Behind every personalized workout is an algorithm making complex decisions. Here's how it works.

How Workout Prescription Works

The Basic Question

Every day, the algorithm must answer: What workout will produce the best training stimulus for THIS runner TODAY?

This requires considering:

• What training stimulus is most valuable now?
• What is this runner capable of today?
• How does today's workout fit into the larger plan?
• What has this runner done recently?

The answer differs for every runner on every day.

From Art to Algorithm

Traditional coaching: Experienced coach observes athlete, considers context, prescribes based on judgment and experience.

Algorithmic approach: System analyzes data, applies learned patterns, optimizes based on defined objectives.

The goal: Replicate (and potentially improve upon) expert coaching decisions at scale.

The Optimization Framework

Workout prescription optimizes for:

• Training benefit (adaptation stimulus)
• Recovery cost (fatigue generated)
• Goal alignment (preparation for target race)
• Risk management (injury prevention)

Balancing these factors produces the prescription.

Factors in the Algorithm

Current Fitness State

Inputs:

• Recent workout performances
• Heart rate at various paces
• Race results if available
• Fitness trend over weeks

Output: Estimated current fitness level—threshold pace, VO2max pace, easy pace boundaries.

Purpose: Calibrate workout intensity to what you're actually capable of.

Accumulated Fatigue

Inputs:

• Recent training load (acute)
• Historical training load (chronic)
• Recovery metrics (HRV, resting HR)
• Subjective feedback if provided

Output: Current fatigue level and recovery status.

Purpose: Adjust workout difficulty to match recovery state. Don't prescribe hard workout when fatigued.

Training Phase

Inputs:

• Current position in training cycle
• Time until goal race
• Phase objectives (base, build, peak, etc.)

Output: Appropriate workout types and emphasis for current phase.

Purpose: Ensure today's workout serves the larger plan.

Recent Training History

Inputs:

• Last 7-14 days of workouts
• Types completed, intensity, volume
• Recovery between sessions

Output: What training stimulus is missing or overrepresented.

Purpose: Balance training appropriately. Don't do three hard workouts in a row. Don't neglect important stimulus types.

Schedule and Constraints

Inputs:

• Available time for today's workout
• Days until next hard effort
• Upcoming race or event proximity

Output: What's feasible within constraints.

Purpose: Prescribe realistic workouts that fit your life.

From Data to Workout

Step 1: Assess State

Algorithm evaluates:

• Where is fitness currently?
• How recovered are you?
• What phase are we in?
• What's been done recently?

Creates a "readiness snapshot" for today.

Step 2: Determine Workout Type

Decision tree (simplified):

If recovery status is low: Easy run or rest.

If recovery status is moderate: Easy or moderate workout.

If recovery status is high: Quality session appropriate.

Within quality options: Select based on phase emphasis, recent history, and training distribution.

Step 3: Set Intensity Targets

Based on fitness estimates:

• Easy run: HR ceiling, pace floor
• Tempo: Specific pace/HR range at your threshold
• Intervals: Target pace for specific adaptation

Personalized to your current fitness, not generic tables.

Step 4: Determine Volume

Based on:

• Training phase volume targets
• Recent volume trend
• Today's workout type
• Recovery status

Output: Specific duration or distance for today.

Step 5: Package the Prescription

Final output includes:

• Workout type
• Warm-up structure
• Main set details (intervals, paces, recoveries)
• Cool-down
• Total expected duration

Clear, executable instructions.

Workout Type Selection

The Workout Menu

Available options:

• Recovery run
• Easy run
• Easy-moderate run
• Progressive run
• Tempo/threshold run
• Interval session (various)
• Long run (various intensities)
• Fartlek
• Hill session
• Race-specific session

Algorithm selects from this menu based on context.

Selection Logic

Phase-based defaults:

• Base phase: More easy, long runs, limited intensity
• Build phase: Adding tempo, interval options
• Peak phase: Race-specific, sharpening workouts

Modified by:

• Recovery status (reduces to easier options when fatigued)
• Recent history (avoids repeating same type consecutively)
• Weekly distribution (ensures appropriate balance)

Example Selection Process

Scenario: Week 8 of marathon build, Tuesday, good recovery metrics.

Algorithm reasoning:

• Phase: Build (tempo/threshold important)
• Day: Typically quality day slot
• Recent: Long run Sunday, easy Monday
• Recovery: Good, ready for intensity

Selection: Tempo run (threshold development, appropriate for phase, fits schedule).

Fallback Logic

If primary selection isn't appropriate: Algorithm has fallback options.

Example: Tempo planned but HRV suppressed → Fall back to easy run or moderate fartlek.

Intensity and Volume Determination

Intensity Calibration

For each workout type, intensity is calibrated:

Easy runs:

• Based on YOUR aerobic threshold
• HR ceiling calculated from your data
• Pace range reflecting your current fitness

Tempo runs:

• YOUR threshold pace (not formula-derived)
• HR target at YOUR threshold
• Sustainable for prescribed duration

Intervals:

• Pace targeting YOUR VO2max or specific system
• Recovery duration matched to YOUR recovery rate

Volume Determination

Total volume considers:

• Weekly volume target for current phase
• What's already been completed this week
• What's remaining this week
• Today's workout type (intervals need less volume than easy runs)

Output: Specific miles or minutes for today.

Dynamic Adjustment

If state changes from initial assessment: Prescription adjusts.

Example: Started workout, HR significantly higher than expected → Alert to cut workout short or reduce intensity.

Algorithm Improvement Over Time

Learning from Outcomes

After each workout:

• How did actual performance compare to prescription?
• How did recovery respond?
• Any reported issues?

Algorithm learns: Your specific response patterns and refines future prescriptions.

Population to Individual

Initial prescriptions: Based on population models (what works for similar runners).

Over time: Individual patterns dominate (what works for YOU specifically).

Result: Increasingly personalized prescriptions with more data.

Feedback Integration

Your input helps:

• Rating perceived effort
• Reporting how workout felt
• Noting any issues

Algorithm incorporates: This feedback into future decision-making.

The Human Element

What Algorithms Can't See

Factors outside data:

• Your motivation today
• Life stress affecting you
• Intuition about your body
• Preferences for certain workouts

Algorithm prescribes based on available data. Some context is missing.

Override Capability

Good systems allow:

• Swapping prescribed workout for alternative
• Adjusting intensity within reasonable range
• Skipping or adding based on feel

The algorithm should serve you, not dictate without flexibility.

Algorithm + Judgment

Optimal use:

• Let algorithm handle routine optimization
• Apply judgment for unusual situations
• Provide feedback to improve future prescriptions

Neither pure algorithm nor pure intuition—both together.

Workout prescription algorithms bring computational power to an inherently individual process. By considering fitness, fatigue, phase, and history simultaneously, AI prescribes workouts optimized for your current state. The result: every workout serves a purpose, calibrated to what you need and what you're capable of today.

Get AI-optimized workouts on your dashboard.