MoTeC, a Journey: Getting The Most From Data Analysis

Thank you for these excellent write-ups, maniac! Keep them coming! I want to learn more!!!

 

Grande! Grazie!!!

 

Is there an issue with Bathurst in motec? I don't get laps recorded as there are no beacons set for the track it seems? Generating the track doesn't work either and gives me weird track layout? Any ideas what's going on? Other tracks are fine.

 

Hi Manic,

Merry Christmas for everyone. I have a question regarding with Damper.xml you shared. On your xml there are a lot of channels defined so could you give and advice how distribute them on the worksheet in channels and which type of graph is the best for those channels?

Thanks a lot.

 

You can also see here I have organized my channel report to look at corner sections only. This lets me look at each section individually while also removing straights sections. This will come in handy once I export all this to a spreadsheet to make some fancy graphs to better visualize all this data.

Here is corner exit speed channel:

Code:

choose(abs(‘G_LAT’) < 0.75
AND 'Throttle' [%] > 99
AND ‘Curvature Filtered’ < 0.50
AND 'Calc Section Time' [s] > 1
,’Corr Speed’ [km/h],
 invalid())

Hi Manic,

which is the formula for Calc Section Time.

Thnks.

 

Hi! Unfortunately I am away from my desktop for the next couple of days so I can't show what by damper report worksheet looks like at the moment.

Thanks for catching that, calc section time is a custom channel I made. I can't remember off the top of my head what the math is, but I'll just offer this alternative until I can get to my rig:

Integrate(derivative('calc outing time'), 1, range_change("Outings:Laps:Track Sections: default"))

This should give you a channel that follows the time and resets every section.

You can also try this one that might eat less CPU cycles:

Code:

'Calc Outing Time' [s]-stat_start('Calc Outing Time' [s],1,range_change("Outings:Laps:Track Sections:Default"))

 

Don't worry Manic. I can wait a couple of days .

Another petitions.

Regarding track bumpiness, could you share the intervals for each "color"? I'm not able to get the same aspect.

About Curvature which kind of graph do you use to insert the channel?

Thanks un advance

 

Hi Manic, could you check It?

Thanks a lot and Happy new year

Enviado desde mi M2102J20SG mediante Tapatalk

 

Laps Report Spreadsheet V0.2 is out! Find it here: ACC MoTeC Laps Template V0.2

- Automatic formatting: Simply export to clipboard from channel report page into 'data dump' tab. Channel report properties page still needs to be in this order.
- New simple graph, stats graph, and residuals tab.
- Various fixes

Grip math channels can be found here: Manic Grip Channels

Getting A Grip On Grip: Measuring A Car's Performance

"People…operate with beliefs and biases. To the extent you can eliminate both and replace them with data, you gain a clear advantage." - Michael Lewis, Moneyball

At any given moment of the day, some budding young e-racer is asking a discord for setup advice. Should I run more rear wing? What camber values should I have at Zolder? What are the ideal damper values? What usually follows is a series of anecdotal experiences from random individuals of unknown skill and knowledge, whose suggestions operate heavily on feel and maybe even a lap time or two. "It could just be placebo," one driver might chime in, suggesting that perhaps there is a deeper level to all this, that perhaps we need a way to separate the wheat from the chaff, that if we could quantify and put some hard numbers to these changes we could get an objective measure of whether our change was effective? Could we even potentially predict just how effective that change potentially was? Wouldn't that be a useful thing to know to help decide where we should focus our efforts for an upcoming race?

Indeed we can! And we can do so by finding the car's grip levels, and at a more advanced level, some powerful yet simple statistical techniques to track just how much of a change we can expect, including variability. But first, let's look into what it is that we are measuring…

Chopping Up The Traction Circle



Are you familiar with the traction circle? Many of you will have seen this, and you will know that maximizing our driving performance is all about living at the edge of this circle during any kind of driving maneuver. For the engineer, it is all about finding ways to make that circle bigger (or to make the proper compromises) in order to maximize the car’s potential for the driver. As useful as the circle is, we have to be aware that different conditions will alter our grip levels all around the track. One instance involves slow hairpins having lower grip levels compared to high speed sweepers due to aerodynamic downforce (though these effects are not missing in GT4 cars, they may not be as drastic of a difference). All we need to do is look at how lateral Gs increase with speed and we can see that our traction circle expands with more downforce:



Along those same lines, driver performance is going to affect how much grip we are producing lap to lap. That said, we can still get an idea of how much grip the car is producing by accounting for this variance, something which I will demonstrate later on in this post.

Another limitation to our circle is that there’s noise in our data, so it can be difficult to parse if we are using all the grip available lap to lap. The way we can overcome this limitation is by gating this data for various maneuvers and averaging them. This gets us a number that we can use to figure out how our car and our driver is performing throughout a stint, and we will use some simple graphs to learn more about our car.

I have already written a post about gating, and we will be using that technique to divide the traction circle into multiple sections. Jorge Segers (Analysis Techniques for Racecar Data Acquisition) identifies five sections for grip analysis:

1.) Overall Grip - Combines both lateral and longitudinal Gs and removes any part of the track that is not grip limited such as on the straights. A useful way to monitor oversteer and understeer events.

2.) Braking Grip - Longitudinal Gs less than -1.0g and lateral Gs less than 0.5g

3.) Cornering Grip - Lateral Gs greater than 0.5g

4.) Aero Cornering Grip - Lateral Gs greater than 0.5g and speed > 120km/h

5.) Traction Grip - Lateral Gs greater than 0.5g and longitudinal Gs greater than 0

You will notice that we are setting up the gates for each of these channels according to where each of these conditions exist on the traction circle. Mind you, these values are car dependent and subject to change according to how each car performs. I have found these values to work well with the GT3 cars. Also feel free to create additional grip channels - you could create a low speed cornering grip channel, a trail braking grip channel, grip while traction control/ABS is on, grip while coasting, and so on. The possibilities are endless! (Link for math channels are at the top of this post)

Once we have all our channels set up, we can put them all into a channel report to see how we performed over the course of a stint:



This is how our gated channels should look like. Some channels will not have as many points to analyze as others, but that simply means we’ll just need a few more data points (i.e. laps) to reduce some of the variation.

We can then put them all in a channel report and find out max and average grip we produced each lap:



Now, you can certainly try and parse all these numbers as is, but the absolute best way is to visualize them in a graph. You can copy and paste this data to any spreadsheet you wish and see what’s going on with all these channels (I recommend using my spreadsheet linked at the top of this post). So how do we use this information? There is an endless list of tests and experiments we can do with this data, so let’s go through a few of them.

Experiment #1: What is the performance difference between two cars?

We can use these grip channels to explore comparisons such as what makes each car unique, compare track grip levels, and variations in grip with setup changes. Currently in the 1.8 meta, the focus is on the BMW M4 as the leading contender for ‘best’ car. Most drivers will intuitively feel these advantages, but we can also put numbers to those observations. Let’s look at overall grip, for instance:



Here we have a comparison of the m4 vs. the m6 1.7 version at Spa in similar weather conditions (I do not have more recent data for the m6 but I encourage others to test and see if these grip numbers stay consistent). Both drivers won their league race with these cars, but based on other data, Driver 1 is more skilled. In this graph we are using what is called a box plot, which is just a way to represent the spread of our measured values. At a glance, we can see the minimum and maximum values, the median, represented by the line within the box (outliers not pictured here). Values on the higher end are obviously going to be harder to attain, even for the best drivers, whereas values within the box and around the median are what we are more likely to expect. The dots represent the actual data points. We can see that there seems to be an advantage for the m4 of about 0.01G - .015G. That might not seem like much, but averaged over the course of a whole race this can have a significant effect. As a reference, BoP ballasts can range from 10-40kg, which is close to adding .01G by typical car weights. Let’s look at brake performance:



As many have observed, the braking performance of the M4 blows the M6 out of the water. We are looking at almost a full 0.1G improvement, which remember, is equivalent to a tenth of the car’s weight. The car does gain significant drag on the straights (will show later), but an advantage of this magnitude can be quite difficult for other cars to overcome.



Overall cornering grip averages. This data is quite noisy partly due to aero effects, which is why we have created an aero cornering grip channel as well. While it may seem that the M4 has better cornering ability, the overall spread is the same as the M6. Cornering is also quite skill dependent. More data points for the M4 could help us get a better idea, or we can look at aero grip alone, like so:



Now we get a better picture of the car’s cornering performance, especially at Spa where aero plays a big role. You will notice an outlier blue dot at the top here. Examining the lap data, Driver 2 going into Eau Rouge put a lot of steering into the car, producing a high amount of cornering grip, but ultimately took the section quite slow because of it. Thankfully, our box plot correctly recognized this event as an outlier. The M4 outperforms the M6 in high speed corners, producing again about 0.015Gs more grip on average.



Quite a spread here, there's actually quite a lot of noise with this grip channel. It is important to note again that skill does play into some of these grip channels, traction included. Still, we can see that the M4 certainly is capable of performing better with traction on average than the M6 can. How about trail braking phase?



The M4 again produces a significant amount of grip versus the M6 at its best, almost 0.04 - 0.05G’s better. The M6 is just fantastic in the braking zone.

So how does this all translate into lap times? Let’s look:



Some of you might be surprised to see that the driver of the M4 is nearly 2.5 seconds slower than the driver of the M6, on average. The big reason is that we are looking specifically at grip limited sections of the track and not taking into account performance on the straights, as well as driver performance for that matter to a large extent. Take a look at this graph:



Driver 1’s box looks a bit strange, but we are looking at a full race weekend and weight differences can have big effects on top speed. In any case, it should be quite clear now where the discrepancy lies. All that additional grip means significant straight line drag, and so the M4 will lose out on the straights versus the M6 every time. It is not conclusive whether the M4 would consistently lose out to the M6 at Spa because of the importance of driver skill during cornering, but hopefully these graphs demonstrate that we can to some degree separate the performance of the car from the driver, even if we are not driving perfectly on the limit or significantly behind on pace. Thankfully, this also demonstrates that our data channels are observing what they are supposed to - the parts of a track where we are grip limited, rather than power or skill limited. For those, we can use other performance metrics like top speed or throttle percentages to figure that out.

(I used Jamovi to produce box plots, Google sheets unfortunately does not have the option)

Experiment #2: How much influence does grip have on lap times?

Or rather, how will a change in grip affect my overall lap time during an entire session? As the last experiment showed, a very important point to remember is that higher grip is not always associated with faster lap times when looking between cars. In order to find that out, we need to control what variables we can to develop an estimate, such as comparing two different setups by using the same driver, car, track, and weather conditions. We can then examine the results of our experiment by using a very powerful statistical technique, linear regression analysis. Once you understand the basic concept, you will have a useful tool at your disposal to find out just how much any variable you can pull from the data influences lap time, grip levels, or any other factor that you want. We will use this technique in all of our experiments so let’s see what it looks like:



*Please note: This technique is what I consider ‘advanced’ for those just getting into data analysis and can lead to any number of false conclusions if you are not careful about how to interpret this data. You have been warned!

Let’s go over the graph and define all the different parts:

The Regression line/Prediction line/Trend line - The blue line running through our data, created using a method called the sum of least squares. It is a mathematical representation of the direction and position of our scatter plot, and helps us to quantify any kind of linear relationship we might be observing in our data set. Forms the basis for the assumptions and predictions we will be making.
Slope - the slope of the regression line from the equation y = mx + b, representing the amount of change in our dependent variable (the y-axis) for every one unit of our independent variable (the x-axis). It tells us that in this data set for every one unit of overall grip, we reduce our lap time by 30.576 seconds. Because we tend to only affect grip by tenths or hundredths, it is better to scale this value like so: -0.306 seconds per 0.01G. This gets us an idea of what is a meaningful measure for our independent variable.
Intercept - unimportant for us, only used to draw the regression line.
R-squared - a percentage value that describes the variation explained by our independent variable (overall grip). Higher the percentage, better the correlation. This does NOT imply that grip is directly responsible for 60% of our lap times, or that a low percentage is somehow worse than a higher one, only that it is correlated in some way. A lot of variables go into lap times that are also correlated with each other. Low correlations can still be meaningful depending on our slope, the standard error, and our data set.
Standard Error Of The Regression/Estimate - measures the average spread above and below our regression line (similar to standard deviation) and is represented by the upper and lower yellow lines. In a lot of ways, this is more meaningful than R squared as it shows us how close the data points fit to the regression line on average in the units we are measuring. Lower numbers, the more precise and less variance in our prediction. Higher the number, greater the variance, the less precise our prediction. In this instance, we experienced a variation of ±0.572 seconds on average for any given value of overall grip.
95% Prediction Interval: Standard Error multiplied by 2. Approximates where we can expect 95% of our data to end up along the regression line. Represented by the red lines above and below the regression. Considered more of a ‘safe bet’ estimation since it represents practically all of the data points.
Standard Deviation Of The Y Variable: As it says. A measure of the variance in our spread. I use this to roughly estimate how useful my standard error is.

Evaluating The Regression

Now that we have some of the definitions and basics covered, we need to evaluate our regression, lest we fall victim to any number of biases that can impact our experiments. Here are a few quotes I made up that we should keep in mind:

1. “All the residuals matter.” - Before we draw any kind of conclusion we should have a quick look at the residuals tab and check if our fitted values look random like so:



This helps us check if our data is biased or not. Biased data is very easy to tell, because they will very often stand out like this:



The points are not random and we are clearly looking at two different data sets. We have to check for bias because the R squared for this graph is 30.78%, which might suggest a relationship, but in reality, we are looking at biased data. This does not mean we cannot glean useful information from this relationship, just that we cannot run a regression on it, unless we remove the specific data points that are causing the bias. This is very easy to do and takes all of one second. Check your residuals. Do it.

2. “Not all regressions occur on a straight line.” - Should not come as any surprise, but not all things in nature occur in a linear fashion. The limitations of google sheets’ graphing capabilities and also the complexity with quadratic regressions would make that kind of analysis more burdensome than it needs to be. Luckily, a lot of the relationships you find here actually appear as a linear relationship, and the residuals plot will tell you if it is not. When assessing driver skill and lap times, very often we are looking at a specific time frame in their skill development, which appears as a straight line up close, but over months and years can appear curved, much like looking for the Earth’s curvature from the surface vs. from up in space. But it is important to keep this in mind so that we do not try to make any grand assumptions with our prediction model.

3. “When comparing setups, best practice is to do an A-B-A test.” Once you start running these regression models on your setups, you really do want to control for psychological impacts on lap times, as well as any other variation that may occur. Practice, hunger level, Hawthorne effect, placebo, nocebo - the list goes on and on when it comes to the driver’s mental state and its effect on driving ability. So having at least three data points is critical to making the regression run accurately. Run your baseline setup (A), make the change (B), then run the car again with the baseline (A). 5 laps per setup is a good minimum. Once you get down to the really fine tuning bits like dampers, it might even be the case that A-B-A-B-A will be necessary because of how fine an impact they can potentially have.

4. “Regression models can get really complex.” As I had mentioned before, regression models are complex tools and what I present here is very much the simplest form of regression analysis, but it is more than enough to make some useful observations about the building blocks of our lap times.

5. “Use what you know to make smart decisions.” - Racecraft and vehicle dynamics are very important in understanding how to use regressions to their fullest. Fuel capacity, tire wear, camber values, weather conditions…these things are not automatically collected by motec but they will show up in the data in other ways. Always keep this in mind when testing these regressions - it is a big part of the reason that I set up the graphs the way that I have. If something does not look right or does not conform to our expectations about vehicle dynamics, more than likely something is not right in our model or there are other variables we are not considering.

What’s the point of all this stuff, anyway? Will it make me a better driver?

For the engineers out there, it’s very simple - you need this information. An engineer cannot work without data, even if that data is just lap times. I believe it is especially important for sim race engineers, because the margins can be so slim compared to real life racing, that extracting every last ounce of time from the car becomes critical. Consider the regression model I presented above - we are talking about finding 3 tenths in lap times with a variance of ±1.144 seconds! And that’s not including outliers, mind you. And these effects get smaller depending on what you are adjusting. How you choose to test and evaluate the data is going to drive your conclusions, and it is why so often you hear that “dampers don’t do anything,” even from pro drivers. The truth is with these cars and the level of driving possible, it simply is not possible to notice a change without using a more robust methodology to suss out the benefits. A race very likely will not be decided on dampers alone, but multiple minor changes can. Regression analysis can help us find what those minor changes should be.

For the drivers, well, I think it depends on what kind of problems you need to solve. I like using music analogies, and data is like using a metronome to practice - as valuable as it is, it does not make you a better musician, it just keeps you honest. Data does not lie, but we cannot expect it to give us the exact answer. If your problem is “I want to improve my driving overall,” it might be better to just get a coach. If it’s something more specific like “I want to check how consistent my brake point is at turn 1,” or "I want to know what impact a 0.5 degree change in camber has on braking and cornering grip," then data becomes much more useful. Regressions can also give you a clearer idea of how your performance changes over time, which can be a confidence booster (just plot lap numbers vs. lap times). Typically in other sports the coach takes this data to make better informed decisions about setting up an athlete’s workload or practice routines, and there’s no reason you cannot apply that here.

Experiment #3: Do damper changes affect lap time?

I will save this for another post as it requires a lot of testing and data, but everything that I have written here will give you everything you need to start looking for an answer. If you do end up doing any testing, please send me your log files my way! Would love to see what you engineers can come up with. Good luck!

 

I use 25mm/s as my minimum, 200mm/s as my maximum/greater than value.

For curvature I just use an ordinary time/distance graph. If the scale is off, there is a quantity in the edit math channel page that is for curvature, and you can put in 1/m as your unit.

 

Hi Manic,

is it posible to show your damper worksheet to see how have you organized it?

Thanks for your work and share your knowledge.

 

Honestly I just throw it all in a channel report and export it to a spreadsheet. I hate looking at these numbers without a graphical representation.

Is there something specific you wanted to do with these channels?

 

I'm thinking on how you organize yourselve on Motec dampers worksheet the math channels you shared. Which channels are you grouping to compare what...?

I've read the dampers post a lot of times, I've downloaded the XML file with math channels but I don't know how I have to use all of them

Thx

Enviado desde mi M2102J20SG mediante Tapatalk

 

Easiest way is to group them according to what each represents. So I have my damper channel report organized as such:

Standard Dev. of FL,FR,RL,RR damper vel - One metric to describe how fast my dampers are moving
Zero bin % of FL, FR, RL, RR dampers - Another similar metric, can also be used to track which damper setting 'feels' the best and thus gives a benchmark to apply to other tracks with the same car.
Bump/Rebound avg speeds FL, FR, RL, RR: Tells me how fast my dampers are moving in each direction. Lower speeds mean my damper adjustments will have a greater effect than at higher speeds
Bump lo speed, Rebound lo speed, Bump hi speed, Rebound hi speed % for FL, FR, RL, RR dampers: A simplified version of the histogram. Gives me detailed info about each damper and what's happening in the lo/hi range, for both bump and rebound. Not strictly necessary, but can get you some really interesting information about how the damper behaves throughout each part of the track. Averaging these values will get you a better idea of where your 'true' percentages are. Can also be used for more advanced graphing purposes.

 

Oversteer/Understeer: Getting Balanced

"When you see the tree you're going to hit, that's called understeer. If you can only hear and feel it, it's oversteer." - Walter Röhrl

Link to math channels: Download

As anyone who has spent any time in a rally car, one becomes deftly acquainted with the car’s ability to spin in every direction other than the one you intended. Trees are a consistent threat in rally - circuit racing, however, does not see as many of these natural barriers. Instead, the nature we do fear is the car’s ability to hold the racing line. Indeed, the very best drivers among us make their millions managing the car’s balance while on a razor’s edge, daring the car to break traction and glide across the tarmac like rocks on ice. It should not be any surprise, then, that managing the car’s balance is the key to becoming an elite driver, and the engineer needs to understand how this phenomenon behaves in order to give the driver the confidence they need to push the car in all the right places.

Car behavior for a given steer angle. Note that speed, pedal inputs, load balance, etc., all contribute to the car balance.​

Even if you think you know their meaning, let’s start with a few definitions so we can establish some ground rules. Oversteer should be immediately obvious to any driver - the car turns in faster than you expect. It does not take but one or two code brown moments to know that steering too quickly is going to cause a race incident. Poor drivers will oversteer a GT3 car with ease, and so do poorly set up cars. Oversteer on a simple technical basis is the rear tires losing grip relative to the front tires.

Understeer is a little bit trickier to understand and feel because you need a little understeer to even get the car to turn, but the concept is the same - the more steering you put into the car, the less the car responds to your input. Less skilled drivers will not perceive this as a bad thing, simply because the car will eventually catch up to where the driver starts to unwind while taking a less than optimal line. The problem here is not recognizing the damage this causes to the front tires - understeer will wear out the tires faster than necessary, put more heat into them, and potentially cause damage. Understeer may be preferred to oversteer but you really do not want either. In the United States, understeer might instead be called “push”, which arguably is a better descriptor of what’s happening since it feels like you are literally being pushed away from your desired racing line.

What you really want is something you do not hear used as often, which is neutral steer. Neutral steer says that the car responds in the way that the driver expects. Notice that I do not append this definition with any kind of technical requirements - this is very much a condition that is pertinent to a driver’s sense of the car! Some drivers want the car to feel edgy and respond quickly to steering inputs, while others prefer something less sensitive. Adjusting steering ratio is one of many ways to find a comfortable steering window. And if it is not immediately obvious to anyone who has spent more than a few hours in the game, a perfectly neutral car in every condition simply does not exist and is as much an indicator of driver performance as it is about the car’s balance (more on that later).

It is important to note that I want to establish these definitions early on, because once we start getting into the discussion of the car’s actual rotation (yaw rate) it will become very easy to start confusing these concepts with other phenomena. So quickly, to recap:

• Oversteer = car turns more than the driver expects (driving line bends in)
• Understeer = car turns less than the driver expects (driving line bends out)
• Neutral steer = car behaves as driver expects it to (turning behavior predictable)

So with something as important as car balance, is there any way to measure this balance using the logged data? You betcha! And you probably didn’t know you already had it in your workbook!

Observing Balance With The Understeer Angle

You may have noticed in the default ACC workbook a little channel called “Oversteer” sitting in the US-OS tab. What is that data trace all about?

The Oversteer channel. Observe the yellow trace, which represents the difference between the steering angle and the lateral movement of the car.​

This channel is a filtered version of what Milliken calls the Understeer angle. By comparing the car’s lateral trajectory to the angle the front wheels are pointing, we can create a data trace that gives us a relative indication of how much slip angle we are producing. Kunos/Aris’ oversteer channel uses this math to produce a channel that shows you when the car is oversteering (positive) versus understeering (negative). This helps us get around the problem of not knowing the actual slip angle of all four tires. (Listen to this interview at 30:00 if you are curious how it works and compares to actual slip angles)

We will, however, use a different math channel to measure this effect. The oversteer channel uses a unique filtering equation but introduces errors and excessive filtering. Here is what the understeer angle channel looks like:

​

Unlike the oversteer channel I mentioned above, the understeer angle goes in a positive direction when there is understeer, and negative when there is oversteer. This is a better representation of what is happening at the wheels because inducing a slip angle will have the front tires turned at a greater angle than the car’s angular trajectory. Understeer will be characterized by a gradual increase in angle, indicating the car is not responding to the driver’s steering input as much as desired. This is a very rough estimate, but inducing an angle greater than 4 or 5 degrees suggests you are heading towards excessive understeer. Your wheel should give you feedback, and you can audibly hear the wheels scrape if the angle is too high. But what exactly are we looking at? How are we able to discern this?

Negotiating the last right hand corner at Paul Ricard
​

The understeer angle is the difference between the steered angle (determined by the steering ratio in the car setup) and an Ackermann angle (not to be confused with Ackermann steering), which is the angular representation of the car’s lateral acceleration versus its cornering radius. In the screenshot above, we get a sense of how the driver is reacting to the lateral development of the car - the driver steers in quickly, but the Ackermann angle does not increase at the same rate (understeer). So the driver parks their steering angle and waits for the car to catch up. At the point where the Ackermann angle starts to decrease is exactly the point where the driver hits the throttle (i.e. the car wants to straighten out as you convert lateral acceleration back into longitudinal). Rather than reducing their steering angle, the driver instead turns even more into the corner, thus producing even more understeer, which only increases as the driver tries to manage the speed at which the Ackermann angle decreases. As you can guess, this large understeer moment requires a moment of countersteering to correct. Given the lack of smoothness in the driver’s steering and relative smoothness of the Ackermann angle, we might be able to conclude that the driver is using these more aggressive inputs to manage the car’s trajectory. There’s a thin line between aggressiveness and being too abrupt, and the better drivers will be able to straddle that line consistently.

When we superimpose the understeer angle onto a track map, we get a sense of how the car behaves around the track:

The Understeer Angle Track Map​

It also allows us to average all the values over the course of a lap and lets us see generally how the car behaves over the course of an entire race.

​

Having a track map and a run chart are really where this math channel shines. We in effect not only get a sense of how the car behaves over the course of a lap, but we can get a sense of how the driver responds to the changing behavior of the car. Notice how this driver is steadily putting more understeer into the car as the race goes on. This in effect is telling me that the driver is not making appropriate adjustments to the car’s changing balance. Why? Because the understeer angle is a representation of how much steering we are putting into the car versus what the car is giving us. With tire wear/changing weather conditions/spent fuel, our Ackermann angle will not respond as much to our inputs, so we need to be adjusting our racing line to accommodate these changes. A pattern like this suggests to me that the driver is not adjusting some aspect of their racing line, such as not taking an earlier brake point or adjusting their throttle response. This is the downside of muscle memory, which can put you into robot mode and is ultimately slowing you down/making things harder for yourself.

We can even compare drivers and cars:

​

In this case, we see how two different drivers respond to different cars. Driver 01 has a larger variance in terms of how much understeer they are putting into the car, which increases ever so slightly as fuel load decreases. Driver 02 puts comparatively less understeer into the car and with less variation, which either suggests the car feels more neutral or has a tendency to oversteer. Understeer values also look flat across the stint. Either the driver is responding well to the car’s changing behavior, or they are missing some potential by not being more aggressive. In examining a variety of metrics and comparisons and driver feedback, it was concluded that driver 02 was indeed being passive, partly invoked by a lack of smooth braking which sacrificed cornering potential.

Final Notes

One thing that is important to realize about this math channel is that it won't tell you exactly if a car has more understeer or oversteer due to car setup. As was implied earlier, how the Ackermann angle develops is just as much a part of driver performance as it is the car's ability to develop that angle. This is why so often you hear the need to be driving at the limit before you can fine tune the car's balance. It is best if you are still learning how to drive fast to use this channel as a performance metric. Get a sense of how smooth your steering is, how you respond to changes in the car behavior, how you steer at various stages of a corner. Find a racing line through a corner and make yourself aware of how you maintain that line - are you pushing out? Getting pulled in? If you can pick a racing line, you can start adjusting your inputs to help keep that line.

You also have to be aware of how the understeer angle changes with speed. We can plot speed vs. understeer angle in a scatter plot to observe this effect:

​

We can see at low speeds we are able to induce much higher slip angles than at high speed. At high speeds we have significantly lower acceptable slip angles in both directions. This is partly why that 4-5 degree reference point I mentioned only works as a guideline - different forces at different speeds mean a changing degree of acceptable understeer angle. Much like we do with grip factors, we can use gating and section analysis to get a more granular look into how understeer behaves at different parts of the track and at different speeds if we wanted to get more advanced. If you are curious about the relationship between corner radius, speed, and downforce, I highly recommend watching this video to understand how these limits are understood. If you are trying to turn the car at a speed that simply is outside of the tires' corner radius limits, no amount of setup is going to get rid of that understeer. Same with entering a corner too fast, or jamming the throttle too fast and for too long on corner exit. Determining the balance of the car requires being able to hit the exact same line with near the exact same speed each and every time, and then and only then can you discern if the problem is either driver or car related.

In part 2, we will explore how to use yaw rate to observe understeer/oversteer characteristics and how it differs from using the steering angle. Stay tuned!


Advanced sectional analysis of understeer angle. Consistency and variance can be observed in the length of each box plot. Can suggest greater experimentation/higher risk taking/aggressive driver, or excessively abrupt, non-smooth driver. Take note of other driver metrics like lap times and section times to make that judgement.

 

Can you show us a picture of your workbook about this?. Thanks

 

Just wanted to share a quick time saver that you may not have noticed in one of the recent MoTeC updates:



In the options menu circuit tab, there's now a feature that automatically removes trash laps! Before this had to be done manually and was time consuming. Just switch that on and you are good to go. 3% is actually quite loose, those are laps usually within 3 seconds, but adjust accordingly to what's appropriate for you.

It would look something like this:

 

Hope everyone is enjoying the new DLC cars! Wanted to chime in with something that showed up in the v1.8.12 changelog: Fast bump dampers had a reverse effect in the UI!

A few thoughts on this: A.) I'm kicking myself for not having seen this as an obvious bug B.) It does not fundamentally change how to approach damper tuning or any setup change for that matter, which is assume nothing, test everything! We can observe the effects of different settings, we do not want to assume something just because a technical book said so - always be testing, confirming, and then measuring the magnitude of their effects. Ultimately I don't think this was a game breaking bug and in fact it might give bigger credence to the idea that most people actually would not benefit from altering damper settings, as they didn't much notice a difference anyway.

In other news, I'm going to be putting up my workbook soon for anyone to use, along with a video explainer. Hopefully I'll be able to get that done sometime in the next few weeks while I in addition to my examination of yaw rate. Thank you all who take the time to read my posts!

 

@Manic_Driver Thank you very much for sharing your knowledge, it not only help me and other with AC1 and ACC but for other sims too !, much appreciated hard work.

 

impatiently waiting for the workbook and video

Enviado desde mi M2102J20SG mediante Tapatalk