The W′ Bank: Why Every Surge Costs More Than You Get Back

Somewhere around your third or fourth attack of a group ride, or the third time you punch up a short rise on a hilly bike leg, a quiet transaction happens that most riders never see. You spend a chunk of a finite anaerobic reserve exercise physiologists call W′ ("W-prime"), and you assume that easing off for a bit pays it back. It doesn't, not quickly, and not for free. This article is about what that transaction actually costs, how the repayment really works, and why the number your bike computer shows you for it is a model's guess, not a measurement.

If you've read our critical power and FTP coverage, you already know the shorthand: above critical power (CP), you're drawing down W′; below CP, you're not. What most coaching software doesn't tell you — because the physiology is genuinely unresolved — is how fast that reserve comes back, what determines the refill rate, and how far off the on-screen W′ balance can be from what's actually happening in your legs.

Why this matters if you're not racing crits

Age-group triathletes rarely attack anyone. But every non-drafting bike leg is full of small, unplanned above-CP moments: a punchy roller you take a bit too hard, a headwind section you push through rather than settle into, a red light you accelerate hard away from, a moment you close a gap to a group you shouldn't be near. None of these register as "intervals." All of them make a withdrawal from the same finite account. The practical question for a 70.3 or Ironman bike leg isn't whether you'll surge — you will, because roads aren't flat and neither is your attention — it's whether you understand how slowly and conditionally that account refills, so you can decide when a withdrawal is worth it and how to actually get the money back before you need it again.

What the evidence says

The refill is slow, and it's not linear

The most direct data on W′ reconstitution comes from Caen and colleagues, Medicine & Science in Sports & Exercise, 2019, who put 11 male participants (V̇O₂max 55 ± 4 mL·kg⁻¹·min⁻¹) through 12 trials of two exhaustive bouts separated by active recovery. Averaged across recovery conditions, W′ came back to only 46.0 ± 2.7% after 2 minutes, 51.2 ± 3.3% after 4 minutes, and 59.4 ± 4.1% after 6 minutes (p = 0.003). Read that again: after a full six minutes of recovery, these riders still had less than two-thirds of their anaerobic reserve back. That is the single most useful number in this whole topic, because it directly contradicts the mental model most riders carry — that a couple of easier minutes basically "resets" you.

Evidence grade: moderate. This is one study, one lab, 11 young men of unspecified training background, tested in a laboratory rather than on the road. It has not been independently replicated at scale. But the direction — slow, partial, non-linear recovery — is exactly what the underlying two-compartment model of W′ predicts, and the magnitudes are the best measured numbers available.

How hard you pedal during recovery changes the refill rate — recovery is a skill

The same study found that the intensity of the recovery itself matters, not just its duration. Recovering at 33% of critical power restored 9.4 percentage points more W′ than recovering at 66% of critical power (56.9 ± 3.9% vs. 47.5 ± 3.2%, p = 0.019) (Caen et al., 2019). In plain terms: soft-pedalling refills the tank meaningfully faster than "easing off" to something that still feels like effort. Riders who back off from 300 W to 220 W after a surge are recovering far more slowly than riders who back off to 150 W — even though both would describe themselves as "recovering." The same study also found that how you spent the W′ affects how it comes back: bouts that depleted the reserve more slowly (lower, more sustained power) produced slower reconstitution than bouts that depleted it quickly and violently. Depletion and reconstitution are asymmetric, and neither is a simple function of watts alone.

This reframes recovery from a passive wait into an active skill. The instinct after a hard punch is to hold a moderate pace that still moves you forward — psychologically satisfying, physiologically counterproductive. If the goal is to have something left for the next rise, the next surge, or the run off the bike, the recovery has to be genuinely easy, not merely "easier."

The number on your screen is a model, and the models disagree with each other

Most modern power-based coaching platforms display a running W′ balance using some version of the Skiba W′BAL model. It's worth being honest about what that number is: a mathematical estimate built on assumptions about exponential recovery, not a direct measurement of anything happening in your muscle. Two studies that have actually checked the model against measured W′ disagree with each other on how good it is.

Skiba, Jackman, Clarke, Vanhatalo & Jones, MSSE, 2014 tested the W′BAL formulation in 11 subjects and found a mean error of only −1.6 ± 1.1 kJ across conditions — a reassuringly small miss. They also found that measured W′ correlated with how elevated V̇O₂ still was at the start of the next effort (r = 0.79, p < 0.01): how hard your aerobic system is still working when you start the next surge is itself a decent predictor of how much anaerobic reserve you have left.

But Caen et al., 2019 — the same study behind the 46%/59% recovery figures — found the standard W′BAL model underestimated actual reconstitution by up to 29.7% after short recoveries. That is not a rounding error; it is closer to a third of the reserve the model tells you that you don't have, that you actually do. These two papers are not measuring identical protocols, so this isn't necessarily a flat contradiction — but it is a genuine, unresolved disagreement in the literature about how accurate the model most training software relies on actually is, and it should temper how much weight any rider puts on a live W′ balance number mid-ride. Treat it as a rough guide to the shape of your depletion, not a precise fuel gauge.

The fatigue signature of variable riding shows up in your muscles, not just your metabolism

If the metabolic accounting is uncertain, the neuromuscular evidence is cleaner. Theurel & Lepers, European Journal of Applied Physiology, 2008 had 10 subjects complete two 33-minute rides at identical average power — one held constant at 70% of maximal aerobic power, the other deliberately variable (10 seconds at 200% of maximal aerobic power, 15 seconds at 150%, 20 seconds at 100%, interspersed with recovery at 50%). Despite matched average power, the variable ride produced significantly greater reductions in maximal voluntary contraction torque, voluntary activation, and peak doublet force — markers of both central drive and the muscle's own contractile capacity — along with higher heart rate and blood lactate.

Evidence grade: moderate. The sample is small and the protocol's 200% surges are considerably harsher than most triathlon terrain will demand. But the finding matters because it shows variable-power riding leaves a measurable fingerprint on the muscle itself, distinct from anything the average power number would suggest. Two riders who finish a bike leg with identical average watts are not necessarily in the same physical state; the one who got there by surging is carrying more neuromuscular fatigue than the average alone predicts.

Does that fatigue cost you on the run? The two best studies disagree

This is the part of the picture that is genuinely unresolved, and worth stating plainly rather than papering over. Two comparable studies in triathletes point in opposite directions.

Bernard et al., EJAP, 2007 had 10 male triathletes ride a 20 km time trial three ways — constant power, variable power (68–92% of maximal aerobic power), and freely chosen — then run 5 km. The constant-power condition produced a significantly faster run (1,118 ± 72 s) than the variable-power condition (1,168 ± 73 s), with no metabolic differences detected between conditions. That supports the intuitive story: surging on the bike costs you on the run.

Suriano, Vercruyssen, Bishop & Brisswalter, Journal of Science and Medicine in Sport, 2007 found the opposite in 8 male triathletes cycling 30 minutes at 90% of lactate threshold, comparing constant power against power varied ±20% every 5 minutes, followed by a run to exhaustion. The variable-power condition produced a longer time to exhaustion (15:09 ± 4:43) than the constant condition (10:51 ± 3:32, p < 0.05) — variable power was better for the run, not worse.

The authors of the Suriano study offer a specific explanation: their variable protocol's final five-minute segment happened to be a low-power block, which may have functioned as an unplanned taper into the run — a kind of built-in recovery window right before transition. That is a genuinely useful hypothesis, and it reframes the whole question. It may be less about the variability index of the whole bike leg and more about what the last 10 minutes before you dismount look like. But it remains a hypothesis extracted from one small study, not a demonstrated finding, and both trials share the same limitations: eight to ten male subjects, short protocols (20–30 minutes of cycling), and nothing close to 70.3 or Ironman duration. We are not able to tell you, from the evidence that exists, whether variable bike power reliably helps or hurts your run. Anyone claiming otherwise is going beyond what two small, contradictory studies actually show.

Nobody has measured any of this in a fatigued state

Here is the gap that matters most for anyone racing longer than an hour: every W′ reconstitution study, including Caen's, tests fresh riders. Nobody has measured how W′ depletes or refills after three, four, or five hours of prior riding — which is exactly the state a 70.3 or Ironman athlete is in when a late climb, a headwind section, or a competitive urge to close a gap shows up. We don't currently know whether a fatigued W′ reserve refills at the same 46%-in-2-minutes rate, faster, or slower. It is entirely plausible that a depleted rider recovers more slowly — accumulated fatigue tends to work that way in other domains — but this is inference, not evidence. If you are pacing a long-course bike leg using a W′ balance number, understand that the model underneath it was validated, imperfectly, in fresh athletes doing short efforts, not in the state you'll actually be in at hour three.

Practical application: treating recovery as a skill, not a pause

None of this means variability is always wrong. Variability that matches the terrain — easing off downhill, pushing on a climb where the gradient actually changes your speed materially — is a different thing from variability driven by inattention, ego, or chasing an average number. The W′ kinetics above are about what happens after a surge, whichever kind it was, and what to do about it.

• After any hard, above-CP effort, actively soften your pedalling rather than settling into a moderate pace that still feels like work — the Caen data show a real, measurable difference between recovering at roughly a third of CP versus two-thirds of CP.
• Give the recovery real time. Expect to have less than half your W′ back after 2 minutes and still under two-thirds back after 6 — a short lull before the next rise is not a full reset, whatever your screen says.
• If your device shows a live W′ balance, treat it as a directional signal (getting low vs. getting high) rather than a precise number — the model behind it has been measured to underestimate how much reserve you have actually recovered by as much as 29.7% after short recoveries, though another study found a much smaller error.
• Save intentional above-CP efforts for terrain that pays you back — a climb steep enough to change your speed, not a flat surge to close a gap you don't need to close.
• In the last 10–15 minutes before T2, favour settling into an easier rhythm rather than a final push — not because the evidence proves this helps your run, but because the one study suggesting variable power can help ties the benefit specifically to a low-power finish, not to variability in general.

Common mistakes

• Assuming a couple of "easier" minutes after a surge is a recovery. If you're still riding at 60–70% of CP, you are refilling W′ far more slowly than you think.
• Treating the on-screen W′ balance as ground truth. It's a model output with a documented error margin, built from fresh-state laboratory data, not a sensor reading from your legs.
• Reading "variable power hurts your run" as a settled fact. The two best studies on this exact question disagree with each other, and both are small and short.
• Believing the physiology of surging in a fresh state tells you what happens at hour three or four of a long-course race. It might not — nobody has tested it.
• Spending W′ on surges that don't matter (traffic lights, ego moves, closing gaps you don't need to close) instead of saving it for terrain where it changes your speed.

How to apply this week

You don't need a dedicated session to put this into practice — you need a habit change inside the rides you're already doing.

• On your next group or terrain ride, after any effort that clearly felt "above threshold," deliberately drop to an easy, conversational pace for at least 3–5 minutes rather than a moderate one — notice how much harder it is to resist holding a "still-productive" pace.
• On a rolling long ride, pick one or two climbs where you'll intentionally spend W′, and commit to genuinely easing off on the following descent or flat, rather than letting the next rise arrive before you've recovered.
• If you train with a power meter and a W′ balance display, glance at the trend rather than the exact figure — use it to notice when you're consistently riding into the red, not to fine-tune pacing to the watt.
• In your next race-simulation long ride, plan the final 10–15 minutes as deliberately easy, and log how your legs feel off the bike compared to sessions where you finished hard — your own data point is more relevant to you than either single study above.

References

• Caen, Bourgois, Bourgois, Van der Stede, Vermeire & Boone, Medicine & Science in Sports & Exercise, 2019
• Skiba, Jackman, Clarke, Vanhatalo & Jones, MSSE, 2014
• Theurel & Lepers, European Journal of Applied Physiology, 2008
• Bernard et al., European Journal of Applied Physiology, 2007
• Suriano, Vercruyssen, Bishop & Brisswalter, Journal of Science and Medicine in Sport, 2007