You Suck at Aero

Guest post from Mario. Editorial comments in red.

Every year I make the same New Year’s Resolution: 1) drink more water, and 2) stretch. It’s simple, free, and I still can’t fucking do it. Same thing every year. At the start of 2019 I had a crazy vision to do a bunch of aero tests at Watkins Glen. I didn’t call it a resolution (because then I wouldn’t have done it), but I resolved to do the tests just the same.

I wrote the story of Real World Testing Miata Aerodynamics on my website, and so I won’t reiterate that here, but I thought it would be good to give my brother’s readership a fresh angle on the tests, and recap some of the high points. Or low points, as it were.

The resolution I make every year is “talk less, listen more”.

You Suck at Watkins Glen

The three biggest obstacles to testing at Watkins Glen are weather, Armco, and the combination of weather and Armco.

Ian and I grew up 25 miles from Watkins Glen, and have suffered a lifetime of events that have been hampered (or is that hammered?) by weather. Upstate NY weather just plain sucks. Nevertheless, Watkins Glen is an historic track, attains high speeds, and has a great back straight for testing aero. So despite knowing better, that’s the location I chose. Of course I got bit by it.

The night before the test it rained, and the track was still wet track in the first session. I took the opportunity to rush home to get some more aero parts, while the rest of the test team sent the car on track for some initial shake downs. When I arrived back at the track, I see three guys literally hammering on the hood of my car. What the fuck did I just miss?

Apparently they left the hood pins unengaged, and as soon as Anthony entered the track, the hood smashed flat against the windshield.  Thankfully the window didn’t break, so he got to make a full parade lap looking through the gap under the hood. We borrowed a hinge from Evan’s Miata, but we got the hood fixed. (In the tests, the bent hood probably affected the coefficient of drag, but at least it was the same for all of the testing.)

After we got the hood fixed, we had a 90-minute delay due to thick fog. So it wasn’t until nearly noon that we got out first test done, and then the track closed for an hour to break for lunch. So here it is 1:00 pm, and we haven’t done shit yet.

The other thing that’s special about Watkins Glen are the steel guardrails, or Armco. You see these on the sides of roads, and on race tracks that don’t have sufficient runoff. Watkins Glen paints their Armco in a distinctive light blue color that’s not dissimilar from Gulf livery.

Cool little 914 in Gulf Livery at Watkins Glen

Whenever someone crashes at Watkins Glen, it’s a long delay. WGI doesn’t do hot tows, and so they have to close down the track to bring out the tow vehicle and clean-up crew. The Armco virtually guarantees that any crash is a wreck. I know this first hand – I came together with an E30 in an AER race and we bounced off both the inside and outside walls of Turn 6. Both cars were kaput for the rest of the weekend. Armco is not kind.

In a weekend that started with delays, we really didn’t need any more delays. But when you mix a wet track and steel guard rails with impatient drivers in race cars, you get more wrecks and more delays.

Those delays meant we didn’t get to test everything I brought. Most significantly, I didn’t test a stock front end vs R-package lip vs airdam vs airdam and splitter. I also wanted to remove the mirrors and see how much that affected drag, do more open-top tests, and other etceteras. Maybe next time. Maybe never.

If you want all the details, check out my site,, but I’ll recap the high points of the test here.

Really, go check out the site. This post has only a small fraction of what is there.

Open top

I see a lot of convertibles with rear wings: Miatas, S2000s, Corvettes, etc. I’ve often wondered about the effectiveness of a wing with an open top, and now I can answer that question. On my Miata, the open top generated the least downforce and reduced the effectiveness of the rear wing a lot – to the tune of 2.5 times less downforce than an OEM hard top. But that doesn’t mean you shouldn’t use a wing with an open top. If you run simulations in OptimumLap using my data, the open top with a 9 Lives Racing wing beats any combination without a wing every time.

Chop top

The Treasure Coast “Chop Top” is a partial mold from the OEM hard top. It’s primary purpose is to enclose the cockpit so that you don’t have to wear arm restraints when racing. It also helps aerodynamics slightly by reducing drag and lift. When compared to an open top or OEM hard top, the chop top is slightly faster.

However, once you add a wing, the Chop Top performs barely better than an open top. This is interesting, because you’d think airflow over the roof is considerably smoother than an open top. However, it’s what’s happening on the underside of the wing that’s more important, and the Chop Top roof can’t defeat the turbulence coming from the open sides of the cockpit and going beneath the wing.

I enclosed the sides of the Chop Top, and that makes it about the same thing as using an OEM hard top without a rear window. And so if you have a hard top and you’re not using a rear wing, removing the rear window will make you go a bit faster. Likewise, if you have an OEM hard top and a wing, don’t remove the rear window.

OEM hard top

The OEM hard top generated more drag and lift than expected from published data. This is likely due to the open windows and wide canopy, which turns the cabin into a parachute. The drag is supposed to be around .38 with closed windows, but we measured over .5. Lift is also supposed to be around .30-something, and we measured in the .5s again. Sucky.

All told, if you run simulations in OptimumLap using numbers from the test, the OEM hard top is only a bit faster than an open top, and actually gets beaten by the Chop Top.

However, once you add a wing, the OEM hard top wins by a lot. It’s all about getting clean air to the wing, and most importantly, beneath the wing, and the hardtop crushes them by a wide margin.

Adding AirTab vortex generators reduced the effectiveness of the wing by about 20%, and increased drag substantially. If I can save just one poor soul from adding vortex generators to their top, this test was kinda worth it.

DIY Fastback

My fastback uses the Chop Top for the roof, to which I attached a long sloping back. It’s quite light, weighing about 17 pounds less than the OEM hard top, and when you consider there’s no trunk, it’s lighter still. But weight isn’t why I made a fastback, I did it because it looks cool.

I really had no idea how well it would work, and I don’t mind saying I’m pretty impressed. The fastback beats up every other top and takes their lunch money. Compared to the OEM hard top, the fastback made 20% more downforce with the wing. I suspected that cleaner air to the wing would help, but I didn’t imagine it would be that much.

The fastback also reduced drag by 15%, which not only helps top speed, but fuel economy. Combined, the downforce and drag created a lift/drag ratio that was 50% better than the OEM hard top with a wing.

With the wing removed, the fastback was less impressive. It still beats all of the other tops in endurance racing simulations, but the lap times weren’t that much faster. Also, every top with a 9 Lives Racing wing beat the fastback without a wing.

9 Lives vs Cheap Wing

I tested two wings, a 9 Lives Racing “Big Wang” and a cheap Chinese double decker wing. I had to modify the cheap wing a lot to make it work. Nevertheless, the 9LR wing simply outperformed the cheap wing in every way possible, and it only contributed .03 to the coefficient of drag. The cheap wing, on the other hand, was like dragging an anchor.

There is some reason to use one of the cheap wings, however. At autocross speeds, where drag is inconsequential, the wing helps. I ran a simulation using the 2010 SCCA Solo Nationals West course, and the double decker wing was three-quarters of a second faster than without a wing. The 9LR wing was another .5 seconds faster than that, but still, a cheap wing is better than nothing.


I intended to try four different front ends. OEM, R-package front lip, Supermiata style airdam, and the airdam with a 4” splitter extension. I already mentioned the many delays, and so the only front-end test I got to do was airdam vs airdam and splitter. The splitter made more downforce, adding .38 to the coefficient of lift, and reduced drag by .01. It’s clearly a win-win situation, use one.


I keep referencing OptimumLap simulations because it’s the best way to use the comparative data. Nobody drives every lap exactly the same, the track changes ever lap, and so does the weather. In the end, it’s hard to quantify real-world lap times. Case in point: we had an 11 mph headwind at one point, and that totally skewed the data until we corrected for it. If we hadn’t run environmental sensors, I’d be telling you to buy a cheap double-decker wing. As it is, I’m telling you not to.

I put the various aero combinations into OptimumLap and ran endurance racing simulations at Watkins Glen. You can find those here. Watkins Glen is a high-speed track and drag matters more here than just about anywhere else. If I re-run these simulations at different tracks, the margin between the various configurations are a bit closer, because drag factors into it less.

You still suck at aero

If I come off sounding like I know a lot about aerodynamics, it’s just me regurgitating various things I’ve read. I still suck at aero, but I’m learning as I go. When I look around at other amateur race cars, I see a lot of other people suck at aero as well.

Here’s a quick recap of dumb shit I see all the time.

  • Exposed front tires are a large source of drag. Cover them.
  • Splitters without dams.
  • Splitters that are too flimsy. They should be able to support your body weight.
  • Wings with too much angle. The roof creates downwash, and if that angle combined with your wing angle is more than about 10 degrees, you’re making the wing stall. That means less downforce and more drag. A wing at zero degrees still creates a lot of lift and not much drag. Check it out, planes fly around like that!
  • Wings assembled incorrectly. I saw a Lemons team that had the wing on backwards. Not intentionally, but because it came from China that way. For realz.
  • End plates on the wrong way. The low pressure zone (the important part) is low and forward on most wing shapes. Cover that part with the end plates.
  • Wings set too low. If air can’t get underneath the wing, it isn’t a wing. Get it roof height, at least.
  • Dual wings with gaps that are too large, don’t converge, aren’t adjustable, or are otherwise defeating the purpose of the second element.
  • Cockpit venting done wrong. This is things like installing vents at the base of the rear window, thinking that air will go out. It goes in.
  • Removing weight at the expense of aero. This one is aimed at my brother, Ian. He was chasing weight and enlarged the openings in his front windows and removed the rear windows on his Yaris. This made his hatchback into a drag chute, and we lost 4 mph on the front straight at Thunderhill.

I admit to being (a) curmudgeonly and (b) skeptical about aero. The curmudgeon in me hates race-inspired cosmetic enhancements from fake factory air ducts to stick-on vortex generators. The skeptic in me wants someone to “show me the data”. It’s difficult to model the cost/benefit of various aero components when it’s so difficult to measure exactly what they are doing. You need really expensive equipment and someone who knows how to operate it. So Mario hired someone with his own money and actually got the data. Fuck’n-A.

I’m no longer a skeptic. I still hate ricers though. And while the benefits of a sorted aero package are absolutely clear, I prefer cars with terrible aero. Back in the 60s, sportscars looked like WWII airplanes, and in my mind that’s what they should look like. I don’t give a shit if the shape has a CoD of 0.5 and generates lift. Now you may be wondering why I race a Toyota Yaris, which has a pretty clean CoD and dog-shit looks. Half of that answer is that it’s the cheapest car to run. The other half is that beating the snot out of other racers is more fun when you have no business doing it.

Software review: Optimum Lap

Not long ago I wrote some software that would predict how fast a car can lap a track given vehicle data (mass, power, grip) and a track model. I was reluctant to distribute it or even work on it because it wasn’t something I wanted to support. Thankfully, I discovered Optimum Lap, which does everything my software did plus more. This free piece of software is really fun to use once you understand the interface. Optimum Lap let’s you design a car with just a few parameters.

  • Mass
  • Power (entered as torque at various RPMs)
  • Tire grip expressed in G
  • Tire radius
  • Transmission and final drive ratios
  • Drag coefficient
  • Frontal area
  • Down force

Here’s a screenshot showing how I entered data for my B-Spec Yaris. I’ve selected the driveline model. This shows that the gearing isn’t matched well with the engine. Either the gear ratios need to be closer together, or the engine needs to rev higher. Well, it’s not like a Yaris was designed for racing, so that’s understandable.

Optimum Lap has a track database containing a several common tracks, but it turns out that it’s not very difficult to make your own with Google Maps. That’s good because at least in one case, their tracks are really bugged (Laguna Seca). Here’s an example of a couple tracks I made (New York Safety Track and Pineview Run).

Mario and I have been playing with this quite a bit over the last couple months as a way to imagine what modifications we’ll make to our race cars or fantasize about what my next car will be. Let me take you down that highway a little.

Next Car?

I used to have a 1986 BMW 325es. I loved that car, but it was getting a bit old so I gave it to Ben and his mates to turn it into a racecar. I got to drive it again at Carolina Motorsports Park and it was like putting on an old pair of jeans: comfy. I’ve driven a bunch of cars on track since then, but I just keep coming back to wanting a BMW 3 Series as a track car. But which one? What does Optimum Lap say about each?

Optimum Lap predicts an e30 325is will lap Thunderhill in 2:15.67. The e36 is about a half second faster at 2:15.10. There’s no way I could get an e30 M3, so let’s look at later models. The e36 M3 is at 2:12.45 while the e46 M3 is at 2:10.59. How fast do I need to go? I actually don’t care that much about speed. Any of these will be fine. But what’s interesting is what happens when you change tracks or change parts of the car.

Pineview Run

Mario lives in New York not too far from Pineview Run (PVR), which YSAR reviewed last Summer. PVR is a twisty little driver’s track that is an absolute hoot to drive. While the distance is much shorter than Thunderhill, the lap times are even more compressed. The difference between the E30 325is and E46 M3 isn’t the Thunderhill 5 seconds, but 0.63 seconds. Over-rotate and screw up your steering correction and it could cost you 3 tenths of a second. It’s easy to make up for those mistakes with power at Thunderhill, but not PVR. That means lap times at PVR are going to be more about driver than car. Speaking of lap times, PVR recently announced their own time trial series, and it has a unique classing system: tire rating only. There are 3 classes, R-comps, 200+ treadwear, and 300+ treadwear. The differences among the cars become even less when tires get harder. In my simulation, the difference between an E30 325is and an E46 M3 is just 0.31 seconds. If you want to know who has the fastest car, mount up some R-comps and go to Watkins Glen. If you want to know who is the best driver, go to PVR with 300 treadwear street tires.

Final Drive Swap?

After I gave my E30 away, one of the performance modifications the new owners made was to put in a higher final drive ratio. I considered doing the same thing for my Yaris. The immediate change is that you get more acceleration in lower gears. That might help for drag races at the stop light, but the reasoning for doing it on a race track is that you can use the higher gears where the ratios are closer. That way you can be at optimal power more often. In order to do this, I would have to source a Scion xB transmission, crack it open, and take the final drive gear and swap it into my Yaris. Sounds like a good deal of labor. What does Optimum Lap say it’s worth? 0.03 seconds at Thunderhill. That’s the equivalent of losing 9 lbs. It wouldn’t even be noticeable and yet I’ve heard drivers who swear it’s the best thing to do to your car. Whatever.

Weight Matters. Power Matters More

One of my recent performance goals for the Yaris was to lose 200 lbs. That’s worth about 1.15 seconds. Pulling another 200 lbs out of a car isn’t exactly easy, but it would be worth another 1.26 seconds. You know what else is worth 2.41 seconds on this car? 15.3 hp. It’s a lot easier to add power than remove weight. Swapping in a Corolla motor would give me 30 hp and a drop of 4.27 seconds. This is why many of the successful endurance racing teams do motor swaps or forced induction.

Grip Matters Most

One of the best ways to improve lap times is by increasing grip. This can be done with tires or downforce. While both parameters are available in Optimum Lap, I don’t really know how to estimate aerodynamic parameters accurately. So I’ll focus on tire grip. Assuming tires have a grip of 1.0g, increasing to 1.1g reduces lap time by 2.1 seconds and decreasing to 0.9g increases lap time by 4 seconds. The more power you have, the more you need grip to take advantage of it. If everyone raced on 400+ treadwear tires, lap times would be more about driver skill than power:weight ratio. If I was ever going to design a racing series, that’s the first rule I’d make. Cars would be cheaper to run and budget builds would have a better chance against expensive builds.

Winning B Class: part 1, fuel consumption

When I built my 2007 Toyota Yaris for racing, I aimed it at the SCCA B-Spec rules. Only after competing in an SCCA sprint race did I realize that sprint racing is not for me. It’s much more expensive per hour, the “win every corner” mindset makes it more dangerous, and it’s lonely not hanging out with a team. With that in mind, the decision was clear: re-build it for endurance racing. Sadly, it’s a little too slow for most applications. In Lucky Dog, it’s slower than most class C cars (when it doesn’t get protested for being too new). In ChampCar, the build is 120 out of 500 points so there are plenty of points to work with. But in order to compete it would take an engine swap or forced induction. Given that I want to keep it emissions legal in California, these options are mostly out of the question. Neither World Racing League nor American Endurance Racing league run events out West, so the target is Lemons. In 24 Hours of Lemons, it would probably be placed in class B. Could we win the B class with a little luck and a lot of planning? Well, this post is the first in a series where we document our efforts.

So what are our advantages? Reliability and economy. Unlike half of the cars in B class, we have a very good chance of running the whole 14.5 hours of a typical race (8 hours on Saturday and 6.5 hours on Sunday). However, we will be competing against much faster cars. We need to be on track as much as possible. This means zero black flags, of course, but it also means as little time as possible in the pits. In fact, we’re hoping to cut out one pit stop.

Most endurance driving stints are 2 hours or less. Lucky Dog and ChampCar actually limit drivers to 2 hours. Lemons has no such rule. However, most cars burn fuel fast enough that they pit between 1.5 and 2 hours. That means that a typical team will run 4-5 stints on Saturday and 3-4 stints on Sunday. I believe our best chance to win means driving only 3 stints on each day. The question is, can a single tank of fuel last 2 hours and 40 minutes on Saturday?

Our previous racing at Thunderhill, Laguna Seca, and Buttonwillow shows that the Yaris burns about 4 gallons per hour. With its 11.1 gallon fuel tank, it should be able to run 2:45. That’s no problem for Sunday but Saturday could be. If our calculation is off by 10%, we might find ourselves running out of fuel, and there would be no chance of winning if that happened. So we need to figure out how to extend our range.

The simplest answer is to install a fuel cell. That would instantaneously solve the range problem but would bring up new problems. They’re expensive. It would require removing the stock fuel tank and fabricating a new structure. The car would also no longer be street legal. The center of gravity would be higher. Too many negatives, so I’m not getting a fuel cell. Lemons does not allow one to modify OEM fuel fillers, so I can’t increase capacity with a fat intake tube either. So if we can’t increase fuel capacity, we’re going to have to increase efficiency.

Economic driving

Who knows how to get the most miles from a tank of gas? Hyper-milers. I’m sort of a closet hyper-miler myself. On the street, I often drive under the speed limit, conserve as much momentum as possible, pump my tires up pretty high, and draft trucks on the highway. I don’t go as far as making aerodynamic improvements though. But we will on the racecar. However, that’s a topic for another day. Today we are going to consider the act of racing more conservatively. The driving can’t change so much that we do more harm than good, though. We have these two connected questions to consider.

  1. How much fuel do we save by changing our driving style?
  2. What style of driving optimizes our chance of winning?

To answer these questions, I’ll be using Assetto Corsa, Brands Hatch, and the NA Miata. While I do have a Yaris model for Assetto Corsa, I don’t think it’s very accurate. The NA Miata is one of the highest quality models and besides, Miata Is Always The Answer. The car is loaded up with 5 liters of fuel, “Street” tires at 30 psi, max camber, and zero toe.

So let’s define a few different driving styles.

  • Hard – Hit the brakes hard. Hit the throttle hard. Steer like a mad man. In slow, out fast. Brake in a straight line. Shift at red line (7k). Lots of amateur racers drive like this, especially those in powerful cars. Clearly we’re not considering this, but I wanted to investigate the efficiency of a typical sucky racer. Intensity 9/10ths. Intelligence 3/10ths.
  • Soft – Conserve momentum as much as possible with early apex lines. Coast slightly before braking zones. Shift at 6K and choose a higher gear if there’s any question. Steering corrections are unnecessary driving like this. Intensity 5/10ths.
  • Enduro – Drive fast but with a lot of margin for error (not much yaw). Shift at 6.5k. Use lots of trail-braking but only a little brake-turning. Intensity 7/10ths.
  • Sprint – Drive faster with plenty of yaw. Still keep some safety in reserve. Shift at 7k RPM. Intensity 8/10ths.

The most interesting finding for me was that Soft driving increased fuel economy by an amazing 40% over Hard driving while having nearly identical lap times. My typical Enduro style results in decent fuel economy and speed. I’m only about 1% off my Sprint pace but my economy is up 13%. Compared to driving Soft, Enduro is 3.3% faster at the cost of 22% less economy. So which style is best for endurance racing? Is it better to drive slowly to get a tank of gas to last 148 minutes, drive as fast as possible while only getting 103 minutes, or something in between?

Style Laps Fast Median Laps Minutes
Hard 10.3 63.96 64.18 98.88 105.8
Soft 14.5 63.84 63.97 139.2 148.4
Enduro 11.9 61.31 61.91 114.2 117.9
Sprint 10.5 60.84 61.29 100.8 103.0

The track is live for 480 minutes on Saturday. But not all of those 480 minutes are hot. Lemons does live towing and when there are several tow trucks on track at once they will fly full course yellows. Sometimes that goes on for 5 minutes and sometimes for 30. I recall one race where they threw a red flag and I waited nearly 20 minutes with the engine off. It’s hard to predict how much of the 480 minutes are green and how much are yellow. So we need to investigate what happens with 10 to 120 minutes of yellow flag time, which results in 470 to 360 minutes of race time.

The next thing to consider is how many driving minutes there are. The car isn’t lapping when it’s in the pits. My calculations use a pit stop time of 10 minutes. It doesn’t take that long to fuel a car and change drivers, but Lemons pit stops occur in the paddock, outside the timing loop on the track. So every time you pit, you lose 1 lap in addition to transit time.

Taking into account lap times, fuel burn, yellow flag time, and number of pit stops, we arrive at the table below. I have highlighted the driving style that produces the most laps in red.

Yellow Hard Soft Enduro Sprint
10 401 412 426 420
20 392 403 416 411
30 383 393 407 401
40 373 393 397 391
50 364 384 387 381
60 364 375 377 372
70 355 364 368 372
80 345 356 358 362
90 336 347 348 352
100 327 337 339 342
110 317 328 329 332
120 308 318 319 323

When Enduro beats Sprint, it does so by 5.67 laps on average. Conversely, Sprint beats Enduro by 3.67 laps on average. The difference comes down to how many stints there are. Enduro sometimes runs one less stint, and when it does, it has a huge advantage. It doesn’t impede lap times that much and has the added benefit of reducing fatigue and the chance of a black flag (which pretty much guarantees we won’t win B class). Driving Soft never wins. It can be as much as 10 laps better than driving Hard, and there are a few situations (highlighted in blue) where it is better than Sprint. But it never beats Enduro. There isn’t much point in driving super Soft. The hyper-miler in me wanted that to be useful, but it isn’t.

Telemetry or it didn’t happen!

The line colors are:

  • Red Hard
  • Blue Soft
  • Green Enduro
  • Black Sprint

The panels from top to bottom are:

  • Brake pressure
  • RPM
  • Speed
  • Steering angle
  • Throttle
  • Time delta

Click on the image to open it in a new window and then follow along with the text below.

There are 4 braking zones (top panel). In Soft style I only applied brakes in 2 of these. Note how low the RPMs are in general. You can also see long periods of coasting in the 5th panel (throttle). But the speed graph (3rd panel) isn’t that terrible. Driving economically is a kind of intellectual challenge, which is why I hyper-mile in real life. I have to do something to make street driving entertaining.

Hard style sees me sawing the fuck out of the steering wheel (4th panel) and mashing the brake and throttle pedals mercilessly. The brake trace (top) shows early and hard application of the brakes followed by no trailing pressure. Just on/off. It’s not a fast or economical way to drive.

There isn’t a huge difference in driving style between Enduro and Sprint. I use more brake pressure in Sprint mode to turn the car and I also choose a lower gear in a couple of places. I consciously take an earlier apex line in Enduro to favor momentum over engine.

Tire pressures don’t matter

I remember reading a recent article comparing 200 treadwear tires and one of the initial concerns was setting tire pressure. Shockingly, they found that varying tire pressures had little affect on lap time. Whoa there! I did not spend good money on a needle pyrometer for no reason! Did I? Did I?

Clearly this is something YSAR needs to investigate. In theory, raising tire pressures does several things.

  1. Decreases rolling resistance
  2. Decreases grip
  3. Improves steering feel

I can imagine that these forces offset each other to some degree. Straight speed vs. corner speed: it’s 6 of one, half-dozen of the other. It makes some sense that tire pressures might not change lap time by much. But making sense isn’t the goal here. I’m a scientist by profession and passion, so I just have to conduct some experiments. Since I don’t have immediate plans for a semi-private test day, I’m testing this in simulation first. Later in the year I hope to revisit this study on a real track.  Let’s begin with the usual sim testing environment: Assetto Corsa, Brands Hatch Indy, NA Miata.

Experiment #1: Ideal tire pressure

In order to remove any human sources of variability, I’m going to let the AI drive first. Assetto Corsa sets the Miata pressures at 28 psi by default and allows a range from 15-40. I chose to change pressures in 4 psi increments. As you can see in the table below, 28 psi seems optimal. Interestingly, all laps are within 0.25 seconds using pressures from 24-40. If I had seen these numbers in real life, I would probably conclude that all lap times were roughly equivalent. But the AI drives each lap within hundredths of a second, so the differences are real, though small. Overall, I have to agree with the initial premise: tire pressures don’t affect lap time very much.

Front Rear Seconds
16 16 65.41
20 20 64.68
24 24 64.32
28 28 64.09
32 32 64.26
36 36 64.29
40 40 64.34

Experiment #2: Asymmetrical tire pressure

One of the things I like doing at the track is running non-square setups. I’ll mount completely different tires on the front and the rear. The two ends of a car are doing very different things, so there’s really no reason to run square setups. One of my favorite ways of goofing around on a skid pad is to mount sport tires on the front and all seasons on the rear. That’s a good way to train your oversteer recovery skills! Note that I said skid pad not HPDE session. I don’t think it’s a good idea to mess around too much in the presence of other drivers on a fast track.

So what happens when the AI drives a non-square setup? As it turns out, Assetto Corsa doesn’t allow you to have different compounds for the front and rear. But you can change individual tire pressures.

My first thought was to change the psi by 4 lbs on either side of 28. So 24F 32R and 32F 24R. The faster combination was to have more pressure in the rear. It wasn’t much of a difference, so I decided to go extreme and set one pair of tires to the ideal 28 psi and the other to 40. The result is sort of shocking. 28F 40R (64.04) is not only faster than 40F 28R (64.41), it’s also slightly faster than 28 square (64.09).

Front Rear Seconds
24 32 64.22
32 24 64.33
28 40 64.04
40 28 64.41

A stopwatch doesn’t give many details, so let’s load up the telemetry and take a closer look at what’s happening in Experiment #2. Green is 28-28 (because green is in the middle of the rainbow). Red is 28-40 (because oversteer feels red). Blue is 40-28 (because understeer feels blue).

For some reason, the AI chooses a different line on the square setup. The green line shows that the AI attempts to hold too much speed which results in being later to throttle. While initially faster, this ultimately causes the square setup to lose nearly 2 tenths by 1800 feet. It maintains that loss for a little while but then recovers most of it by the end. Apart from one bad decision in one corner, the square setup is actually faster everywhere else. This is why we don’t rely solely on the stopwatch.

What’s happening with the understeer and oversteer setups? The reason the oversteer is faster is that it’s able to use more mid-corner throttle, and it gets to full throttle sooner. It also has more yaw early and requires less steering effort in a few places. You have to zoom way in to see this. These are very subtle differences, but they add up to 4 tenths of a second by the end.

Experiment #3: Human driver

OK, time for me to drive. The first thing I did was run some square setups at a couple different pressures. There’s a little difference in the way they feel but not that much. I’d rather focus on what happens when you run different pressures in the front and rear.

Front Rear Fast Median M – F Cuts
28 28 60.93 61.25 0.32 0
28 40 61.80 62.26 0.46 1
40 28 61.25 61.36 0.11 0

The fastest was the square setup. That’s not really surprising. What is surprising was that the understeer setup was very close. The median lap was only 0.09 seconds off. If you look at the difference between the median and fast laps (M – F) you can see that the understeer laps have the most consistent pace. That was my impression while driving too: “oh well, another uneventful lap”.

The big shock is how bad the oversteer setup was. Its fast lap was 0.55 seconds slower than understeer and the median is even worse: 0.90 (some of the laps were not pretty). I was having to make steering corrections in nearly every corner as the back stepped out under braking and also under throttle. I also had one lap where I went a little too much off course and got a cutting violation.

In the graph below, the panels are speed, steering angle, throttle, and time. I have plotted the top 5 laps of each run. As you can see from the red steering angle trace, the position and magnitude of the steering corrections are quite variable. This indicates that an oversteering car is hard to drive consistently (and possibly also that I suck at racing).

Let’s take a closer look at the fast laps to dissect how understeer and oversteer affect driving style. I’ve zoomed in on the first corner (a fast, descending right-hander) below. Again, the panels are speed, steering angle, and throttle from top to bottom. The area under the blue steering angle trace is relatively large. I’m having to crank the steering wheel quite a bit because the front of the car is sliding (understeer). On the green trace, there is very little steering because the rear is stepping out just a little. This is what Paul Gerrard calls zero steer. On the red trace, the back has stepped out so much (oversteer) that I have to make a steering correction in the opposite direction to prevent myself from spinning. Note that the green trace also has a steering correction (it’s bowed down in the middle), but it is very mild.

Looking at the throttle trace (bottom panel) you can see the disadvantage of the understeer setup: it’s late getting to full throttle. So in addition to the loss of speed from scrubbing the front tires, it has an additional opportunity cost in throttle time. The oversteer setup should get to full throttle first because it’s pointed straight first, but I’m fighting the wheel so much I don’t manage it. A better driver could make this work better than me.

Here’s the whole graph. Note that the understeer setup isn’t always the last to full throttle. Sometimes the initial application is delayed. But once applied, the throttle can be used as an on/off switch. You don’t really have to balance the back end when the back end isn’t sliding. In contrast, the oversteer setup requires a soft foot and live hands to keep it on track.

Tire pressures do matter

The AI was relatively unfazed by non-square changes in tire pressure, but I was not. Having a loss of grip specifically on one end of the car or the other completely changed how I drove. I can sum up the driving experience as follows:

  • An understeering car
    • feels boring
    • requires a lot of steering effort
    • requires trail-braking to rotate
    • requires patience before throttle
    • may see you running off track at the exit
  • An oversteering car
    • feels exciting
    • practically turns itself
    • requires steering corrections to prevent over-rotation
    • requires throttle modulation
    • may see you spinning at the entry, middle, or exit

Why is the AI behavior (oversteer fast) so different from mine (understeer fast)? I’m not sure exactly what to take away from the AI driver. It’s several seconds slower than me and doesn’t even know how to trail-brake (data not shown). The AI sucks at racing. However, it is very good at controlling oversteer. Its steering corrections are always exactly the right amount. I don’t think we should read too much into the AI performance.

Although I set out to determine if tire pressures affected lap times, what I ended up focusing on was how tire pressures affected grip balance. Why? Because the handling of the car is what will ultimately dictate lap times. Too much oversteer not only results in a car that is difficult to control, it’s also slow. But what of too much understeer? It’s a little annoying but can be mitigated by trail-braking. Ultimately, it’s easier to deal with a little extra understeer than a little extra oversteer. For many inexperienced racers, the natural reaction to stuff going wrong is to lift off the throttle. If the car naturally understeers, the stuff is mostly understeer and lifting is the appropriate response. In an oversteering car, lifting is going to make matters worse.

Going Forward

All of the experiments here depended on the Assetto Corsa tire model. How accurate is that? No idea. I don’t think of these experiments as the end of anything, but rather the seeds for the real-world tests I’ll do later in the year. Stayed tuned (pun intended).

Endurance tire testing: part 2

Before getting to the latest post, I want to remind people that there is a YSAR author contest with the top prize being a Rumblestrip DLT1-GPS lap timer. For details, click the Contest link at the top of the page.

Last week I blogged about how my brother and I tested a few tires to find out which one was the best. What exactly does best mean? Lap time certainly matters, but also consistency. We do endurance racing, not time trials. So our ideal tire is one that gives the driver the confidence to lap consistently fast. I’ll summarize Mario’s impressions from last time.

  • RS-4 is the best tire because it feels best and records fastest laps
  • RE-71R is a good tire, but hard to drive consistently fast
  • RT615K is a good rear tire, but it lacks grip and feel on the front
  • R1R is soft and feels weird

This week, let’s take a look at telemetry traces and see if we can get a little more resolution on the differences among the tires.

RS-4 (first run)

The two laps in the graph below have nearly identical lap times (1:37.4). But they are really very different laps. Looking at the time difference in the bottom pane, you can see that one lap gets ahead of the other by about 0.25 seconds. Then it loses all that time and ends up 0.12 seconds behind. While the fastest lap recorded was 1:37.4 it could very easily have been a high 1:36.


Here are 3 laps on the R1Rs in the same 0.2 second span (1:38.3 – 1:38.5). Again, while that sounds pretty consistent, it isn’t. Mario drives the first and second halves of the course very differently. The first half of Thunderhill West has more compromises and high speed corners. The second half features several hairpins. He drives the hairpins very consistently, but not the high speed corners. Why? Maybe he has better braking markers in the hairpins? In any case, when trying to figure out which tire is best, we have to take into account each corner, not the lap time.


The 3 laps graphed are all 1:38.1X, so amazingly close together. I haven’t plotted the fastest lap here. Overall, the laps look more consistent than the R1R laps. Is that because the driver is getting more accustomed to the track or because the tires give better feedback?

RS4 (second run)

These 3 runs are within 0.1 seconds of each other. The fastest lap was not plotted. Overall, consistency is much better.

All runs combined

Overlaying all the runs, you can see just how much variation there is in first half of the track. Mario felt much more confident on RS4s, and this translates into braking much later. This produces a transient time gain that is partially lost by braking a little too deep. These are bumps in the time graph at the bottom relative to the fastest blue lap. The corner where RS4s appear to help the most is T6 through T7. Here, the blue lines pull away from the others (see bottom time lost). Once in the hairpins, RE71Rs appear to be just as good as RS4s despite having narrow tread and wheel widths.


So what did we learn? For one thing, laps that look the same from the perspective of a stop watch can be really different in detail. There’s too much variability in the first half of the lap to say much about the relative grip of the tires. We can say, however, that the feel of the tire matters very much to the driver. In the second half of the track, where comparisons are more robust, RE71Rs may be slightly better than RS4s. Despite having only one session on RE71Rs, he drove them at least as well as the RS4s.

Let’s finish this off with a few bullet points

  • Because feel is such an important characteristic, you really should try a few tires rather than settling on what is cheap or convenient.
  • It’s probably easier to fit tires to the driver rather than asking your driver to change their style to fit the tire.
  • Don’t rely on a stopwatch to tell you which tire is best.
  • Doing tire tests on a 5-run HPDE day with a driver who hasn’t driven the track in over a year isn’t going to get the most consistent data.
  • It was a great day of driving and data mining, and I can wholeheartedly recommend taking tires and timers to the race track.

Endurance tire testing: part 1

Before getting to the post this week, I want to turn your attention to the 2018 YSAR Author Contest. Also linked in the menu above. Write an article for YSAR and you might win a great prize.

I had an MRI recently that shows I have a herniated disc. So my back problems are pretty fucking real as well as being debilitatingly painful. As such, I can’t do any performance driving for a while. Good thing I have a twin brother who can step in and drive for me. In this case, it was a tire test day at Thunderhill West (follow link for instructional video). The car was my Yaris in mostly B-Spec trim but I recently upgraded the calipers to increase the brake pad choices. The Yaris has such a huge cargo area that we were able to fit several tool kits, 2 people, and 6 tires inside even with a full cage. I also have a custom tire rack that fits on a mini hitch that increases its capacity to another 4 should the need arise.

Here is a brief list of the tires we were testing. A more thorough description is given below.

  • Falken RT615K+ 205/50/15
  • Bridgestone RE71R 205/50/15
  • Hankook RS4 225/45/15
  • Toyo R1R 225/45/15

Five 20-minute sessions isn’t an ideal way to test tires. There aren’t that many runs, the runs are a bit longer than necessary, and there isn’t enough time to make a tire change in the middle of a run with a pit crew composed of exactly one gimpy 51-year old (me). So how do we figure out which is the best tire when the test conditions are so volatile? We have 3 methods.

  1. Lap times
  2. Feel of the car from the driver’s perspective
  3. Telemetry analysis

In part 1 of this post we’ll look at lap times and how the tires feel from the perspective of the driver. Next week I’ll present the details and show why telemetry is so important. Now let’s hear from Mario, whose contributions are in blue text.

I tried to stick with a pace that I would do in an endurance race, and not to find faster lines or experiment with driving style too much. The point was to go for consistent laps, and measure all the tires on a level playing field. Naturally the air and track temperature changed throughout the day, and as this was a HPDE session, I had to throw away some good laps to manage traffic. Also, I don’t know the track that well, and so my driving was certainly going to improve throughout the day.

Let’s talk about the tires in the order they were run.

Falken RT615K+ 205/50/15 15×7

Fast 1:40.9, Median 1:41.6

Our team has used a lot of different performance tires over the years (Bridgestone RE11A, RE71R; Dunlop Z1, Z2; Falken RT615, RT615K, RT615K+; Federal 595 RSRR; Hankook RS3, RS4; Hoosier SM7; Toyo RR, RA1; Nitto NT01, NT05, Yokohama S.Drive). Historically, some version of Falken RT615K has been our go-to tire. The reason for this is that it strikes a nice balance in expense, life, and grip. It’s not the fastest tire, but it is one of the more durable in the 200TW category. Given that we’re more cheap than fast, we like Falkens. We almost always mount these on 15×7 rims even though 15×8 (or even 15×9) are supposed to be faster. Why? No good reason. Possibly because Spec Miata uses a 15×7.

I’ve heard that the RT615K+ is made in the same plant as the Dunlop Z3 Star Spec. The difference is the tread pattern. Falkens are typically $10 cheaper per tire. If they truly are the same thing, the choice comes down to camber wear. Dunlops have a symmetrical tread pattern, which means you can flip the tires on the rims, which may extend tread life quite a bit. If you read reviews on the RT615K+, people say they get greasy if you run them hard. That’s not a bad thing if you ask me. If you’re driving the limit, any tire will get greasy.

The tires we had for the test day were mounted on 15×7 Kosei K1 rims. They had seen quite a bit of track action, but only as rear tires on the Yaris. That means they were hardly worn at all. For the rears, we had a set of old RE71Rs (see below). Let’s hear what Mario has to say about the RT615K+..

I’ve driven the Falken Azenis 615K probably more often than any other tire, and they are the gold standard which I measure everything against. I hadn’t tried the 615K+ yet, so I was happy to go out on the Azenis first and see what all the plus was about.

My experience with them is that they warm up quickly, and usually the fastest lap is the second lap, but then they get a bit greasy when hot. Traction drops off a bit then, but stays at that level forever. And so they felt the same as always, with good audible feedback and a generous traction limit that doesn’t suddenly go away.

However, after trying the other tires, the gold standard is now the old standard. By comparison, the turn in was vague, and they simply don’t have as much stick. Perhaps on a 8” rim they would have worked better, but I doubt an inch of rim width was going to be the night-and-day difference I’d experience with the other tires. Later in the day we’d put these on the rear, and for that, they are my tire of choice.

Hankook RS4 225/45/15 15×8 (first run)

Fast 1:37.4, Median 1:38.4

The RSR has become one of the most popular endurance racing tires. One reason is that much of the competition has reengineered their tires for the larger autocross market where grip is more important than longevity. RS4s are a more traditional 200 TW tire that last a long time.

The tires used in the test were mounted on 15×8 TR C1 rims. The tires had been used in a previous Lemons race and had about half of their tread remaining. The rear tires were the same as the test above (old RE71R).

The second time I turned the wheel I knew I was on a totally different tire. I was initially a little bit nervous because the steering was so different than the Azenis, but the tires warmed up quickly, and I to them: super accurate turn in, great feedback, and you can hear them working. I put down a few consistent laps and was surprised to see they were over 3 seconds faster than the Azenis!

That’s pretty astounding considering these were back-to-back sessions an hour apart. The track and weather conditions were probably as similar as they were going to be, and I don’t think my driving line or technique changed much from the previous session.

The RS4s seemed to take a bit longer to come in than RT615K+, and were fastest on the 4th or 5th lap. After that they seemed to fall off about a quarter second.

Toyo R1R 225/45/15 15×9

Fast 1:38.3, Median 1:39.3

Originally a 140 TW tire, Toyo later rebranded the R1R as a 200 (probably to get more sales). Magazine tire reviews consistently report that the R1R is a pretty soft tire that wears quickly. It’s supposedly really good in the rain. It has one of the more interesting tread patterns. The brand new tires in the test were mounted on 15×9 Konig Dekagrams. Rear tires were as above.

I thought these were going to be the fastest, because they were the only ones on 9” rims, and they are basically rebadged 140 TW tires. But I just didn’t have much confidence under braking. They didn’t have as much audible feedback, and even through the wheel and pedals I never knew what they were doing. Any corner which required some initial braking cost me time, and while I thought these were the fastest tires around T2 (I could floor it all the way around), it was simply because I didn’t have the confidence to go through T1 faster. The run up to T7 requires the longest and hardest braking, and I felt like I was going to flat-spot them every time.

When we pulled the tires off, they looked totally different. The way the rubber was melting off the tire made them look like they were much softer than the other tires. I’d worry about the longevity of these in an endurance race. However, they were also brand new tires with full tread, and probably got the hottest because of that. I hear that R1Rs are good rain tires, and that’s probably where these will be used now.

Bridgestone RE71R 205/50/15 15×7

Fast 1:37.1, Median 1:38.2

The RE71R is well known as a cheater tire because it’s more like a 100TW than a 200TW in grip and longevity. I’ve heard some stories of them lasting only a few hours. My experience is that they are actually more durable than other tires on my Yaris. Despite its low power, my Yaris has caused blistering and chunking on most of the tires it has seen. It’s pretty frustrating to see a tire with very little wear except the shoulder has been completely chewed away. RE71Rs don’t do that because they can handle the heat.

We had planned to use these on 8” rims, but due to some fitment problems, we had to use the ones that we’d been using as rear tires, which were on 7” rims. That meant we had to move the Azenis to the rear, and so this wouldn’t be an apples-to-apples comparison with the other rubber.

But that turned out to be not such a bad thing, as the RE71R front and RT615K+ rear made a balanced combination with neutral handling. I was able to rotate the car much easier and play with balance more effectively.

As such, the RE71Rs set down the fastest time of the day (so far), but I didn’t feel they were as consistent. I felt like the one fast lap was an outlier, and that I couldn’t drive them that way lap after lap. These were also the oldest tires, and I’m not sure the effect of that.

Hankook RS4 225/45/15 15×8 (second run)

Fast 1:36.7, Median 1:37.3

We went back to RS4 front leaving the 615K+ on the rear, and again the handling was very neutral, similar to the RE71Rs. Taking some traction away from the rear definitely helps me.

Compared to the RE71R, the RS4s instilled more confidence, and this might be down to simply the sounds they make. A better driver might go fastest on the RE71R (or maybe even the R1Rs), but I felt better on the RS4s. And they were more fun.

So much so that on the last two laps I decided to screw consistency and up my pace. I immediately dropped half a second and did a 1:36.4. On my second flying lap I looked at the RumbleStrip and saw I had another .3 seconds in hand and thought I had a sub 36 lap in there… But they threw the checker on me in T8, and so I didn’t get a chance to find out.

Conclusions Part 1

Although we had come to the track to determine which tire was fastest, one of the most important lessons we learned was how much feedback is important to the driver. A simple tire swap can make a huge difference in the way a car feels and consequently what performance a driver can extract from a car.

Mario drove the RS4s faster and more consistently than any of the other tires. He also felt more confident with RT615Ks on the rear rather than the stickier RE71Rs. Moving forward, we’re now considering new combinations of tires. Perhaps 245 width RS4s in the front? Maybe something in 195 width for the rears? More tests will follow, but not until my back heals or my brother visits again.

Next week we’ll take a higher resolution look at the data using telemetry and see why you should always run telemetry.

Power, grip, and aero in theory

Six months ago I did some simulator tests where I used Assetto Corsa to answer questions about the relative contributions of power, grip, and drag. I wanted to follow that up a little with something a bit more rigorous. So I took my driving inconsistencies out of the equation and had the AI drive the car. I did a bunch of experiments on a lazy Sunday using the original rFactor. That was fun and informative, but I’m not reporting on that today because I decided to write a program that simulates a car driving around a track. Why? Well, honestly it’s because I wanted to implement the various equations myself. Most of the math is pretty easy in isolation. Equations for acceleration, lift, drag, etc. aren’t too complicated. Putting them together sometimes is though. For example, as you increase speed, you increase drag. So acceleration gets worse the faster you go.

The Model


The track is modeled as a series of alternating straights and corners. The simplest description would look something like this.


This means a 2000 foot straight followed by a 200 foot radius corner with a 60 degree arc. You can chain together any number of straights and corners to create whatever track you like. The sections don’t actually need to connect in a closed shape. I decided to use Thunderhill as that’s one of the most popular tracks in the region. I used Google satellite images and scale bar to rough out the track. It comes out as 2.94 miles, which is pretty close to the actual length. Note that my track model doesn’t take into account elevation or camber (yet).


Under the assumption that the driving line is circular, corner speed depends only on the radius of the corner and the grip of the tires. This is made a little complicated by aero modifications that increase grip and speed by adding downforce, but only a little. Because the corner speed is constant, it’s trivial to determine how much time was spent in the corner.


Straights are somewhat complex to model because the vehicle increases speed for some time, and then brakes to arrive at the correct speed for the next corner. This calculation depends on initial speed, engine power, gearing, aerodynamic drag, frontal area, and grip of the tires. A simple way of thinking about it is that the total time is the sum of the accelerating time plus the decelerating time. The way I solve it is by binary searching the transition from throttle to brake. At some number of seconds the distance covered and the final speed will be correct: it’s just a matter of making refined guesses.


Since Miata Is Always The Answer, I decided to do experiments with a virtual Miata. People sometimes say “the answer is always Miata” but that would spell out TAIAM, which doesn’t mean shit. Let’s give some parameters on the typical Miata that we will vary to see how the lap times change.

  • 2300 lbs with all fluids including the driver. We’ll strip some weight out of this to see what happens. We’ll also add a little.
  • 100 HP. My model assumes an engine of constant output. I don’t take into account the torque curve or gearing yet. It’s best to imagine “100” as a placeholder for both torque and horsepower, and the value of 100 is not a very healthy example of the breed.
  • 0.40 Coefficient of Drag. A hardtop Miata with windows down has a drag of something like 0.4. But topless it’s worse, and you could always add theme and make it terrible. For reference, a Prius is below 0.3 but it would be hard to get a Miata that low. However, a Prius has a larger frontal area.
  • 0.0 Coefficient of Lift. I abstract the various aero components into a single item rather than wing, splitter, diffuser, etc. A wing can be flat surface made from plywood with a CoL of 0.75  or something wing-shaped with a CoL of 1.0-1.6. The default value is 0.0 but a base Miata probably has some lift.
  • 0.0 sq-ft wing area. I’m not sure how to convert the various aero surfaces into wing area, but 0-12 feet in 4 foot increments seem like a reasonable range. The default value is 0.0.



It’s probably no surprise that more power reduces lap times. This is especially true if you have an anemic engine. Adding 20 HP sees lap times dropping by 3.08 sec. Another 20 HP is 2.57 sec. While there are diminishing returns, there are significant benefits to 200 HP and beyond. What’s amazing about engines is that you can realistically have them vary over a huge range. A turbo or supercharged Miata can make 300 HP. It might not make a good endurance racer at that point. However one of the most successful Lemons cars is the turbocharged Miata from Eyesore racing. In a race situation, high HP is doubly useful because it’s much easier and safer to pass under acceleration than braking or cornering.


The more corners a track has, the more grip becomes the key factor in performance. On a circular track, grip would be the only factor (assuming you have enough power to drive a given speed). Even a small change in grip can make a large difference in lap time. For example, a change from 1.00g to 1.05g drops lap time by 2.77 seconds. If you look at the telemetry of different drivers in the same car, you’ll see some people can extract more grip than others, and I think this is largely why some drivers are a couple seconds faster than others. In this model, an all season tire is about 0.90 grip. Summer tires 0.95, 200TW 1.0, Semi-slicks 1.1, Slicks 1.2. While these figures may not be correct, it’s the relative difference that’s important. If you want to win, get the grippiest tire allowed by the rules. The UTQG rating is only a rough indicator of the grip. In the crowded 200 treadwear class, I’ll bet there’s more than 0.05g of variation, especially when you consider differences in rim widths and tire pressures.


There are two components to aerodynamics, drag and lift (three if you count aesthetics). Drag has a relatively mild effect on lap times. Slipstreaming the heck out of it won’t see more than 1 second improvement. Similarly, ruining your CoD to a tune of 0.5 won’t see you slower by more than 1 second. Of course, every second counts, but this is the least useful area to tune. However, cosmetically, not much says racecar more than a wing.

Because lift affects grip, and grip is incredibly important, an aero package that increases downforce has a reasonable effect on lap time. Simply adding a wing could see your lap times dropping by 1.3 seconds (this is the Ideal 4 column below). There is some drag associated with wings, however, and on a track that is more straight than corner, a wing may do more harm than good. Note that a splitter can both decrease drag and increase downforce, so not all downforce increases drag. While you won’t see huge improvements in lap time from aero, it’s a one-time cost, unlike tires, and a well made aero package could see you dropping 1-2 seconds.


Removing 100 lbs will see lap times dropping by only about 0.6 sec. Weight reduction appears to have a relatively minor effect because it varies over such a restricted range. It’s a lot easier to improve your power:weight ratio by adding horsepower than removing pounds. So weight reduction might not seem like it’s worth doing, but it is. Out in the real world, the relationship between load and grip is sub-linear, so dropping weight is better than the model shows. There are also gains to be had in component longevity and fuel economy. The simple weight loss associated with angle grinders is relatively cheap, but when you start replacing structural parts with lightened versions it gets costly.

Some example builds

Let’s close this out with some example builds and lap times. Note that for a variety of reasons, the absolute lap times aren’t exactly as you would see at the track, but they aren’t very far off. It’s more important to think of the relative differences.

  • First day at the track – untuned engine (120 HP), all season tires (0.90g), open top (0.45 CoD, no downforce), and a coach in the right seat (2500 lbs) = 2:33.20
  • Solo – as above, but with Summer tires (0.95g) and no passenger (2300 lbs) = 2:28:91 (4.29 sec faster than above)
  • Sport build – engine is mildly tuned (130 HP), 200 TW tires (1.0g), hard top (0.40) and enough weight reduction to offset the top = 2:24.25 (4.66 sec faster than above)
  • Budget enduro – 100 lbs of weight reduction (2200), an additional 5 hp (135), DIY splitter and wing (0.35 CoD, 0.75 CoL, 8 sq-ft area) = 2:21.14 (3.11 sec faster than above)
  • TT build – as above, but using stickier, wider tires (1.05g) and professionally designed aero (CoL 1.3) = 2:17.36 (3.78 sec faster than above). Now I’m sure you’re wondering if it’s the tires or aero. It’s mostly tires (2:18.44 vs 2:20.09).
  • Eyesore – a famous Lemons car with a ghetto-charged motor that was dyno’d at 197 hp. It’s light (2200), has 200TW tires (1.0g), and theme for aero (0.45) = 2:17.29.

At some point I need to put this theory through some real life testing. I honestly can’t imagine anything more fun than going to a track day with 4 sets of tires, removable aero, and some ballast. It would be a long, hard day of work, but what a day. It costs $2200 to rent Thunderhill West for a 2-car test day. In the off-season, they sometimes cut that in half. I just need to find another car to share the session with and a crew to help out with the pit work.

Power, grip, aero…

Everyone wants to improve their car a little (or a lot). How much do power, grip, and aero improve lap times? There’s lots of anecdotal evidence out there, but not much rigorous study. One reason for that is that every day is different. Lap times in the morning can be more than 1 second faster than the afternoon in California. One way to test these parameters on equal footing is through simulation. So I decided to embark on such an activity with Assetto Corsa. Why AC? Because it’s easy to modify. I can simply edit the power.lut file to change horsepower. Aero is similarly easy. Tires are more difficult, but there are already 2 tire choices for most cars, so I did that.

The baseline car I started with is 2400 lbs, 120 hp, 0.40 CoD, 20 sq-ft. frontal area. This doesn’t represent any specific car. It’s not far off from a Spec Miata though. The track I used was Brands Hatch Indy. I almost always test stuff on this track because it has a very small layout, about half the size of most tracks, which lets me get consistent numbers in a short time. I also think the mixture of turns and straights represents the average race track pretty well.


Let’s imagine changing horsepower from 1 to 300. I didn’t do that exactly. I drove the car with 60, 80, 100, 120, 140, and 160 hp. Then I did a curve fit to smooth the data points and extrapolate from 0 to 300. The extremes may be inaccurate, but they are also somewhat unrealistic given the starting vehicle. In the graph below, the X-axis is horsepower and the Y-axis is lap time. You can see the diminishing returns with increased power. If your car has 50 hp, adding 10 more makes a big difference. However, above 200 hp, there’s very little to be gained. If your car is a momentum car, increasing the power will lower your lap times. If you’ve got a muscle car, you need to search elsewhere.


For this next study, I once again drove the car with 60-160 hp but with harder tires. The take-home lesson here is that the higher the power of the car, the more sticky tires are important. Let’s illustrate that with a couple data points. If you’re on the red tires and have 173 hp, your lap time is 57.000 seconds. Switching to the blue tires will lower you to 56.336 seconds. You could also increase horsepower to 217 to get the same lap time. It’s a lot easier switching to stickier rubber than finding another 44 hp. Let’s look at a low horsepower example. At 100 hp on red tires, the lap time is 60.454. On blue tires, it’s 59.940. You can also get to 59.940 on red tires with 106 hp. So, sticky tires could be worth 6-44 hp depending on your starting hp of 100-173.


Aerodynamics are modeled with frontal area and coefficient of drag. It’s kind of hard determining what these values are for your car. You can look these figures up online, but I’ve found really conflicting figures. You can also try to estimate these at, which is a pretty awesome website. I urge you to check it out. Another great website is if you want to do some curve fitting. Below is the lap time as a function of CoD. This was actually done at 100 hp not 120 hp.

Fantasy Enduro Builds

Let’s imagine 2 builders, Mario and Ian, who decide to build endurance racers from Spec Miatas. Mario decides to perfect the aero while Ian decides to reduce weight.

Mario’s aero mods add weight, but since he’s no longer playing by SM rules, he is able to remove an equivalent weight. So his car stays at 120 hp and 2400 lbs, but his coefficient of drag is now 0.30 CoD. I kept the 20 sq-ft frontal area. The aero mods are worth ~6 hp.

Ian guts and chops his Miata, turning it into a freakish dune-buggy like thing. Same power and frontal area, but weight is just 2150 and CoD has increased to 0.45. The decrease in weight effectively increases hp by 14. But the CoD effectively decreases hp by 8. Overall gain ~6. (Note: I don’t believe simply summing up the hp losses and gains this way is very accurate, but for small values around the defaults, it’s probably okay).

So what’s the difference between these two builds on track? To test that, I loaded up Laguna Seca and input the new vehicle parameters. The two cars were nearly identical in lap times. The aero car had a very slight advantage at the end of the main straight. The lightweight car had a very small advantage in the infield. But after T6, the cars were neck-n-neck the whole way back to the finish line. It was a little surprising to me that the advantage of shedding 250 lbs, which is over 10% of the weight, could be completely mitigated by bad aero. I guess I had better add an air dam to my dune buggy.

FWD Drifting: Part 2 and Cones in Practice

Last week I talked about some of the tuning and techniques for drifting FWD cars. Some readers may be asking “why bother?” Well, because it’s a driving skill. And if you can drift a FWD car, it will help you drift a RWD car. Inducing oversteer by dynamically changing the balance of the car is important regardless of which wheels are providing power. I shot the video below on the skid pad at Thunderhill between coaching sessions.

The first part shows an exterior view of some switchbacks. It’s sort of comical how slow I’m going and how little my car looks like a racecar. But even at slow speeds it will slide around corners. In the second part, the camera is inside the car. You can see that I don’t use the hand brake. The car oversteers by changing the balance of the car, not locking the rear wheels. It’s also set up with a lot less grip in the ear. The car has RE-71R tires on front at 26 PSI (cold) and Hankook runflats on the rear at 38 PSI (cold).

The next series of shots are what I’m calling point to point. It’s just going around two cones but with different turn radii. I start with a large radius and progressively shorten it. Which one do you think takes the least time? Back in December, I posted on this topic. See Cones in Theory. If you don’t want to read that whole post, here’s the short version: I make the statement that path A takes less time than B, C, or D. That’s the experiment I’m performing in the point to point videos above.

I timed the various runs and indeed, the tiny radius is the fastest (path A). It’s also in a very bad spot in the power band. I’m driving in 2nd gear the whole time, and there just isn’t as much power when driving the tighter radii. But it didn’t change the outcome. Path A is the fastest way around a brace of cones.