Science — Energy System Progression (ESPE) & Phase Development

8 min read·Applies to: Montis AppChatGPT & ClaudeOpen in the Montis App

The question this layer answers

Is the athlete actually improving, remaining stable, or losing capability over time?

Adaptation Progression looks beyond acute load and recovery.

Montis currently uses the Energy System Progression Engine (ESPE) to compare rolling power-curve windows and identify changes across different performance durations.

ESPE is stateless: it compares supplied current and previous power-curve windows rather than inventing adaptation from the current week's training load.

Adaptation Pulse

The Adaptation Pulse shown in the Montis dashboard is a presentation of the Energy System Progression Engine (ESPE).

It is not a separate model or additional calculation.

ESPE compares the athlete's current rolling power-curve window with the preceding equivalent window and resolves:

individual energy-system status;
overall adaptation state;
curve dynamics;
supporting power-model metrics.

The dashboard then presents a simplified view of those ESPE outputs as the Adaptation Pulse.

For example, the current comparison may be:

85 days vs previous 85 days

The exact comparison window is supplied by the ESPE curve_window.

The Adaptation Pulse provides a quick view of:

Anaerobic status;
VO₂ status;
Threshold status;
Aerobic Durability status;
Dominant curve shift;
Ride Fatigue Resistance state, trend and primary limiter, when supported.

The detailed Adaptation dashboard exposes the same ESPE output with the underlying power curves, percentage changes and derived metrics.

In simple terms:

ESPE = the adaptation model

Adaptation Pulse = the summary presentation

Adaptation Dashboard = the detailed ESPE evidence

Ride Fatigue Resistance = how well cycling power is retained after accumulated work

Power-curve anchors

For cycling and other power-supported sports, the current ESPE model can use anchors including:

5 seconds;
1 minute;
5 minutes;
20 minutes;
60 minutes.

For running, the required anchor set is reduced where appropriate.

The anchors broadly represent different performance domains:

very short / neuromuscular;
anaerobic;
VO₂-related;
threshold;
aerobic durability.

Montis compares the current power curve with the preceding comparison window and calculates percentage change at the available anchors.

Ride Fatigue Resistance

Ride Fatigue Resistance is an optional extension of ESPE that measures how well cycling power is retained after the athlete has already accumulated substantial work.

Instead of asking only:

What power can the athlete produce when fresh?

it also asks:

How much of that power remains available deep into a ride?

Montis uses the fatigued power curves configured for the athlete in Intervals.icu. These are identified internally as kj0 and kj1, but the actual depletion thresholds are athlete-configured values.

For example, an athlete may configure curves after:

1000 kJ;
2000 kJ.

These values are not fixed by Montis and may differ between athletes. If no fatigued power curves are configured or returned, Montis does not produce a fatigue-resistance assessment.

Fatigue Resistance is currently supported for Ride power curves. It is not generated for Run.

Retention by duration

At each configured depletion threshold, Montis evaluates the available power anchors:

5 seconds;
1 minute;
5 minutes;
20 minutes;
60 minutes.

Retention is calculated by comparing fatigued power with the normal power curve from the same period:

Retention % = fatigued power ÷ normal power × 100

Higher retention means the athlete preserves more of their normal power after accumulated work.

Lower retention means power degrades more substantially deep into the ride.

Montis compares the current retention values with the matching values from the preceding equivalent window:

Retention change (pp) = current retention % - previous retention %

pp means percentage points.

For example:

previous retention: 77.45%;
current retention: 78.09%;
change: +0.64 percentage points, shown as +0.64 pp.

Performance domains

The duration anchors are grouped into fatigue-resistance domains:

Short Power: 5 seconds and 1 minute;
VO₂: 5 minutes;
Threshold: 20 minutes;
Long Duration: 60 minutes.

The primary limiter is the domain with the lowest current retained-power result at the evaluated depletion threshold.

For example, if 20-minute retention is the lowest domain result, the primary limiter resolves as Threshold.

Governed summary

Montis evaluates the highest configured depletion threshold for which sufficient current data exists. Where a matching previous fatigued curve exists, it also resolves longitudinal trend.

The current fatigue-resistance state can be:

Robust: overall retention of at least 90%;
Moderate: overall retention of at least 80%;
Limited: overall retention below 80%;
Unknown: insufficient usable duration domains.

These are Montis operational classifications, not universal physiological thresholds.

Trend is governed using the change in overall retention:

Improving: increase of at least 2 percentage points;
Stable: change within ±2 percentage points;
Declining: decrease of at least 2 percentage points;
Baseline: no usable matched previous fatigued curve.

Confidence reflects the completeness of the current and previous fatigued-curve evidence:

High: two matched thresholds with complete domain coverage;
Moderate: matched previous evidence with sufficient domain coverage;
Low: current-only, incomplete or limited comparison evidence.

State and trend describe different things.

An athlete can therefore be:

Limited but improving;
Limited and stable;
Moderate but declining;
Robust and stable.

Relationship to durability and readiness

Fatigue Resistance is closely related to durability.

Good fatigue resistance means the athlete's power curve degrades relatively little as accumulated work increases. Limited fatigue resistance means one or more power domains fall substantially once the athlete is deep into a ride.

It does not directly measure acute fatigue, recovery or daily readiness.

A decline in fresh power with stable fatigue resistance is possible, just as improving fresh power with declining fatigue resistance is possible. Montis therefore keeps normal ESPE power-curve progression and fatigue-resistance progression as separate signals.

Fatigue Resistance is currently presented as governed ESPE evidence in reports and dashboards. It does not independently change the Adaptive Decision Engine score or directive.

More: Ride Fatigue Resistance

Energy-system status

Each system can be classified from the observed change as:

strong gain;
moderate gain;
mild gain;
stable;
decline;
unknown.

ESPE also derives higher-level information such as:

glycolytic bias;
aerobic durability ratio;
durability gradient;
system balance;
VO₂ reserve relationship;
curve profile;
curve quality;
adaptation bias;
adaptation state.

The model can identify states such as:

aerobic consolidation;
VO₂ expansion;
anaerobic build;
mixed adaptation;
VO₂ / threshold decline;
plateau;
baseline.

A decline in one system does not automatically mean global fatigue. ESPE output must be reconciled with recent exposure, recovery and Performance Intelligence.

Curve dynamics

ESPE also evaluates how the shape of the power curve has changed.

It derives:

overall vertical shift;
rotation index;
dominant shift.

The dominant shift can indicate whether the curve has moved relatively more toward short-duration or long-duration capability.

This matters because adaptation is rarely uniform across every energy system.

Phase detection

Training phase is resolved separately from ESPE.

ESPE determines how performance capability is changing across equivalent power-curve windows.

Phase detection determines what the athlete's recent training-load pattern represents.

Montis infers phases from weekly training behaviour rather than relying only on a manually assigned calendar phase.

Weekly load model

Completed activities are first aggregated into weekly training load.

Montis then evaluates:

weekly TSS;
CTL;
ATL;
TSB;
week-to-week load change;
ACWR;
Load Variability Index;
CTL and ATL direction;
relative weekly load.

Where Intervals.icu CTL and ATL are available, those values are treated as authoritative.

Montis does not normally reconstruct them.

If CTL or ATL are unavailable, the phase detector contains an EWMA fallback:

longer-term load is approximated with a 6-period EWMA;
acute load is approximated with a 2-period EWMA.

TSB is then resolved as:

TSB = CTL - ATL

EWMA-smoothed load trend

Montis does not classify the phase from one isolated week's load change.

The raw week-to-week TSS change is calculated first and then smoothed using an exponentially weighted moving average:

Weekly load delta
    ↓
EWMA smoothing (span = 3)
    ↓
Smoothed load direction

This reduces the influence of a single unusually large or small week.

The resulting trend is interpreted broadly as:

increasing load;
stable load;
unloading.

Fatigue and freshness context

The smoothed load trend is interpreted together with TSB.

Montis groups TSB into broad states such as:

deep fatigue;
fatigue;
neutral;
fresh;
very fresh.

This means the same reduction in training load can have a different meaning depending on the athlete's current form.

For example:

unloading while fatigued can represent Recovery or Deload;
unloading while fresh can represent Taper;
stable load around neutral form can represent Base;
increasing load with acceptable form can represent Build.

Additional load controls

Phase detection also uses:

ACWR;
relative weekly load versus CTL capacity;
Load Variability Index;
CTL slope;
ATL slope.

These provide supporting context for whether the observed weekly pattern is consistent with progressive loading, unloading, recovery or tapering.

Phase resolution

The engine can resolve phases including:

Base;
Build;
Peak;
Taper;
Recovery;
Deload;
Transition;
Overreached.

The phase is therefore not produced by EWMA alone.

A more accurate representation is:

Weekly training load
    ↓
CTL / ATL / TSB context
    ↓
EWMA-smoothed week-to-week load trend
    ↓
ACWR + variability + relative load
    ↓
Fatigue / freshness context
    ↓
Phase classification

This gives Montis a traceable phase model based on how training load is actually evolving.

Deload versus Recovery

Montis treats Deload and Recovery as related but distinct training states.

Deload describes a stronger intentional reduction in training stress while fatigue is still present.

Recovery describes a lower-load consolidation state where unloading is occurring without the stronger reduction required for Deload, or where freshness has already begun to return.

For example, in the current phase classifier while the athlete is in a fatigued TSB zone:

smoothed load reduction below approximately -5% can resolve as Recovery;
a stronger reduction below approximately -12% can resolve as Deload.

When the athlete is already fresh, unloading is interpreted differently and may resolve as Recovery or Taper depending on the remaining load relative to CTL.

So:

Deload = stronger reduction of the training stimulus while fatigue is present

Recovery = consolidation through reduced load and improving freshness

Taper = event-oriented unloading intended to preserve performance while increasing freshness