What can you actually run? Your training data knows better than any calculator.

Why Generic Calculators Fail

The Fixed Formula Problem

Most race calculators use the same approach:

1. Enter a recent race result (say, 22:00 5K)
2. Apply a formula (often based on VDOT tables)
3. Receive predictions for other distances (47:00 10K, 3:25 marathon)

The assumption: All runners have identical relationships between race distances.

The reality: Individual variation is enormous.

Individual Differences

Runner A and Runner B both run 22:00 5Ks, but:

Runner A (Speed-dominant):

• Excellent anaerobic capacity
• Limited endurance base
• 10K: 46:30 (better than predicted)
• Marathon: 3:40 (much worse than predicted)

Runner B (Endurance-dominant):

• Moderate speed
• Exceptional aerobic endurance
• 10K: 48:00 (slightly worse than predicted)
• Marathon: 3:15 (much better than predicted)

Generic calculators give both runners the same predictions. Both predictions are wrong.

Training Context Matters

A 22:00 5K after:

• 12 weeks of structured training
• Indicates fitness on an upward trajectory

A 22:00 5K after:

• 6 months of consistent 40-mile weeks
• Indicates current fitness plateau

Same race result, different implications for future performance.

Generic calculators can't distinguish these scenarios.

What Data Actually Predicts Performance

Training Volume and Consistency

Volume patterns:

• Weekly mileage over recent months
• Long run distances
• Consistency of training (gaps vs. continuous)

What it reveals:

• Endurance base development
• Durability for longer races
• General training load tolerance

Workout Performance

Quality session data:

• Tempo run paces and heart rate
• Interval performances
• Long run pace and drift

What it reveals:

• Current threshold fitness
• VO2max approximation
• Race-specific readiness

Training Response Patterns

How you respond to training:

• Fitness improvements over time
• Recovery between hard sessions
• Adaptation to increased load

What it reveals:

• Your personal improvement trajectory
• Where you are on the fitness curve
• How much more improvement is likely

Physiological Indicators

Heart rate and recovery data:

• Heart rate at various paces over time
• Recovery HR patterns
• HRV trends

What it reveals:

• Aerobic fitness changes
• Fatigue accumulation
• Current recovery state

How AI Generates Predictions

Multi-Factor Analysis

AI predictions incorporate multiple data streams:

Historical baseline:

• Your past race performances (if any)
• Training history and patterns
• Seasonal performance variations

Current fitness estimate:

• Recent workout performances
• Training load and intensity
• Recovery metrics

Training trajectory:

• Are you improving, plateauing, or declining?
• What does the fitness curve look like?
• How much more adaptation is expected?

Race-specific factors:

• Distance-specific preparation
• Course and condition adjustments
• Taper and peak timing

The Prediction Model

Step 1: Fitness estimation Based on recent training data, estimate current race-relevant fitness.

Step 2: Trajectory projection Based on training plan and response patterns, project fitness at race day.

Step 3: Distance adjustment Apply your personal distance-to-distance relationships (not generic formulas).

Step 4: Condition modification Adjust for expected race conditions (weather, altitude, course).

Step 5: Confidence calculation Determine prediction uncertainty based on data quality and quantity.

Confidence Ranges

Good predictions include uncertainty:

High confidence: +-1-2% "Based on your data, expect 3:22-3:28 marathon"

Medium confidence: +-3-5% "Limited data suggests 3:18-3:35 marathon range"

Low confidence: +-5-10% "Insufficient data for confident prediction; rough estimate 3:10-3:45"

Understanding confidence prevents over-reliance on point estimates.

Predictions vs. Goals

Prediction: What Data Suggests

Based purely on your training data and patterns, what time is most likely?

This is descriptive: Here's what's probable given current information.

Goal: What You're Aiming For

Based on your aspirations and circumstances, what time do you want to run?

This can be prescriptive: Here's what you're targeting.

Using Both

Realistic goal setting: If prediction says 3:25-3:32 and your goal is sub-3:00, there's a gap. Either:

• Adjust goal to match realistic prediction
• Understand you're reaching for a stretch goal
• Modify training to close the gap (if time allows)

Pacing strategy: Start the race based on prediction, not aspirational goal. Racing beyond predicted fitness usually produces positive splits and disappointment.

Mental preparation: Knowing a realistic prediction helps manage expectations and reduce race-day anxiety.

Prediction Dynamics

Early Cycle Predictions

8-12 weeks from race:

• Prediction based on current fitness + expected improvement
• Higher uncertainty due to more time for variation
• Useful for general goal setting

Mid-Cycle Predictions

4-8 weeks from race:

• Training response data accumulating
• Prediction becoming more specific
• Good time to finalize goal pace

Pre-Race Predictions

1-3 weeks from race:

• Maximum data available
• Prediction most accurate
• Final pacing strategy decisions

Prediction Updating

Predictions should change as training progresses:

After strong training block: Prediction improves as fitness gains show in data.

After illness or interruption: Prediction adjusts down to reflect lost training.

After exceptional workout: Prediction may adjust up if performance exceeds expectations.

Static predictions set 16 weeks out ignore everything you learn during training.

Using Predictions Effectively

For Goal Setting

Process:

1. Get initial prediction at training start
2. Set goal aligned with prediction (or deliberately above/below)
3. Update prediction as training progresses
4. Adjust goal if prediction changes significantly

Avoid: Setting a goal at week 1 and ignoring prediction updates.

For Pacing Strategy

Race pacing from prediction:

• First half: 2-5 seconds/mile slower than prediction pace
• Second half: Build toward or slightly faster than prediction pace
• Allows for conservative start, strong finish

Why not goal pace from the start: If goal exceeds prediction, starting at goal pace often leads to blowing up late in the race.

For Race Selection

Prediction helps choose appropriate races:

• Want Boston Qualifier? Does prediction suggest it's realistic?
• Targeting specific time? Is it within prediction range?
• Choosing between distances? Which prediction is strongest?

For Mental Preparation

Knowing your range reduces anxiety:

• Expectation set appropriately
• Success defined realistically
• Focus on execution rather than hoping for miracles

Improving Prediction Accuracy

Provide More Data

Better data = better predictions:

• Log all runs (not just impressive ones)
• Include heart rate data
• Rate perceived effort consistently
• Note sleep and life stress

Include Varied Workouts

Predictions improve with diverse training data:

• Tempo runs reveal threshold fitness
• Intervals reveal VO2max
• Long runs reveal endurance
• Easy runs reveal aerobic base

Missing any of these reduces prediction accuracy.

Allow Calibration Time

New users: Limited data means predictions rely heavily on population models. After 4-8 weeks of consistent training, individual patterns emerge and predictions improve.

Validate with Racing

Race results calibrate predictions:

• Tune-up races validate or adjust predictions
• Even time trials provide useful data
• Prediction accuracy improves with each race

When Predictions Are Wrong

Prediction Too Slow (You Beat It)

Possible reasons:

• Training data underrepresented your fitness
• Race-day conditions were better than expected
• Psychological factors (competition, motivation)
• Recent fitness jump not yet reflected in data

Response: Enjoy the PR. Data will adjust.

Prediction Too Fast (You Missed It)

Possible reasons:

• Training data overrepresented fitness (too much hero data)
• Race-day conditions were worse than expected
• Pacing or fueling errors
• Undisclosed fatigue or illness

Response: Analyze what happened. Was it prediction error or execution error?

Chronic Mismatch

If predictions are consistently wrong:

• Review data quality (is tracking accurate?)
• Consider missing factors (life stress, sleep, etc.)
• May have unusual physiology that differs from population patterns

The AI should learn from errors over time.

Generic calculators treat you as a statistical average. Data-driven predictions treat you as an individual with unique patterns, strengths, and trajectory. The more training data you provide, the more accurate your predictions become—giving you the information to set appropriate goals, pace wisely, and race confidently.

See your personalized predictions on your dashboard.