator = EWMACalculator::new(100);
        let constant_value = 50.0;

        // Update with constant value multiple times
        for _ in 0..10 {
            calculator.update(constant_value);
        }

        // EWMA should converge to constant value
        assert_relative_eq!(calculator.current().unwrap(), constant_value, epsilon = 1e-6);
    }

    #[test]
    fn test_ewma_span_parameter() {
        // Test different span values
        let spans = vec![10, 50, 100, 200];
        let values = vec![100.0, 110.0, 120.0, 130.0, 140.0];

        for span in spans {
            let mut calculator = EWMACalculator::new(span);
            let expected_alpha = 2.0 / (span as f64 + 1.0);

            assert_relative_eq!(calculator.alpha, expected_alpha, epsilon = 1e-10);

            // Smaller span = more responsive = higher alpha
            // Update with increasing values
            for value in &values {
                calculator.update(*value);
            }

            // Verify EWMA is computed
            assert!(calculator.current().is_some());
        }
    }
}

#[cfg(test)]
mod ewma_computation_tests {
    use super::*;

    #[test]
    fn test_ewma_formula() {
        let span = 10;
        let alpha = 2.0 / (span as f64 + 1.0); // = 2/11 ≈ 0.1818
        let mut calculator = EWMACalculator::new(span);

        // First value
        let v1 = 100.0;
        let ewma1 = calculator.update(v1);
        assert_relative_eq!(ewma1, v1, epsilon = 1e-10);

        // Second value: EWMA = α * v2 + (1 - α) * EWMA_prev
        let v2 = 110.0;
        let expected_ewma2 = alpha * v2 + (1.0 - alpha) * ewma1;
        let ewma2 = calculator.update(v2);
        assert_relative_eq!(ewma2, expected_ewma2, epsilon = 1e-10);

        // Third value
        let v3 = 105.0;
        let expected_ewma3 = alpha * v3 + (1.0 - alpha) * ewma2;
        let ewma3 = calculator.update(v3);
        assert_relative_eq!(ewma3, expected_ewma3, epsilon = 1e-10);
    }

    #[test]
    fn test_ewma_trend_tracking() {
        let mut calculator = EWMACalculator::new(20);

        // Upward trend
        let upward_values: Vec<f64> = (100..120).map(|x| x as f64).collect();
        let mut last_ewma = 0.0;

        for value in upward_values {
            let ewma = calculator.update(value);
            if last_ewma > 0.0 {
                // EWMA should increase with upward trend
                assert!(ewma > last_ewma, "EWMA should track upward trend");
            }
            last_ewma = ewma;
        }
    }

    #[test]
    fn test_ewma_mean_reversion() {
        let mut calculator = EWMACalculator::new(50);

        // Initialize at 100
        calculator.update(100.0);

        // Spike to 150
        calculator.update(150.0);
        let spike_ewma = calculator.current().unwrap();

        // Revert to 100
        for _ in 0..20 {
            calculator.update(100.0);
        }

        let reverted_ewma = calculator.current().unwrap();

        // EWMA should decrease back towards 100
        assert!(reverted_ewma < spike_ewma);
        assert!(reverted_ewma > 100.0); // But not fully there yet (50 span is slow)
    }
}

#[cfg(test)]
mod ewma_threshold_adaptation_tests {
    use super::*;

    #[test]
    fn test_adaptive_threshold_normal_volatility() {
        let mut calculator = EWMACalculator::new(100);

        // Simulate normal market conditions (low volatility)
        let base_value = 1000.0;
        let volatility = 5.0; // ±0.5%

        for i in 0..50 {
            let noise = (i as f64 * 0.1).sin() * volatility;
            calculator.update(base_value + noise);
        }

        let ewma = calculator.current().unwrap();

        // EWMA should be close to base value
        assert!((ewma - base_value).abs() < volatility * 2.0);
    }

    #[test]
    fn test_adaptive_threshold_high_volatility() {
        let mut calculator = EWMACalculator::new(100);

        // Simulate high volatility market
        let values = vec![
            1000.0, 1050.0, 980.0, 1020.0, 950.0,
            1030.0, 970.0, 1040.0, 990.0, 1010.0,
        ];

        let mut ewma_values = Vec::new();
        for value in values {
            ewma_values.push(calculator.update(value));
        }

        // EWMA should smooth out volatility
        let ewma_volatility = calculate_std_dev(&ewma_values);
        let raw_volatility = calculate_std_dev(&vec![
            1000.0, 1050.0, 980.0, 1020.0, 950.0,
            1030.0, 970.0, 1040.0, 990.0, 1010.0,
        ]);

        // EWMA volatility should be lower than raw volatility
        assert!(ewma_volatility < raw_volatility);
    }

    #[test]
    fn test_adaptive_threshold_regime_change() {
        let mut calculator = EWMACalculator::new(50);

        // Low volatility regime (100-105)
        for _ in 0..20 {
            calculator.update(100.0 + (rand::random::<f64>() * 5.0));
        }
        let low_vol_ewma = calculator.current().unwrap();

        // Regime change to high volatility (100-120)
        for _ in 0..20 {
            calculator.update(100.0 + (rand::random::<f64>() * 20.0));
        }
        let high_vol_ewma = calculator.current().unwrap();

        // EWMA should adapt to new regime
        assert!((high_vol_ewma - low_vol_ewma).abs() > 0.0);
    }

    fn calculate_std_dev(values: &[f64]) -> f64 {
        let mean = values.iter().sum::<f64>() / values.len() as f64;
        let variance = values.iter()
            .map(|x| (x - mean).powi(2))
            .sum::<f64>() / values.len() as f64;
        variance.sqrt()
    }
}

#[cfg(test)]
mod ewma_edge_cases_tests {
    use super::*;

    #[test]
    fn test_ewma_zero_values() {
        let mut calculator = EWMACalculator::new(100);

        // Initialize with zero
        calculator.update(0.0);
        assert_relative_eq!(calculator.current().unwrap(), 0.0, epsilon = 1e-10);

        // Add more zeros
        for _ in 0..10 {
            calculator.update(0.0);
        }
        assert_relative_eq!(calculator.current().unwrap(), 0.0, epsilon = 1e-10);
    }

    #[test]
    fn test_ewma_negative_values() {
        let mut calculator = EWMACalculator::new(100);

        // Use negative values (e.g., returns)
        calculator.update(-5.0);
        calculator.update(-3.0);
        calculator.update(-7.0);

        let ewma = calculator.current().unwrap();
        assert!(ewma < 0.0, "EWMA should handle negative values");
    }

    #[test]
    fn test_ewma_large_values() {
        let mut calculator = EWMACalculator::new(100);

        // Use large values (e.g., Bitcoin prices)
        let large_values = vec![50000.0, 51000.0, 49000.0, 52000.0];

        for value in large_values {
            calculator.update(value);
        }

        let ewma = calculator.current().unwrap();
        assert!(ewma > 0.0 && ewma < 100000.0);
    }

    #[test]
    fn test_ewma_extreme_volatility_spike() {
        let mut calculator = EWMACalculator::new(100);

        // Normal values
        for _ in 0..50 {
            calculator.update(100.0);
        }
        let normal_ewma = calculator.current().unwrap();

        // Extreme spike (10x)
        calculator.update(1000.0);
        let spike_ewma = calculator.current().unwrap();

        // EWMA should increase but be dampened by history
        assert!(spike_ewma > normal_ewma);
        assert!(spike_ewma < 1000.0); // Not fully track the spike

        // Should be closer to previous EWMA due to long span
        let expected_spike_ewma = (2.0 / 101.0) * 1000.0 + (99.0 / 101.0) * normal_ewma;
        assert_relative_eq!(spike_ewma, expected_spike_ewma, epsilon = 1e-6);
    }

    #[test]
    fn test_ewma_reset() {
        let mut calculator = EWMACalculator::new(100);

        // Build up history
        for i in 0..20 {
            calculator.update(100.0 + i as f64);
        }
        assert!(calculator.current().is_some());

        // Reset
        calculator.reset();
        assert!(calculator.current().is_none());

        // Should reinitialize on next update
        calculator.update(50.0);
        assert_relative_eq!(calculator.current().unwrap(), 50.0, epsilon = 1e-10);
    }

    #[test]
    fn test_ewma_very_small_span() {
        let mut calculator = EWMACalculator::new(2);

        // Very small span means high alpha (2/3 ≈ 0.667)
        let alpha = 2.0 / 3.0;
        assert_relative_eq!(calculator.alpha, alpha, epsilon = 1e-10);

        // Should be very responsive
        calculator.update(100.0);
        calculator.update(200.0);

        let expected = alpha * 200.0 + (1.0 - alpha) * 100.0;
        assert_relative_eq!(calculator.current().unwrap(), expected, epsilon = 1e-10);
    }

    #[test]
    fn test_ewma_very_large_span() {
        let mut calculator = EWMACalculator::new(1000);

        // Very large span means low alpha (2/1001 ≈ 0.002)
        let alpha = 2.0 / 1001.0;
        assert_relative_eq!(calculator.alpha, alpha, epsilon = 1e-10);

        // Should be very slow to respond
        calculator.update(100.0);
        calculator.update(200.0);

        let expected = alpha * 200.0 + (1.0 - alpha) * 100.0;
        assert_relative_eq!(calculator.current().unwrap(), expected, epsilon = 1e-10);

        // Should be close to first value due to low alpha
        assert!((calculator.current().unwrap() - 100.0).abs() < 5.0);
    }
}

#[cfg(test)]
mod ewma_span_comparison_tests {
    use super::*;

    #[test]
    fn test_span_responsiveness_comparison() {
        let spans = vec![10, 50, 100, 200];
        let values = vec![100.0, 150.0]; // Sudden jump

        let mut final_ewmas = Vec::new();

        for span in &spans {
            let mut calculator = EWMACalculator::new(*span);
            for value in &values {
                calculator.update(*value);
            }
            final_ewmas.push(calculator.current().unwrap());
        }

        // Smaller span should be more responsive (closer to 150.0)
        for i in 0..final_ewmas.len() - 1 {
            assert!(
                (final_ewmas[i] - 150.0).abs() < (final_ewmas[i + 1] - 150.0).abs(),
                "Smaller span should be more responsive to changes"
            );
        }
    }

    #[test]
    fn test_optimal_span_selection() {
        // Test different spans on realistic data
        let market_data = generate_realistic_market_data(100);
        let spans = vec![10, 20, 50, 100];

        for span in spans {
            let mut calculator = EWMACalculator::new(span);

            for value in &market_data {
                calculator.update(*value);
            }

            // All spans should produce valid EWMA
            let ewma = calculator.current().unwrap();
            assert!(ewma > 0.0);
            assert!(ewma.is_finite());
        }
    }

    fn generate_realistic_market_data(count: usize) -> Vec<f64> {
        let mut data = Vec::new();
        let mut price = 1000.0;

        for i in 0..count {
            // Add trend + noise
            let trend = 0.1 * (i as f64 / 10.0);
            let noise = (i as f64 * 0.3).sin() * 5.0;
            price += trend + noise;
            data.push(price);
        }

        data
    }
}