Reading Your First WOT Pull: What a Good Datalog Actually Looks Like

A wide-open-throttle datalog answers exactly one question: is this engine operating the way the calibration thinks it is, and is it safe? As of July 2026, four numbers decide that — knock retard, delivered versus commanded air-fuel ratio, fuel trims leading into the pull, and intake air temperature. Read them in that order and a log that looked like a wall of columns becomes a short list of decisions.

Most people open their first log and start scrolling horizontally, looking for something dramatic. Nothing looks dramatic. Everything looks like numbers. The problem is not the data — it is that nobody told you the reading order, and the reading order is most of the skill.

Before you read: what makes a log readable

A log you cannot interpret is usually a log that was captured wrong. Three requirements, and they are not negotiable:

One gear, one pull. Start the recording, drive normally for ten seconds or so, then take one uninterrupted pull in a single gear from low RPM up to your intended limit, then lift and keep recording for another few seconds. Shifting mid-pull makes the RPM axis fold back on itself and destroys your ability to see a trend against engine speed.

Context on both ends. The cruise before the pull is where fuel trims live. The coast after it is where you see whether temperatures were climbing. A clip containing only the loud part throws away half the diagnostic value.

The right channels, sampled fast enough. If a channel is not in the log you cannot reason about it, and a slow sample rate turns a knock event into a gap between samples. Which channels to select and why is its own subject — the datalog channel guide covers the list and the trade-off between channel count and sample rate.

With that in hand, here is the order.

1. Knock retard — the only channel that can end the session

Knock retard is the ECU pulling ignition timing because a knock sensor heard something it interpreted as detonation. It is a protective action, and it is the first thing you look at because it is the only channel in the log that describes potential engine damage rather than tuning quality.

Scan the entire pull. What you want to see is zero, flat, all the way through.

What you might see instead:

• Isolated single-sample blips of a degree or less. Common, usually noise, worth noting and watching across repeat pulls rather than acting on immediately.
• Sustained retard of two degrees or more across a range of RPM. Stop. This is a real event until proven otherwise, and continuing to make full-load pulls while investigating is how expensive parts break.
• Retard that appears at a specific RPM every single time regardless of load. That pattern is the signature of mechanical noise rather than combustion — a heat shield, a bracket, an accessory. Worth chasing with a wrench before touching a table.

The distinction between real detonation and false knock matters because the fixes are opposite. Real knock is answered with less timing, more fuel, cooler intake air, or better fuel. False knock is answered by finding the rattle. Pulling timing to hide false knock costs power permanently and fixes nothing.

If the car was pinging audibly before you ever logged it, the cause list is broader than the calibration — the walkthrough on rough running and pinging covers the mechanical possibilities worth ruling out first.

2. Commanded versus delivered AFR — the honesty check

This is the single most informative comparison in the whole log, and it is a comparison, not a value. Two channels:

• Commanded AFR (or commanded lambda) — what the calibration intended to deliver.
• Delivered AFR — what your wideband sensor actually measured in the exhaust.

If those two track each other within a couple of tenths through the pull, your air model and fuel model agree with reality. That is the whole point of the exercise, and it is what makes any subsequent timing work meaningful.

When they diverge, the direction is diagnostic:

What you see	Most likely cause	Where to look
Delivered leaner than commanded, worsening with RPM	Airflow underestimated at high flow, or injectors near their limit	MAF calibration or injector duty cycle
Delivered richer than commanded, worsening with RPM	Airflow overestimated, or injector data scaled too large	Injector data and the airflow table
Error flat across the whole pull	Global scaling issue — injector slope or fuel pressure assumption	Injector slopes
Error only in a narrow RPM band	Localized airflow table error in those cells	The specific cells at that airflow

That last row is the useful one and the one beginners miss. An error confined to a band is a cell problem, not a scaling problem, and correcting it globally will break everything that was already right.

A lean condition at full load deserves particular caution: it raises combustion temperature exactly where cylinder pressure is highest, which is precisely the condition that produces the knock you checked for in step one. Lean-at-WOT and knock retard appearing together are not two problems. They are usually one.

3. Fuel trims — what the ECU was already compensating for

Short and long-term fuel trims live in the closed-loop cruise portion of your log, before the pull. They tell you how much the ECU has been quietly correcting to hit its target under light load — which means they tell you how wrong your airflow model already was before you asked it to do anything hard.

Read them at steady cruise, not during transitions. Trims within a few percent of zero mean the base calibration is close. Trims consistently positive mean the ECU is adding fuel to compensate for airflow being underestimated, or for an actual air leak. Trims consistently negative mean the opposite.

The connection to your WOT data is direct: an airflow model that needs a ten-percent correction at cruise is unlikely to be accurate at full load, and full load is where you have no closed-loop correction to save you. Fixing trims first makes the WOT numbers interpretable. Doing it in the other order means chasing a moving target.

The VE and MAF correction calculator turns a trim reading into a specific percentage change for the affected cells, which is considerably faster than guessing.

4. Intake air temperature — the variable that invalidates comparisons

Intake air temperature does not usually cause a failure. It causes inconsistency, which is worse, because it makes you believe changes worked when they did not.

Watch IAT across the pull and, more importantly, across a session. A first pull on a cool intake and a fourth pull after heat soak are not comparable data. Denser, cooler air means more airflow at the same throttle position, more cylinder pressure, and a greater tendency toward knock. If your third pull shows knock your first pull did not, the calibration may be identical and the air simply hotter.

The practical rule: let the car cool between pulls until IAT returns close to where it started, and note the IAT on every pull you intend to compare. Two pulls with a thirty-degree IAT spread tell you almost nothing about each other.

Putting it together: the go / no-go read

You now have everything to make the only decision the log exists to support.

Go looks like this: no meaningful knock retard anywhere in the pull; delivered AFR tracking commanded within a couple of tenths through the RPM range; cruise trims in single digits; IAT stable and reasonable; and injector duty cycle with headroom left at the top of the pull.

No-go is any of: sustained knock retard; a delivered AFR that runs lean of commanded as RPM climbs; injectors at or near one hundred percent duty; or a trim reading large enough that the airflow model is clearly wrong.

Between those two sits the common real-world case — mostly good, one specific thing off. That is not a failure. That is the log doing its job, which is to point at one table with one correction so the next pull is better than this one. One change at a time, one log per change, and the process converges. Change three things at once and you have learned nothing from either log.

What a log will not tell you

It will not tell you how much power the car makes. It will not tell you whether the tune is finished. And it cannot tell you whether a part is failing mechanically — a log describes what sensors measured, and a sensor cannot see a cracked ring land.

What it does tell you, reliably, is whether the calibration's model of the engine matches the engine. That is the foundation everything else is built on. The complete guide to reading a datalog goes deeper on individual channels, and the datalog analysis service page explains how TuneVault reviews a VCM Scanner pull for commanded-versus-delivered AFR, knock retard, trims and IAT, and returns a go or no-go with the specific cells to change.

If you would rather have a second set of eyes on your first pull than guess at it alone, start here.

Frequently asked questions

What is the first thing to look at in a WOT datalog?

Knock retard. Everything else is a tuning conversation; knock retard is a damage conversation. Scan the whole pull for any non-zero knock retard first, and if you find sustained retard above roughly two degrees, stop reading the rest of the log and treat it as a no-go until you understand the cause. Once knock is clear, move on to delivered air-fuel ratio, then trims, then intake air temperature.

What air-fuel ratio should I see at wide-open throttle?

Most naturally aspirated gasoline builds run somewhere in the mid-twelves to high-twelves lambda-equivalent at full load, and boosted builds run richer still. The exact target belongs to your combination, not to a number off the internet. What matters far more when reading a log is the gap between what the ECU commanded and what the wideband actually measured — a delivered value tracking the commanded value within a few tenths is the sign of a calibration that is doing what it thinks it is doing.

Why does my commanded AFR not match my actual AFR?

Because one of the two air models is wrong for the airflow you just reached. Delivered leaner than commanded usually means the ECU is underestimating airflow or the injectors are running out of duty cycle; delivered richer than commanded usually means airflow is being overestimated or injector data is scaled too large. The direction of the error tells you which table to look at, and the RPM where it appears tells you which cells.

How long should a datalog pull be?

Long enough to include roughly ten seconds before the pull, the full pull in one gear from low RPM to your intended limit, and several seconds after you lift. The before-and-after context is what lets you separate a real trend from a sensor artifact, and the single-gear requirement is what makes the RPM axis readable at all. Short clips of only the loud part are the most common reason a log cannot be interpreted.

What is false knock and how do I tell it apart from real knock?

False knock is the knock sensor picking up mechanical noise — a loose heat shield, an accessory bracket, injector or valvetrain clatter — and reporting it as detonation. Real knock usually appears at high load and high cylinder pressure, scales with timing and heat, and moves when you change fuel. False knock tends to show up at odd, repeatable RPM points regardless of load, and does not care what octane is in the tank. Both are worth investigating; only one is fixed in the calibration.

Do I need a wideband sensor to read a datalog properly?

For anything at wide-open throttle, yes. The factory narrowband sensor is accurate only very near the stoichiometric point and effectively reports rich-or-lean beyond that, so a full-load log without a wideband contains no usable fuel data. You can log part-throttle fuel trims without one, but the pull itself will be unverifiable.

Can a datalog tell me how much power the car makes?

No. A datalog tells you whether the engine is operating safely and whether the calibration is behaving as intended. Power is measured on a dyno or, roughly, with careful timed runs. Treating a clean log as a power figure is the most common misreading of the whole exercise — a clean log means the tune is safe to keep developing, not that it is finished.