This article is for coaches, data enthusiasts, and anyone who wants to understand exactly how the W'bal model works—including its strengths and limitations.
Historical Background
The Origin (1965)
The Critical Power concept was introduced by Monod and Scherrer in 1965. They observed that the relationship between power output and time to exhaustion follows a hyperbolic curve:
P(t) = W'/t + CPWhere:
- P(t) = Power sustainable for time t
- W' = Curvature constant (work capacity above CP)
- CP = Critical Power (the asymptote)
This can be rearranged to the linear form:
W = CP × t + W'Total work (W) equals Critical Power times duration, plus the finite capacity W'.
The Two-Parameter Model
The hyperbolic power-time relationship gives us two parameters:
- Critical Power (CP): The highest power output that can be sustained without progressive W' depletion. Represents the boundary between heavy and severe intensity domains.
- W' (W prime): The finite work capacity available above CP. Historically called "anaerobic work capacity," though this is physiologically imprecise.
The W'bal Model (Skiba, 2012)
The Problem
The original CP model only predicted time to exhaustion for constant-power efforts. Real cycling involves intermittent efforts—surges, recoveries, attacks. How do we track W' during variable-power exercise?
Skiba's Solution
Dr. Philip Skiba proposed the W'bal (W' balance) model in 2012, allowing real-time tracking of W' depletion and reconstitution.
Two algorithms emerged:
1. Integral Algorithm (Original)
W'bal(t) = W' - Σ W'exp(u) × e^(-(t-u)/τ)Where:
- W'exp(u) = W' expended at time u
- τ (tau) = Time constant for recovery
- e = Euler's number
The time constant τ is calculated as:
τ = 546 × e^(-0.01 × DCP) + 316Where DCP = difference between CP and recovery power.
2. Differential Algorithm (Froncioni/Clarke)
A simplified approach that uses only current power:
If P > CP:
W'bal = W'bal - (P - CP)
If P ≤ CP:
W'bal = W'bal + (W' - W'bal) × (1 - e^(-1/τ))This is the default in GoldenCheetah and many other platforms.
Key Physiological Concepts
What W' Actually Represents
W' correlates with:
- Phosphocreatine (PCr) stores in muscle
- Anaerobic glycolytic capacity
- Accumulation of metabolites (H+, Pi, lactate)
It's not purely "anaerobic"—oxygen kinetics also play a role. The term "anaerobic work capacity" is a simplification.
Why Recovery Is Non-Linear
W' depletes linearly above CP (each second costs the same), but recovery is curvilinear:
- Fast initial recovery (PCr resynthesis)
- Slower later recovery (metabolite clearance)
- Recovery rate depends on how far below CP you drop
This is why the exponential recovery model is used.
Validation Studies
Field Validation (Skiba et al., 2014)
Key findings:
- Mean W'bal at exhaustion: 0.5 ± 1.3 kJ (essentially zero)
- In non-exhausted conditions, minimum W'bal: 3.6 ± 2.0 kJ
- ROC curve analysis: AUC = 0.914 for predicting exhaustion
The model successfully identifies when athletes are in danger of exhaustion.
Limitations in Elite Athletes (2017)
Research by Bartram et al. found:
- Elite cyclists recovered W' faster than the model predicted
- Bias of -112 ± 46 seconds in τ values
- "The current SKIBA 2 model is not appropriate for use in elite cyclists"
Implication: τ should be individualized, especially for highly trained athletes.
Recent Research (2024-2025)
Current consensus:
- W'bal is useful for monitoring and pacing guidance
- Recovery time constants (τ) need individualization
- The model works better for recreational/trained cyclists than elite
- Accuracy is approximately ±20% for recovery predictions
Mathematical Implementation
WattPlan's Approach
We use the differential algorithm for simplicity and real-time performance:
class WPrimeBalance {
constructor(cp, wPrime) {
this.cp = cp;
this.wPrime = wPrime;
this.wBal = wPrime;
}
update(power) {
if (power > this.cp) {
// Linear depletion
const expenditure = power - this.cp;
this.wBal = Math.max(0, this.wBal - expenditure);
} else {
// Exponential recovery
const recoveryPower = this.cp - power;
const tau = 546 * Math.exp(-0.01 * recoveryPower) + 316;
const recovery = (this.wPrime - this.wBal) * (1 - Math.exp(-1 / tau));
this.wBal = Math.min(this.wPrime, this.wBal + recovery);
}
return this.wBal;
}
}Time to Exhaustion
TTE = W'bal / (P - CP)Only valid when P > CP. When P ≤ CP, TTE is theoretically infinite.
Estimating CP and W' from Data
Linear Regression Method
Using best efforts at multiple durations:
Work = CP × Time + W'With data points (t₁, W₁), (t₂, W₂), (t₃, W₃):
CP = (W₂ - W₁) / (t₂ - t₁)
W' = W₁ - CP × t₁Three-Point Protocol
Recommended durations for testing:
- 3 minutes (heavily W'-dependent)
- 12 minutes (mixed)
- 20 minutes (heavily CP-dependent)
From Power-Duration Curve
Using non-linear regression on the hyperbolic model:
P = W'/t + CPThis is more robust with noisy field data.
Known Limitations
1. Individual Variation in τ
Recovery rates vary significantly between individuals. The standard τ equation is an average, not a personal value.
2. State Dependence
W' and recovery rate may be affected by:
- Prior exercise (reduced W')
- Glycogen depletion
- Environmental conditions (heat, altitude)
- Accumulated fatigue
3. Supra-Maximal Efforts
The model assumes W' depletion is linear, but at very high intensities (>150% CP), neuromuscular factors may cause earlier failure.
4. The "Slow Component"
At severe intensities, VO2 continues to rise even at constant power (VO2 slow component), which isn't captured by the simple two-parameter model.
Practical Implications
What W'bal Is Good For
- Real-time pacing guidance during hard efforts
- Identifying exhaustion risk
- Comparing effort patterns between rides
- Tracking anaerobic fitness trends
What W'bal Is NOT Good For
- Precise prediction of exact failure time
- Absolute values across different individuals
- Highly accurate recovery timing
- Replacing proper testing with estimates
Best Practices
- Use W'bal as a guide, not absolute truth
- Calibrate CP and W' with proper testing when possible
- Learn your individual recovery patterns over time
- Consider W'bal accuracy as ±20%
References
- Monod H, Scherrer J. (1965). "The work capacity of a synergic muscular group." Ergonomics, 8(3), 329-338.
- Skiba PF, Chidnok W, Vanhatalo A, Jones AM. (2012). "Modeling the expenditure and reconstitution of work capacity above critical power." Medicine & Science in Sports & Exercise, 44(8), 1526-1532.
- Skiba PF, Fulford J, Clarke DC, Vanhatalo A, Jones AM. (2014). "Validation of a novel intermittent W' model for cycling using field data." International Journal of Sports Physiology and Performance, 10(8), 1034-1041.
- Bartram JC, Thewlis D, Martin DT, Norton KI. (2017). "Accuracy of W' recovery kinetics in high performance cyclists—modeling intermittent work capacity." International Journal of Sports Physiology and Performance, 13(6), 724-728.
- Jones AM, Vanhatalo A. (2017). "The 'critical power' concept: Applications to sports performance with a focus on intermittent high-intensity exercise." Sports Medicine, 47(Suppl 1), 65-78.
- Skiba PF, Clarke DC, Vanhatalo A, Jones AM. (2014). "Effect of work and recovery durations on W' reconstitution during intermittent exercise." Medicine & Science in Sports & Exercise, 46(7), 1433-1440.
For a more accessible introduction, see The Complete Guide to W' and Critical Power.