race-prediction

name: race-prediction description: Race time estimation using VDOT, Riegel formula, training-based methods, and prediction tracking

Race Prediction

Estimation Methods

0. Selecting Input Data (Critical)

Before applying any prediction method, you must select the right input data. The best_efforts tool returns lap data for each effort. Analyze the lap structure to determine effort quality.

Data priority for VDOT / Riegel formulas:

1. Official PRs (from manage_personal_records) — chip-timed race results, highest reliability
2. Genuine max efforts — where lap analysis shows dedicated race or time trial structure (warmup → even/negative splits at high HR → cooldown)
3. Training-based estimation — derive from threshold pace, long run pace (see methods below)

DO NOT use for VDOT / Riegel:

• Efforts embedded in a much longer run (e.g., a half marathon split from a 31km long run) — the athlete was NOT racing, they were training at sub-max effort
• Efforts where laps show variable pacing or a fade — not a controlled race effort
• When lap data is unavailable and the effort context is uncertain

When only training data is available (no verified race efforts), skip VDOT/Riegel entirely and use the training-based methods below (threshold pace, long run pace extrapolation).

1. From Recent Race Results (Most Accurate)

Use equivalence tables to predict other distances from a known race time.

Riegel Formula: T2 = T1 * (D2/D1)^1.06

• More accurate for similar distances
• Overestimates for much longer distances (fatigue factor)

Common Conversions:

Examples:

• 20:00 5K -> ~3:16-3:24 Marathon
• 45:00 10K -> ~3:27-3:36 Marathon
• 1:35 Half -> ~3:18-3:25 Marathon

2. From Training Data (When No Recent Races)

From Threshold Pace:

• Threshold pace (tempo run pace) ~ half marathon race pace
• Marathon pace ~ threshold + 15-25s/km
• Use: recent tempo/threshold workout paces

From Long Run Pace:

• Long run easy pace is typically 1:00-1:30/km slower than marathon pace
• If long runs averaging 6:00/km -> marathon pace ~4:45-5:00/km

From Easy Run Pace:

• Easy pace is typically 1:30-2:00/km slower than threshold
• Less reliable but useful as a sanity check

3. VDOT-Based Estimation

VDOT is a single number representing running fitness. Key reference points:

Confidence Levels

High Confidence

• Based on a race within last 6 weeks at a similar distance
• Training data is consistent and extensive (8+ weeks)
• Athlete has run the predicted distance before

Medium Confidence

• Based on training paces only (no recent race)
• Race was >6 weeks ago
• Predicting a distance the athlete hasn't raced

Low Confidence

• Limited training data (<4 weeks)
• Predicting from a very different distance (5K -> ultra)
• Significant fitness changes expected before race
• First-time at the distance

Environmental Adjustments

Heat

• +5-10s/km for temperatures above 15C
• +15-30s/km for temperatures above 25C
• Humidity amplifies the effect

Altitude

• +3-5% time for every 1000m above sea level
• Acclimatization takes 2-3 weeks

Terrain

• Trail races: add 20-50% to road time depending on technicality
• Hilly courses: roughly +1s/km per 10m elevation gain per km

Wind

• Headwind costs more than tailwind saves (net negative for out-and-back)
• Strong headwind can add 10-30s/km

Tracking Prediction Evolution

When saving predictions:

1. Always record the basis (what data drove the estimate)
2. Record confidence level
3. Use save_race_prediction to persist to SQLite
4. Use get_prediction_history to show trends
5. Compare predictions over time: are they improving, plateauing, or declining?

Prediction trends inform coaching:

• Improving: Training is working, stay the course
• Plateauing: May need stimulus change (new workout types, more volume)
• Declining: Check for overtraining, life stress, injury