MoPatterns

"The Detective" — "I find what you can't see"

Status: ❌ Future

MoPatterns will use machine learning to detect behavioral patterns and predict outcomes.

---

Purpose

• Detect patterns in training behavior
• Predict future performance
• Identify optimal training times
• Correlate recovery with performance
• Personalize recommendations

---

Pattern Types

Training Patterns

Pattern	Example
Best training days	"You perform best on Tuesday/Thursday"
Optimal session length	"Sessions under 60 min have better RPE"
Rest day impact	"Performance drops after 3+ rest days"
Volume tolerance	"You handle 20 sets/session well"

Recovery Patterns

Pattern	Example
Sleep-performance correlation	"Below 6 hrs sleep = +0.5 avg RPE"
Energy cycles	"Energy dips every Thursday"
Soreness recovery	"Legs need 72 hrs between sessions"

Progression Patterns

Pattern	Example
PR frequency	"You set PRs every 2-3 weeks"
Plateau indicators	"RPE creep precedes plateaus by 2 weeks"
Deload timing	"Performance peaks 1 week after deload"

---

Data Model

interface PatternInsight {
  id: string;
  type: PatternType;
  confidence: number;        // 0-100%
  description: string;
  recommendation: string;
  dataPoints: number;        // How much data supports this
  firstDetected: Date;
  lastConfirmed: Date;
}

interface Prediction {
  type: PredictionType;
  target: string;            // e.g., "bench press"
  prediction: number;        // predicted value
  confidence: number;
  timeframe: string;         // "2 weeks"
  basis: string;             // explanation
}

type PatternType =
  | 'training_timing'
  | 'recovery_correlation'
  | 'volume_tolerance'
  | 'progression_rate'
  | 'fatigue_pattern';

type PredictionType =
  | 'next_pr'
  | 'plateau_risk'
  | 'deload_needed'
  | 'optimal_weight';

---

Machine Learning Approach

Data Features

interface TrainingFeatures {
  // Session features
  dayOfWeek: number;
  timeOfDay: number;
  sessionDuration: number;
  restDaysBefore: number;

  // Recovery features
  sleepHours: number;
  sleepQuality: number;
  energyLevel: number;
  soreness: number;

  // Historical features
  recentVolume: number;
  recentAvgRPE: number;
  daysSinceDeload: number;
}

interface PerformanceTarget {
  avgRPE: number;
  volumeCompleted: number;
  prAchieved: boolean;
}

Models (Future)

1. Performance Predictor - Predict session quality from recovery data
2. Plateau Detector - Predict when plateau will occur
3. Deload Recommender - Optimal deload timing
4. PR Predictor - When user is likely to hit a PR

---

Implementation Tasks

• Collect sufficient training data
• Design feature extraction
• Train initial models
• Build pattern detection logic
• Create insight delivery system
• Add confidence scoring
• Build explanation generation