The Sim-to-Real Gap: A Honest Look

I get asked about the sim-to-real gap two different ways. Sim racers want to know whether their training transfers to real-world driving when they finally book a track day. Track-day drivers ask the inverse: is sim time a waste before their next event, or is it the fastest preparation they can buy? The honest answer to both is the same. Yes, sim training transfers, partially, and the partial is predictable. The gap lives in a specific place, and naming where the gap lives is the difference between treating sim hours as preparation versus treating them as a substitute.

This post is the founder version of the argument that the sim-to-real transfer article develops in technical detail. The doctrine claim, in one sentence: shapes transfer between sim and real, and magnitudes do not. What follows is what that means in practice.

What sim training transfers cleanly

A non-trivial portion of driving knowledge lives in the driver’s mental model of the track and the car, not in their seat-of-pants calibration. Brake-zone ordering — which corners need a hard stop and which need a brush of the pedal — transfers cleanly between platforms. Line shape, the geometric path through a corner from turn-in to apex to exit, is the same shape on iRacing as on a real GT3 with the same corner radius. Trail-brake patterns, throttle modulation off the apex, race-craft instincts in traffic — all of this category of knowledge belongs to the model the driver carries in their head, and the model is platform- independent.

Repetition is sim training’s other asymmetric advantage. Two hundred laps of one corner in an evening is something no real-world circuit budget supports, and the kind of weakness shape that requires repeated isolation to fix is exactly the kind of thing sim training is unmatched at delivering. A driver who has run two hundred Spa Eau Rouges in iRacing this month has a calibrated mental model of that corner that survives the platform jump.

What sim training does not transfer

The gap lives in the magnitudes — in the specific values the driver feels in the seat. Tire feel under real-world load and surface temperature does not transfer; the rubber does what the rubber does, and no haptic feedback rig has yet calibrated to a real-world contact patch. G-force calibration is the inner-ear signal that tells the driver how much braking force the car is generating; that signal does not exist in the sim, and a driver who has never felt real braking G has only the visual cue of the car’s pitch to tell them where the limit is.

Fear is the third magnitude that doesn’t transfer. The cone of visual perception narrows in the real cockpit because there is an armco at the end of the runoff, and the narrowing changes the driver’s commitment level on entry to high-speed corners. The sim has a reset button. A real circuit has consequences that arrive at the speed the driver is currently traveling. Mechanical sympathy — the sense that a missed gear is a real expense and a clipped kerb is real damage — only develops in the real car.

Weather adaptation in actual rain is the fourth magnitude. A wet circuit with cold tires in real life produces a sliding behavior that no sim’s tire model has fully captured. The shape of the line a driver takes around a wet circuit transfers; the rate at which they trust that line under the conditions does not.

The doctrine claim, made concrete

Shapes transfer. Magnitudes do not. The same brake-release shape that lands cleanly at iRacing’s Sebring T1 will land cleanly at the real Sebring T1, with the same line shape and the same trail-brake overlap. The pedal pressure required to produce that release will differ. The G-force the driver feels under the deceleration will differ. The fear the driver feels at the entry to the corner will differ. The shape of the driving is the same; the magnitudes need recalibration on the day.

This is why I do not claim sim training is a substitute for real-world seat time, and why I also do not dismiss sim training as somehow disconnected from real driving. Both framings are honest about half the picture and wrong about the other half. The framing that holds up across data is shapes-transfer-magnitudes-don’t.

The mistake everyone makes about sim time

The common mistake is treating sim hours and real-track hours as substitutes for each other on the same axis. They are not. They are complements with asymmetric advantages. Sim is unmatched at repetition volume, race-craft scenarios with full thirty-car fields, and rapid setup iteration. Real track is unmatched at tire-degradation feel, real- stakes commitment calibration, and weather adaptation under genuine conditions. Time spent on either is time well spent if the driver is deliberate about which kind of learning they are extracting from the session.

What the development plan article calls the framework runs across both surfaces. The same five verbs — diagnose, prescribe, execute, measure, adapt — apply to a sim session and to a track-day session. The weakness shape extracted from a Tuesday-evening iRacing session and the weakness shape extracted from a Saturday track-day session can be the same shape, and the drill that fixes it can be the same drill, with magnitudes recalibrated on the day.

What this means for LAP

The cross-platform Driver OS — one driver profile that travels with the driver across sim and real-world surfaces — is the positioning argument that distinguishes LAP from the engineering-grade tools the same data lives in. It is also what makes the shapes- transfer claim load-bearing in a product rather than only in a doctrine document. A weakness shape extracted on iRacing on Tuesday becomes a prescription that can be drilled on Saturday’s track day. The same trace-reading framework runs on either surface. The same coaching library, the same prescription format, the same review cadence.

Without that cross-platform unification, the driver who runs both surfaces lives in two disconnected data silos and has to reconstruct the connection in their head every session. With it, the connection is the product.

The track day after sim training

The concrete framing I give first-time track-day drivers who arrive after sim prep is short. Sim work calibrated the shape of your driving. The track day calibrates the intensity. The first session is recalibration of intensity, not a chance to match sim lap times in real life. The shape will be familiar; the pedal pressures, the G-force under braking, the cone of visual perception — all of those need a re-baseline. Treat the first session as a re-baseline session. By the third session, the magnitudes have caught up with the shapes the sim already taught, and the lap times converge.

Why the honest framing matters

Adjacent products in the cluster overclaim sim-to-real transfer or dismiss it entirely; both stances are wrong about the same fact, in opposite directions. The honest framing — shapes transfer, magnitudes don’t — is more useful in practice and more durable across edge cases. A driver who internalises the framing trains differently: sim sessions chase shapes, track sessions chase magnitudes, and the framework runs across both.

The gap between sim and real is real. Naming it precisely is what makes the work between them coherent rather than a guessing game.