VDOT vs Critical Speed: Which Should You Use for Training?

VDOT or critical speed? I choose VDOT when I need training paces from a trustworthy recent race; I choose critical speed when I need an estimate of the heavy–severe intensity boundary from multiple maximal efforts. VDOT translates one performance into race equivalents and pace categories. Critical speed (CS) fits a distance–time model and also estimates D′, a finite distance capacity above CS. After 12 years of running, I have learned to ask what decision the number must support before I calculate it.

My Translate–Boundary–Verify method keeps the jobs separate. I translate a race result with VDOT, locate a boundary with CS, and verify the proposed workout against how I actually run it. The two models can complement each other; they are not interchangeable scores.

My quick choice: If I have one trustworthy recent race and need broad training paces or race equivalents, I start with VDOT. If I have several comparable maximal efforts and need an intensity boundary, I use critical speed. If either input is weak, I use effort and a cautious pace range until I can test again. I never convert the VDOT score directly into CS.

My 60-second comparison card

Decision	VDOT	Critical speed
Primary question	What do my recent race and equivalent training paces suggest?	Where is my heavy–severe boundary, and what is my finite capacity above it?
Minimum input	One representative race or time trial	Two maximal efforts; I prefer three well-spaced efforts
Main output	A performance score, race equivalents, and pace categories	CS in speed or pace plus D′ in distance
Main strength	Simple translation into everyday training language	Individual boundary model using multiple performances
Main weakness	One race may not represent every distance or current condition	Protocol and model choices can shift the estimate
Best first use	Broad pace guidance and race equivalence	Boundary-focused interval design and performance profiling
My safety rule	Never generate training pace from goal fitness	Never treat D′ as a perfectly rechargeable battery

Neither model measures the whole runner. Neither replaces durability, fueling, terrain skill, recovery, or judgment.

The two questions are different

VDOT asks: What level of running performance does this result represent inside the Daniels model? From that answer, I can compare race equivalents and obtain training pace guidance.

Critical speed asks: What speed best describes the boundary between heavy and severe exercise from this runner’s distance–time relationship? D′ then describes a finite distance-like capacity above that boundary.

The distinction matters. A race-translation tool and a boundary model may produce nearby paces without becoming the same construct.

I also keep the units attached: VDOT is a performance score; CS is a speed that I may express as m/s or min/km; D′ is a distance, usually meters. A VDOT of 48 and a CS of 4.3 m/s are not numbers to compare or average. The useful comparison is what each estimate implies for a specific workout or race.

TRANSLATE: what VDOT gives me

VDOT is a performance-derived value associated with the work of Jack Daniels and Jimmy Gilbert. I enter a recent race or well-executed time trial, and the model translates it into a score, equivalent race performances, and training paces such as easy, marathon, threshold, interval, and repetition.

The attraction is immediate usability. One trustworthy result can seed an entire pace vocabulary.

I do not call VDOT a measured VO₂max. It reflects achieved performance through a model that includes the interaction of aerobic capacity, running economy, and the performance curve. A laboratory VO₂max test and a VDOT value answer different questions.

What a VDOT result can and cannot tell me

It can help me	It cannot prove
Compare performances at different race distances	That I have equal durability at every distance
Set broad training pace ranges	That one exact pace is correct in all conditions
Identify a possible speed–endurance imbalance	Why the imbalance exists
Communicate workout categories simply	The physiological boundary used by another model
Update training after a meaningful performance	That a single great day defines permanent fitness

A runner with strong mile speed may receive an ambitious marathon equivalent. That number is a hypothesis about balanced development, not a promise. Long-run durability, fueling, muscle resilience, and experience still decide the marathon.

My VDOT input gate

Before I trust the translation, I grade the input.

Input factor	High confidence	Medium confidence	Low confidence
Recency	Within the current training phase	Several weeks old	From a different fitness cycle
Course	Accurate and comparable	Rolling or moderately technical	Uncertain distance or extreme terrain
Conditions	Interpretable	Heat, wind, or altitude affected it	Conditions dominated performance
Execution	Evenly paced and genuinely hard	Small pacing mistake	Tactical, interrupted, or major collapse
Readiness	Normal taper or training context	Some fatigue	Illness, acute fatigue, or injury symptoms

I attach the confidence label to the output. “VDOT 48, medium confidence” is more useful than a naked 48 printed to several decimals.

My even-split pacing system improves the quality of a race or time-trial input by reducing the distortion created by a reckless start.

The circular mistake I refuse to make

I never enter my goal race time, receive a higher VDOT, and then use those faster paces as the training required to justify the same goal. That turns desire into evidence.

I train from current demonstrated fitness. The goal can shape event choice and long-term progression, but it does not get to manufacture today’s threshold or interval pace.

If a current VDOT pace feels wrong, I inspect the input, conditions, workout design, and recovery before defending the calculator.

BOUNDARY: what critical speed gives me

Critical speed uses the relationship between distance and time across multiple hard efforts. One common linear form is:

distance = CS × time + D′

In that model, CS is the slope. D′ is the intercept—a finite distance-like capacity associated with running above CS.

I interpret CS as an estimate of the heavy–severe boundary. Below it, a physiological steady state may be achievable depending on how far below and for how long. Above it, tolerance is finite and the disturbance continues to accumulate toward exhaustion. CS is a model boundary, not a pace I can hold indefinitely.

I show the basic calculation below, but I do not let a neat formula hide weak test data. I inspect trial spread, conditions, and fit before using CS to change a workout.

When I want to build sessions from that boundary, I use my critical-speed training guide for the workout decisions. Here I focus on choosing between the two models and checking whether the estimate deserves my trust.

D′ is useful—and easy to oversell

D′ helps describe why two runners with similar CS can perform differently above the boundary. One may have a larger finite reserve and handle a faster surge or longer severe-domain effort.

I do not treat D′ like a fuel gauge that depletes and recharges with perfect precision. Recovery above and below CS depends on intensity, duration, prior work, fitness, and the model used. Field estimates can be sensitive to small errors in the trials.

For most recreational training decisions, CS is the sturdier first output. I use D′ conservatively when it genuinely changes interval duration, race analysis, or the profile I am building.

My critical-speed input gate

Test requirement	What I want	What weakens the model
Number of efforts	Three maximal efforts	Only two points produce an exact-looking line
Duration spread	Clearly different efforts across a useful range	Trials clustered at similar durations
Measurement	Track or accurately measured route	GPS uncertainty over short efforts
Recovery	Enough separation for genuine performance	Residual fatigue between trials
Conditions	Same surface, shoes, weather window	Mixed terrain, wind, or footwear
Pacing	Hard, controlled distribution	Sprint start and collapse

I often use efforts that span roughly 3–15 minutes, but I do not pretend one universal protocol fits every runner or purpose. The tests must be long enough to define the curve and short enough to avoid becoming redundant endurance trials.

Testing is training stress. I schedule it inside a reduced-load period instead of adding three maximal efforts to an already demanding week.

A transparent two-point example

Suppose a runner completes:

• 1,200 meters in 240 seconds;
• 3,000 meters in 660 seconds.

Using the two-point linear calculation:

CS = (3,000 − 1,200) ÷ (660 − 240)

CS = 1,800 ÷ 420 = 4.286 m/s

That is approximately 3:53 per kilometer.

I convert m/s to seconds per kilometer with 1,000 ÷ speed. Here, 1,000 ÷ 4.286 is about 233 seconds, or 3 minutes 53 seconds per kilometer. The pace is an interpretation of the model, not a new test result.

Then:

D′ = distance − (CS × time)

Using the 1,200-meter trial:

D′ = 1,200 − (4.286 × 240) ≈ 171 meters

This example is intentionally transparent, not ideal. Two points always create a perfect line, so they cannot reveal whether one trial was poor. I prefer a third effort and inspect how well all points fit before prescribing from the result.

Why model choice and rounding matter

Critical-speed estimates can be produced from different mathematical forms. Results may be close, but CS and especially D′ can shift with the model, test range, and error in one performance.

I record:

• exact protocol and dates;
• distances and times;
• surface, shoes, and conditions;
• pacing notes and session RPE;
• model used;
• error or fit information when available;
• rounded training range, not just the raw decimal.

If software returns CS at 4.2873 m/s, I do not prescribe training with four-decimal confidence. The runner’s daily state and environmental cost are far noisier.

VERIFY: how I make either model earn trust

Verification layer	What I check	What changes my decision
Session opening	Is the modeled pace immediately too hard or too easy?	Large mismatch after a complete warm-up
Mid-session	Are breathing, RPE, and mechanics appropriate?	Continuous drift or strain at stable output
Repeatability	Can all repetitions stay controlled?	Large fade or repeated surging
Recovery	Is normal training available afterward?	Threshold-like work creates race-like recovery
Race specificity	Does the model agree with relevant workouts?	Short-test fitness overpredicts long-race ability
Conditions	Does pace change logically with heat, hills, and wind?	Athlete forces the number regardless of demand

My training-by-effort guide supplies the internal controls, while my RPE scale for runners gives those sensations a shared language. Pace starts the session; the runner’s response decides whether the model remains useful.

VDOT and CS are not threshold synonyms

VDOT includes a threshold training pace category. CS estimates a heavy–severe boundary. Lactate threshold comes from a defined blood-lactate method. Ventilatory thresholds come from gas-exchange behavior.

These values may sit near one another in a runner, but proximity does not erase different inputs, protocols, and definitions.

My two-threshold training map separates the lower and higher threshold concepts. My guide on estimating lactate-threshold pace shows why I preserve the method label instead of calling every hard sustainable pace “threshold.”

How I use VDOT to seed training paces

I begin with the pace ranges associated with current VDOT, then apply purpose and conditions.

Training type	How I use the VDOT output	My field veto
Easy	Broad ceiling, not a pace target	Full conversation and low recovery cost must remain
Marathon	Starting estimate for specific work	Current durability and fueling must support it
Threshold	Range for controlled sustained work	RPE, speech, mechanics, and repeatability decide
Interval	Pace anchor for VO₂-oriented work	Rep duration and recovery must preserve workout identity
Repetition	Coordination and speed reference	I stop before mechanics deteriorate

The pace category does not write the workout. Duration, recovery, total volume, terrain, and weekly placement determine what the session actually becomes.

How I use CS to design boundary work

I do not run every threshold session exactly at CS. A boundary is not a bullseye.

For longer controlled work, I typically begin below CS and look for repeatable breathing and mechanics. Near-CS intervals can be shorter and carefully recovered. Above-CS work belongs to a different severe-domain purpose and draws on finite tolerance.

My threshold-run method turns the boundary into a controlled Find–Hold–Finish session. When the objective is broader aerobic control rather than boundary precision, I use the distinctions in my tempo-run guide instead. The model informs the opening range; execution controls the final prescription.

My model-to-workout map

Question	Lead model	Supporting check
What paces fit my recent 10K result?	VDOT	RPE and completed workouts
What equivalent half-marathon is plausible?	VDOT	Long-run durability and fueling
Where is my heavy–severe boundary?	Critical speed	Threshold behavior and test quality
How should I individualize hard interval duration?	CS and cautious D′ context	Rep consistency and recovery
Why are my short and long races uneven?	Both	Training history and event specificity
What pace should I run today in heat?	Neither alone	Effort, conditions, and session purpose

I maintain two models only when each changes a decision. More dashboards do not automatically create better coaching.

When VDOT and critical speed disagree

Disagreement is not a software emergency. It is information about inputs, model scope, or the runner’s profile.

Pattern	My likely explanation	My next move
Short-race VDOT looks strong; long equivalents fail	Speed is ahead of durability	Keep current speed; build long aerobic support
CS looks high; threshold work repeatedly fails	Tests or model may overestimate the boundary	Audit trials and lower the working range
VDOT paces feel easy; CS is conservative	VDOT race may reflect stronger current fitness	Validate with a controlled session, then retest CS
Both look strong; races underperform	Pacing, fueling, heat, course, or race skill	Fix execution before raising fitness targets
Models change after different shoes or surfaces	Tests were not comparable	Standardize protocol before interpreting fitness
D′ swings while CS stays stable	D′ is sensitive to trial error and model fit	Avoid precise above-CS prescriptions

I do not average the two numbers. I identify which question matters and which input best matches it.

A worked disagreement: 4:05 VDOT pace versus 3:53 CS pace

Suppose my recent-race VDOT suggests a 4:05/km threshold training pace, while the clean two-trial CS example above gives me 3:53/km. Those numbers answer different questions. I do not average them to 3:59/km or announce that one model is broken.

I first check whether the race and CS trials describe the same fitness period, course, weather, and footwear. I then run a controlled threshold session, such as three eight-minute repetitions beginning near the more conservative 4:05/km, with easy recovery. If breathing, mechanics, and the final repetition stay controlled, I can cautiously narrow the working range in a later session. If it already feels like a race, I back off and inspect the inputs.

This is an illustration, not a universal 12-second adjustment. The point is to let the session test a hypothesis without making a workout pay for disagreement between models.

A practical case: speed ahead of endurance

Suppose a recent 5K generates a VDOT that predicts an ambitious half marathon. The same runner’s CS is also strong, but long threshold blocks deteriorate and long-run mechanics fade after 90 minutes.

I do not raise the half-marathon goal because two models look flattering. Both may be accurately describing the runner’s current shorter-duration power while failing to supply the missing durability.

I use steady running and the durable aerobic work described in my gradual mileage-build guide to build the bridge. In my intermediate half-marathon plan, race-specific durability earns the goal rather than inheriting it from a 5K calculator.

A practical case: endurance ahead of speed

Another runner may produce modest short maximal efforts but hold pace exceptionally well across longer races. VDOT equivalents may underestimate the half-marathon relative to the 5K, while CS sits close to a large share of the runner’s race speeds.

That profile tells me not to force every session faster. I preserve the endurance strength and introduce carefully dosed faster work. The weekly architecture in my intermediate 10K plan provides room for that without turning each workout into a retest.

Conditions can move output without moving fitness

Heat, humidity, wind, hills, altitude, surface, and accumulated fatigue can slow performance at the same internal demand. A lower pace on a difficult day does not automatically mean VDOT or CS has fallen.

I standardize formal tests when possible and record the conditions. In ordinary training, I preserve the session purpose rather than forcing the modeled pace.

Heart rate may add context, but it also drifts and can be wrong. My guide to heart-rate drift while running keeps one late-run rise from becoming a false fitness verdict.

My test-week design

I do not stack a VDOT time trial and three CS trials into a normal quality week.

Day	Example test-week role	My reason
Monday	Rest or easy	Arrive fresh enough to interpret effort
Tuesday	Short CS trial	Capture the short end without prior fatigue
Wednesday	Easy	Absorb the maximal effort
Thursday	Medium CS trial	Add a distinct duration
Friday	Rest or short easy	Protect the longest trial
Saturday	Long CS trial	Complete the spread
Sunday	Easy or rest	Review before prescribing

If I also need a current VDOT race, I use an existing race result or schedule the race in a different week. Testing should not consume the block it is meant to improve.

How often I update the models

I update VDOT after a meaningful recent race or time trial that represents current fitness. I do not change it because one interval rep was fast.

I retest CS after a meaningful training block, often six to eight weeks, when the result will change training. I repeat a comparable protocol rather than casually replacing one point in an old model.

I delay testing when illness, injury symptoms, extreme weather, or accumulated fatigue would make the result hard to interpret.

My confidence scorecard

Confidence	VDOT	Critical speed	How I prescribe
High	Recent, hard, well-paced measured race	Three clean, well-spaced comparable trials	Normal ranges plus field checks
Medium	One meaningful limitation	Two trials or one compromised condition	Wider ranges; validate early
Low	Stale, tactical, uncertain, or poor execution	Clustered, GPS-dependent, or fatigued trials	Do not let the number lead

Confidence can differ between outputs. I may trust CS more than D′ from the same test set. I may trust a VDOT 10K translation for threshold pace but not its marathon equivalent.

Precision should shrink when confidence falls.

Mistakes I no longer make

I do not call VDOT measured VO₂max

It is a performance-model output. That distinction preserves what the number can honestly claim.

I do not create current training from goal VDOT

Goal fitness is not evidence. I train the runner I have today.

I do not calculate CS from one effort

A relationship requires multiple points. One time trial can inform performance but cannot define the distance–time line.

I do not cluster CS tests

Trials that are too similar in duration create a weak lever for estimating slope and intercept.

I do not treat D′ as an exact battery

Depletion and recovery are model-dependent and context-sensitive. I use it as a guide, not a countdown clock.

I do not average incompatible paces

If VDOT threshold pace and CS differ, I investigate their inputs and intended meanings. Averaging can hide the problem.

I do not prescribe decimal precision

I convert outputs into practical ranges and let the athlete’s response refine them.

I do not let testing replace training

A model earns its value through better sessions and races. Frequent maximal testing can make the plan worse.

My final decision tree

1. If I have one trustworthy race and need usable pace guidance, I start with VDOT.
2. If I have several clean maximal efforts and need a boundary model, I calculate CS.
3. If I need race equivalence, VDOT leads and event-specific durability verifies it.
4. If I need heavy–severe boundary context, CS leads and controlled sessions verify it.
5. If inputs are weak, I use effort and broad ranges until better evidence arrives.
6. If both models agree, I still test the prescription rather than celebrating the agreement.
7. If they disagree, I choose the model that answers the current question and audit the other input.

That hierarchy keeps the tool subordinate to the training decision.

Frequently asked questions

Is VDOT better than critical speed?

No. VDOT is simpler for race translation and broad training paces. Critical speed provides a boundary-focused model from multiple efforts. The better tool depends on the question and input quality.

Can I convert VDOT directly into critical speed?

Not reliably. The models use different inputs and structures. I calculate each from appropriate performance data, preserve the method label, and compare them only in context.

Is critical speed the same as threshold pace?

Not automatically. CS estimates the heavy–severe boundary from performance. “Threshold pace” can refer to several physiological or coaching definitions. I state the method before comparing values.

Is VDOT the same as VO₂max?

No. VDOT is a performance-derived model value, while VO₂max is measured through a specific laboratory protocol. They may relate, but they are not interchangeable measurements.

How many efforts do I need to estimate critical speed?

Two can fit the basic linear model, but I prefer three well-spaced maximal efforts because a third point helps expose a poor trial. Protocol consistency matters as much as quantity.

Which model predicts marathon pace better?

Either can overpredict when durability, fueling, weather, or pacing is weak. I use marathon-specific workouts, long-run response, and race history alongside the model.

How often should I update VDOT and critical speed?

I update after a meaningful race or training block, not after every fast workout. Six to eight weeks is a useful review interval when retesting will change the plan.

What should I do when VDOT and CS disagree?

I audit recency, pacing, course, conditions, test spread, and event specificity. Then I use the model that answers the current decision and validate it in a controlled workout.

My bottom line

I compare VDOT and critical speed with a Translate–Boundary–Verify method. VDOT translates one performance into accessible race equivalents and training paces. Critical speed uses several efforts to estimate a heavy–severe boundary and D′.

The best model is not the one with the most impressive number. It is the one built from trustworthy inputs, attached to the right question, and willing to be corrected by the runner’s actual training.