When you load a session, AB Racing Line analyses your telemetry and produces two separate estimates of recoverable lap time: a Technique Gain from changes to your driving inputs, and a Setup Gain from car setup improvements. In the app they're the Time still on the table card on the Overview: your driving is the technique gain, car setup the setup gain, and the big number is the two combined. This article explains exactly how those numbers are built — what signals are used, what formulas are applied, and where the uncertainty comes from.
Technique gain is built up corner by corner from signals that are directly measurable in your telemetry. Only signals with a clear, calculable relationship to lap time are included. Signals that improve the feel of the car but cannot be directly converted to seconds — such as smoothness or trail braking style — are excluded from the number.
This is the highest-confidence estimate we produce.
We measure where you begin braking on each lap, averaged across all your clean laps. We compare this to the latest braking point consistent with the corner's speed requirement — derived from the car's average deceleration rate and the minimum apex speed needed to make the corner cleanly.
The time saved is calculated from basic kinematics:
For example: if you are braking 20 metres earlier than necessary at 150 kph (41.7 m/s), the recoverable time in that braking zone is approximately 0.48 seconds — before accounting for what you can do with that extra distance. We apply this calculation to every corner individually and sum the results.
Each marker shows where braking started on a single lap. The gap to the optimal late brake point is recoverable time every single corner.
We measure your average apex speed at each corner and compare it to the maximum the car's available lateral grip supports. The lateral grip ceiling is estimated from your own best G-circle utilisation across the session — we use your data, not a theoretical maximum.
The time relationship is:
Arc length is estimated from the GPS trace. Speed values are in metres per second. For example: a 5 kph apex speed gain through a medium-speed corner with an 80-metre arc saves approximately 0.07 – 0.12 seconds through that corner alone.
Coasting is defined as any moment where throttle is below 5% and brake is below 2% simultaneously. It represents time where no meaningful force is being applied to the car — the driver is in transition between inputs rather than using the available grip.
We measure the total coasting time per lap from your telemetry — and then compare it against your own quickest lap of the session, not against a fixed target.
That matters more than it sounds. How much coasting a lap needs is a property of the circuit, not of the driver. At NASCAR Charlotte a driver's own best lap still coasts 17.2% of the time, so the fixed 4% benchmark we used to apply declared nearly 16 points of that "waste" — time the shape of the track makes unavoidable. Measured against their own best lap, the honest figure is 2.7 points.
Not all of even that gap is recoverable; some coasting is deliberate weight transfer. We apply a 30% recovery factor, and cap the result at 3% of lap time.
Across 30 sessions, the quickest lap coasts less than a typical one in 27 of them — a median of 0.70 percentage points. So this is something you demonstrably already do better when the lap goes well, which is exactly the kind of gain worth quoting.
The amber zone is where neither throttle nor brake is applied. Only the coasting ABOVE what your own best lap needs counts, and 30% of that gap is treated as recoverable.
This page used to carry a fourth technique component based on how much of the lap you spend at
full throttle. We removed it, and it is worth explaining why rather than quietly
dropping it.
Full throttle percentage is an outcome, not a lever. You do not go faster by
holding the pedal down longer; you hold it down longer because you carried the corner
better — which the apex-speed component above already counts. Measuring it again was counting
the same gain twice.
The data said so too. Across our session corpus a driver's own fastest lap beats a typical one
by a median of just 1.1 percentage points of full throttle, and in 5 of 30 sessions the fastest
lap had less full throttle than a typical one. The sign is not even stable. Meanwhile
the fixed 60% target it was measured against failed on over half of all real sessions, where
the honest range runs from 37% to 74% depending on the circuit.
In one real case it was contributing 6.3 seconds of an 8.3 second estimate.
Removing it is the single biggest reason the numbers on this page are smaller, and truer, than
they used to be.
The conservative end of the range uses only high-confidence signals (braking point) at 80% of their calculated value. The optimistic end adds medium-confidence signals at their full calculated value. Both ends exclude low-confidence signals entirely.
Setup gains are inherently harder to quantify than technique gains because they depend on how the driver responds to a changed car. We use two measurable signals.
Incorrect tyre pressures reduce the size of the contact patch. An over-pressured tyre crowns in the centre; an under-pressured tyre cups at the edges. Both conditions reduce peak lateral grip.
The relationship between pressure deviation and grip loss follows established tyre physics:
A front-view cross-section of each tyre meeting the road. The shape of the tread tells you everything: a crowned tyre touches in the middle only, a flat tyre touches across its full width, and a cupped tyre touches at the edges only. The strip on the road below shows the actual contact patch — both over and under-pressured tyres leave grip on the table.
We measure your actual hot pressures from the telemetry (80th percentile of on-track readings to exclude outliers) and compare them to the target derived from tyre temperature profiles. The lap time equivalent is:
Cornering time fraction is the proportion of your lap spent cornering — calculated from your GPS trace and speed data. For most road courses this is between 0.55 and 0.70.
When the car has a handling imbalance — entry oversteer, exit understeer, or similar — the driver unconsciously applies corrective steering inputs to keep the car on line. These corrections are measurable in the telemetry as steering angle applied in the opposite direction to the corner.
We do not count steering inputs directly. Instead we measure attitude velocity — how fast the car is rotating relative to how fast it is actually turning — for the entry and apex phase of every corner. A phase where that exceeds ±2°/s is one where the car is genuinely loose or genuinely pushing, rather than one where you simply moved the wheel.
A setup change that resolves the imbalance removes the need to correct it. We apply a 50% recovery factor, because some of that time reflects habit rather than the car, and habit takes longer to unlearn than a spring rate takes to change.
Both numbers are deliberately small. A badly imbalanced session — three affected phases, which is the 90th percentile across our corpus, where the median is zero — lands on about 0.30s. That is the figure actually observed in the app's own records after a recommended balance change was applied.
The combined gain is not the technique gain and setup gain simply added together.
The reason: some of both gains occupy the same territory. If the car has an entry balance problem, the driver may be braking earlier than necessary to compensate for snap oversteer on turn-in — not because of their natural style. Fix the balance, and some of the "braking point gain" disappears because the driver was already at the right point for that car's behaviour.
We call this shared territory the overlap correction:
The overlap percentage is higher when attitude velocity data shows the car has a clear imbalance — suggesting more driver compensation is occurring — and lower when the car is well-balanced and the gains are genuinely independent.
Every technique component on this page is really a consistency measure. Each one asks the same question in a different way: what if you did, on every lap, what you already did on your best one?
That means the total cannot honestly exceed one number — the gap between a typical lap and the best lap you could assemble from your own fastest sectors. If your median lap is 1.2s off that, then 1.2s is what perfect consistency is worth. Claiming more is claiming you would go faster than you have ever actually gone, which is a different promise entirely and not one this page is entitled to make.
So we calculate it, and we cap the total at it. This is the single biggest change to how these numbers are produced, and it was overdue: before it existed, the old model exceeded that ceiling in 31 of 34 sessions. In one it offered seven times the available gap; on an oval, twenty-six times. It once offered 3.82 seconds to a driver whose median lap already equalled their best — a perfectly consistent session, where the honest answer is zero.
Across the same corpus, the largest figure the app will now show fell from 29.49s to 4.12s, and the typical one from 3.63s to 0.82s.
0.8s technique + 0.4s setup − 0.1s overlap = 1.1s combined potential gain
To keep the numbers honest, the following are intentionally excluded:
We show a range rather than a single number because lap time prediction is not an exact science. The range represents our genuine uncertainty, not a hedge.
If your session had variable conditions (wet/dry transitions, significant tyre degradation), the estimates are flagged as lower confidence and the range is widened accordingly.
The goal is for the number we show you to be one you can actually achieve — not a theoretical maximum that makes the app look impressive but leaves you frustrated on track.