Predictive Timing vs. Machine Learning Driver Coach

Published in Speed Secrets Weekly #266

Are lap times really that helpful?

Early in my time behind the wheel, I rode right seat in a Porsche Boxster, with my friend John Lewis. John was offering hot laps on behalf of the Porsche Sport Driving School, at our local Porsche Clubs’ Drivers’ Education event at Barber Motorsports Park. That ride was when I began to build muscle memory, feeling the chassis and the tires interact with the road. John would exclaim “feel this!” when the car would compress and gain grip. He made me realize what the car should feel like, and he gave me the power of knowledge. I now knew with absolute certainty that the car had more in it. I could be confident in that, I just had to learn to extract it…

“Everyone knows that lap times are the most obvious indicator of auto racing performance. But lap times are a very low-resolution tool for driver training and vehicle tuning – lap times can determine a winner at the end, but do not tell the complete story of the path to victory.” – Dr. Austin Gurley –  Creator of the Machine Learning Coach

So how do we uncover the above mentioned “path to victory”? Let’s start with what we learn at racing school. Attending Skip Barber, Bondurant, or a manufacturer school like those from Porsche & BMW is a very popular way to begin down the path of TRO (total racing obsession). If you have attended one of these programs, you’ll notice how much time the instructors spend teaching you how important a good lap time is… Just kidding! What you will notice, however is that those instructors focus on two things: Vision and Grip/traction/ “the limit.”

So why is it, when we graduate from these schools and strike out ambitiously on our own, that we focus so intently on lap times? I’m not saying times are not important, quite the opposite! The Lap Time is the end goal, but we’re not going to achieve that time by obsessing over it. Just as the above quote implies, lap times are a “very low-resolution tool for driver training and vehicle tuning.” Ask any good driver coach out there, and they’ll express that it’s not just about going fast, it’s about knowing how you’re going fast, and what actionable goals helped you to get there. The same can be said for theoretical or predictive lap times. At this point in time, the predictive or running +/- lap time (compared to a previous best lap or sector) is the predominant means of in-car driver feedback.

My goal with this article is to flesh out the differences and similarities between predictive timing, and other forms of real-time feedback; primarily the new machine learning style feedback seen in the APEX Pro device.

Predictive Lap Timing

The most familiar and commonly accepted form of real-time feedback is an in-car predictive lap timer. These types of systems are provided by most data systems, and even some phone apps. When I define what a predictive lap timer is, I can’t speak for every system out there, so I use some generalities. Predictive laps are computed based on a combination of your best “sectors” on the track. The more sectors there are, the better the resolution is on the predicted time. Some predictive or real time indications show you a +/- indication vs your best lap as well.

The type of predictive timing I have used for myself and with my students is a combination of your best sector times to make up an ideal lap. It’s displayed in real time to the driver, showing if you are slower than, or faster than, the predicted achievable lap. Sometimes you can also configure it to show if you’re slower, or faster than your best in the sector that you’re driving at that moment. A plus or minus indicator Is commonly displayed to show that you are slower (+) or faster (-) the predictive lap or sector.

Most systems require at least one full lap before a predictive time is displayed, because the system needs a baseline lap first. Like regular old lap times, predictive can be misleading when looking at complete performance, especially if you’re using a predictive that was computed during an earlier session. Conditions change and consumable items are used (tire wear, brake wear, fuel level). All this impacts that predictive time, so it’s usually best to have it base line during each session to reflect those changing conditions. Let’s make a quick list of pros and cons to predictive lap timers:

Pros:

  • Usually relevant, easy to understand metric (lap time)
  • Incentive to improve
  • Helps pinpoint WHERE you are slow and WHERE you are fast on the track

Cons:

  • Most systems display the delta time as text = hard to read/interpret while on track peripherally
  • Can be a false indication of performance – improvement in sector can equal time lost in the next
  • Is lap time good encouragement while on track? Does it reinforce the two things mentioned above?
  • Psychology – Am I capable of analyzing lap times to make driving changes in the car?

How does using a predictive timer reinforce the two things that you’ll learn at any and every driving school in existence? You tell me. What answers the question: “How can I go faster in turn X?” It usually has to do with a combination of two things:  Vision and Grip. Predictive timing helps you to determine the areas of deficiency on the track, and that may help narrow your focus – which is helpful. But, what we’re really looking for is a way to reinforce good vision habits, and sense the grip level. Those two things will not only make us faster, but more adaptable and versatile. Great racers aren’t just fast; they know WHY they are fast. They can adapt to changes conditions and car variables, and can always find a way to go a little quicker albeit consistently.

Let’s take a step back for a moment. If you are not a full-time racer, think about what other activities you do. Your career, a sport you play or played, or another hobby you have. Ideally, think about something that you do well. It might even be driving! Now, think about your progression in that activity. Think about when you went from consciously thinking about each step, to performing subconsciously. At this point you can process information about something entirely different while you’re driving at speed. Usually you turn this corner into subconscious consciousness because you trained your brain (i.e. muscle memory) to do so. For most of us, we train our brain by learning WHY we do things:

Why do we relax on the throttle and not add more steering when the car understeers? Why do we brake the hardest initially? Once you begin to break those things down and truly know WHY, you can commit that act to muscle memory. Building muscle memory is CRITICAL to our learning progression when driving on track. There’s no way we can automatically have that muscle memory, we must develop it.

Machine Learning Coach Feedback

We’re specifically talking about the real time feedback offered by the APEX device, but because it’s really its own segment in this market, I’ll refer to it as the Machine Learning Coach (MLC). The MLC presents something like GSum (display of maximum G-force achieved) BUT as a % relative to the “limit.” The “limit” varies with speed, track characteristics, and vehicle characteristics. Since GSum is simply displaying where you numerically, it’s not useful in real-time without a lot more information. Just like looking at vehicle speed at a corner’s apex without a reference for your target speed. That’s because the maximum G’s a car has reached in a session is not always achievable! Not every corner, braking zone, or track-out point allows the car to attain peak G forces.

The Machine Learning Coach compensates for vehicle grip and corner characteristics like banking and cresting, and other things that cause grip changes. The MLC extrapolates by building a model of “good behavior” corners and can show you the lost potential in areas where you struggle. The more laps the device learns, the more accurate it becomes. It can provide useful feedback within the first lap. MLC is not a time-based system, lap time is not used as part of its algorithm. It’s 100% grip based. That means that the sensors in the device are constantly defining the edges of the friction circle, albeit in 9 axes.

The Machine Learning Coach reinforces working with the track. Fast drivers don’t wrestle the car into submissions, they seduce it. They do this by working with the natural undulations and grip level on the race track. Because the MLC is measuring everything inertially, it understands that when the car is going uphill, you have more grip! It understands how on-camber and off-camber corners affect the grip level too. MLC is constantly adapting and is always redefining the “limit” based on new information. It’s great for beginners and advanced drivers alike because of that adaptability. To summarize, let’s make another pros and cons list for the machine learning coach:

  • Grip based – not time based. This means you’re not out there pushing for a lap time!
  • Answers both ‘where’ and ‘how’ to gain speed because driver inputs can be directly associated with the LED display. It also helps to uncover ‘why’
  • Adaptive technology: constant learning, meaning the “limit” is accurate and constantly redefined. Helps drivers at all levels
  • Helps to build muscle memory – you can associate the way the car feels with validation from the LED display

So which one is right for me?

Both real-time indications have their advantages. Ultimately, the systems you should use depend on your goals and why you’re at the track. The great thing is that predictive and APEX can work together. They don’t have to exist in a vacuum, in fact, they probably reinforce one another! I’d personally recommend a predictive timer only for drivers that are competing. When in competition, it’s worthwhile to have that information, whether used on the track or not. But, I’d also want to know how much of the available grip I’m using, which will necessitate the MLC as well.

Ultimately, reading a lap time while at speed is nearly impossible for our peripheral vision. I’d recommend steering clear of using text for in-car indication. There’s a reason shift-lights use LEDs, and those awesome radar rear view cameras on professional racecars use color indication as opposed to text! Lastly, how do you know if your theoretical lap time is achievable or not? I’m not sure, but I’d rather be learning to feel the limit, than frustrated I can’t match an unattainable lap time…

 

 

One thought on “Predictive Timing vs. Machine Learning Driver Coach

Leave a Reply

Your email address will not be published. Required fields are marked *