Can AI Read Your Karting Data? An Honest Answer

Let's be honest: AI is revolutionizing our lives, and it's probably going to do it even more in the future. I'm one of the power users of Claude, for example. I started using ChatGPT a couple of years ago, then swapped to Claude early this year, and it's working really well for me. I do lots of tasks on it, and I've been working on PURPL a lot thanks to Claude.

But the thing is that now you can really use AI in racing, and in karting specifically. If you have no knowledge of reading data — or even if you do have data analysis experience — you'll probably be better off using AI to do the analysis for you. Because the reality is that we as human beings are not going to be at the same level as AI when it comes to computing all the numbers and the math required across all those laps, corners and sectors. You will not be able to do that anywhere close to the speed of AI.

Not just any AI — it has to be built for karting

But obviously not just a random AI. Not just using ChatGPT, Claude or Gemini. You need a specific app, an AI app that has been built and developed fully karting-centric, that knows exactly what to do, that has predefined templates of what results to spit out, and gives you the useful insights you need in order to find lap time.

Here at PURPL I've been really pushing myself to get the best possible outcomes and get drivers to find lap time as easily and effortlessly as possible. Of course, you still have to go out there and do the steps required to get faster — the AI is not going to do that for you. But identifying the corners where you're not doing well right now, and where you're losing time in a specific session, is super useful, because then you can go into the following session with a clear idea of what to work on and where to push your limits.

It's really the tool I wish I had when I drove 20, 15, 10 years ago — and even when I'm driving today.

When I go and drive myself, I use it all the time. But the question many karting drivers ask themselves is whether AI can actually read karting data, and do it in a real way — not hallucinated, with fake data and fake coaching.

The rule that decides whether you can trust it

That is the priority here at PURPL, and I'm super strict on it.

If there isn't high confidence in the coaching being provided, I'd rather let the user know there isn't enough confidence than give wrong feedback.

That's the test I'd apply to any tool in this space, ours included. It's easy to build something that always has an answer. It's much harder — and much more useful — to build something that will tell you when the data doesn't support a conclusion. A confident-sounding wrong answer costs you a session, or a weekend. If you want the mechanics of what's happening under the hood, that's how an AI racing coach actually works.

What AI genuinely brings: memory and scale

We're currently in a position where AI can reach a large amount of data and parse it all together. But the great thing is that AI has memory. It's going to remember everything you did in previous sessions at that specific track, or at other tracks. It's going to know your driver DNA, and have an idea about the areas you've been struggling with most lately, and the areas where you should be working to improve — even though you may not have data from reference drivers.

That's the thing: with AI you can find lap time against yourself, and not necessarily against some reference driver, whether that's your teammate or somebody else. And that's where PURPL excels, because it lets you see, among all the laps and all the sessions you did, which were your best sections and what theoretical best lap comes out of them.

I always prefer to do that from a single session, because comparing different sessions in karting is most of the time not a fair comparison. It's not apples to apples: every session the track changes, it develops, it either gets a little bit faster or a little bit slower. So I believe it's better to compare yourself within the same session, and see how far you were from reaching the ultimate perfect lap.

How the lap gets split — and why fake splits ruin everything

How does the AI do that? It splits the lap using an algorithm we created at PURPL that detects where one sector begins and another ends. Not the same way the MyChron does it — in a different way, so that you end up with more splits than the three you get from AiM. Depending on the track you end up with five, six, seven, eight.

And it's great because those sectors are real. They're not splitting the middle of a chicane, which would be unrealistic for a theoretical lap. In a fast or slow chicane, if you're driving for the entry and the split lands in the middle of it, you'd gain a lot of time on one lap but lose so much on the exit — and the theoretical would be unreal. That was the problem we faced: we had to make sure it got fixed, and we needed a strong algorithm, which we built. So now we can confidently give you splits that are real and don't affect the previous corner or the next corner in order to make that split faster. They're autonomous splits, which is really important. The underlying idea is the same one behind splitting a lap into sectors, just done so the pieces respect where corners actually are.

Once you have those splits, you can see: in my fastest lap I gave up a tenth in mini sector one, half a tenth in mini sector two, two tenths in mini sector three. And maybe at the end of the lap, after seven splits, you're four tenths off your ultimate theoretical best time. That's showing you what you were capable of doing with whatever you had in that moment — with your engine, your chassis, your skills. It just needed to be put together.

How far off your theoretical means "not ready to win"

I'm a big fan of this comparison against yourself, because the goal is to become the best version of yourself. And what I've noticed is that when drivers aren't ready to win yet, they're typically very far away from their theoretical. To me, far away means two or three tenths. That's really too much.

Imagine qualifying for the CIK world championship: within one tenth you find 15 or 20 drivers in a Junior or Senior race. In a WSK qualifying session, if you're more than a tenth off P1 you'll be at least P15. So it's crucial to match whatever performance you had underneath you in that moment.

You can't blame the engine or the chassis or the setup for a tenth lost, if you have three or four tenths in yourself displayed in the theoretical.

I see it in myself too. Whenever I'm driving at the top of my game, I come out of the kart and I'm not even half a tenth off the theoretical — the ultimate lap theoretical from PURPL — and I'm probably matching the theoretical from the AiM MyChron. That means I was able to reach the peak performance of the kart and of the driving at the same time, and put everything together.

It's fairly easy to do that in qualifying, because the performance comes in the last couple of laps: if you have a clear lap it's quite easy to match your theoretical, since by the last lap there's more grip, and unless you make big mistakes you should match your fastest lap in every sector. In a longer stint it's much harder. If you do 20 laps, with at least 10 or 15 in the optimal operating window of your tires and engine, and you're doing a really good job pushing your limits — at some point you'll make mistakes. So getting close to your theoretical over a stint is hard. But if you can sustain that level of consistency and closeness to your theoretical, you're set for many, many successful races. That's the moment when you realize: I'm ready to win now.

Why comparing against yourself beats comparing against a rival

That's the really cool thing about AI — you can use yourself as a benchmark and extrapolate the maximum performance from yourself, even without data from other drivers.

And even if you did have their data, you'd never have certainty about their kart compared against yours. You can always say: oh, they had a different engine, different tires, a different chassis, a different setup. There's always an argument. Someone else's data is genuinely useful — in a race, or with multiple teammates, to find the best lines and the best technique, whether you have to drive harder on the entries or open the entry up a bit more. But when you're driving by yourself you still have lots to work on, lots to learn, and still the goal of hitting that ultimate perfect lap.

The perfect lap doesn't refer to any driver around you. It's you against yourself — you and the kart against yourself.

That's what drivers like Lewis Hamilton, Max Verstappen and Ayrton Senna refer to. And I feel PURPL lets you track your performance against yourself, so you know whether you had a good session, how close you are to your theoretical, and whether you're improving across recent sessions.

Read relative values, not absolute ones

Always remember that in karting there's a huge difference in tires, in grip level on the track, and in air temperature. Of course you can readjust the jetting on the carburetor for different carburation, but the reality is the track is never going to be the same, and neither is your kart.

If you're not planning to change tires — because you're trying to save money, and not everybody can run new tires every single day — then by definition every session your tires are a little older. And it's not just age: they accumulate heat cycles, so they get harder, ever so slightly, with less tread and therefore less grip. It wouldn't be fair to compare your last session against your first.

So be open-minded: yes, the last session I wasn't as quick as my first ones, probably because my tires were worse. But you can still see how consistent you were — your standard deviation — and how consistent your braking points were, and how many meters away you were driving from your average line. You can see whether you've been consistent in your lines, or too inconsistent. That's something really cool to analyze, and it's what we provide inside PURPL. I notice it in myself every time.

You can also see how far you are from the theoretical in the last session. Maybe you have old tires and you're two tenths off your personal best from the morning, but you're driving a bit more consistently and a bit closer to the theoretical — because you built up muscle memory during the day and removed some rust, maybe it had been a few weeks since you last drove. By the end of the day you're more precise, more consistent, closer to your potential, but maybe a bit tired. If you do that well and then fit fresher tires the next morning, when you're fresher in body too, you'll be going in the right direction.

In karting it's not an absolute value you should look at — it's the relative value. You'll never have the exact same track with the exact same grip level, because you'll have a different number of drivers out there and different rubber laid down. Some drivers run hard tires, which makes the track slower than if everyone runs softs. If everybody runs soft tires the track gets really fast, even though you may still be on old hard tires — and the opposite applies too. So take every session and every day with a grain of salt, and just try to improve yourself. That's the same reason raw lap time is a poor progress metric.

So — can AI read your karting data?

Yes, AI can definitely read your karting data. But the role of a driver coach is still very useful, because AI is not going to tell you what you did wrong in a race in terms of racecraft, or about your body posture, your psychology, fixing your mindset during the day, or reading your body language outside and inside the track. Experienced racing drivers and coaches know how to deal with all of that. AI isn't there, and probably won't be — unless you have a camera that literally watches you and tracks your body posture during a race. And even if that happened, I still believe the driver coach would have the upper hand on everything requiring judgment. I've gone through that split in detail.

But when it requires numbers and data, I feel AI is going to beat a driver coach and a data telemetry guy every single time.

And coaches shouldn't be ashamed of using AI, feeling like it's a shortcut, feeling guilty. You shouldn't feel guilty. Think about all the professionals in the world — marketing professionals, finance professionals, sales professionals. They all use data and AI software to improve their workflows. That's nothing to be ashamed of. It's just the way humans work: they try to improve productivity. Telemetry already changed karting once; this is the same shift, one step further.

AI is not here to replace humans and driver coaches. It's here to augment them.

I genuinely feel driver coaches are going to become even better coaches if they implement AI. There are many different websites trying to solve the same problem — the goal is to pick the ones that solve it in the best possible way and have a proven track record. I'm sure PURPL has it.

So let me know how it works for you, sign up today, and feel free to hit me up and tell me what you think, or if you have improvement requests. I'll be here to listen. Take care, and remember to keep sending it. Ciao ciao.