Skip to main content

Plot the trend before you judge the change

Generated from content/lms/data-interpretation-ii-advanced/06-metric-driven-analysis/02-run-charts.md; edit the source file, not this page.

Source path: content/lms/data-interpretation-ii-advanced/06-metric-driven-analysis/02-run-charts.md

Course: Read the data your hands can't feel

Module: Build metrics that survive the noise

Estimated duration: 45 minutes

The skill in this lesson is simple to state and easy to violate: before you decide that a setup change worked, that your driving improved, or that the car has developed a problem, plot the metric across the run and look at the trend. A single lap can tell you that something happened. A run chart helps you decide whether that something belongs to the car, the driver, the tires, the track, or the randomness of one imperfect lap.

This matters because logged data is not naturally useful. A logger can produce a large series of numbers during a practice session or race. Speed, throttle, brake pressure, steering angle, longitudinal g, lateral g, and other channels are valuable only after you reduce them into something you can reason about quickly. The source material frames this in two connected ways: you can visualize the numbers graphically, or you can summarize them with statistics. Metric-driven analysis joins those two ideas. You build a metric that answers a specific question, then you plot that metric lap by lap, run by run, or stint by stint so the important portions of the data become detectable without staring at every squiggly line.

A run chart is not just a prettier table. It preserves order. That is the whole point. If you only compare your best lap to your previous best lap, you can miss the shape of the session. If the first three laps are improving, the fourth is the best, and the next five fade away, that is a different story than eight laps that all move steadily in the right direction. If one lap jumps after a setup change but then returns to the old level, that is different from a step change that stays present for the rest of the run. The same number can mean different things depending on where it appears in the sequence.

Use run charts when you are trying to make a decision from more data than your eyes can hold at once. That includes driver performance analysis, vehicle performance analysis, setup evaluation, tire behavior, and driver consistency. The bonded material supports each of those use cases: logged cockpit activity can show driving style, logged vehicle behavior can be combined with driver comments to locate handling problems, comparative analysis can reveal the effect of setup changes or driver performance, and metric/run chart analysis can accelerate interpretation of a large dataset. The intermediate driver version of this lesson is not about building a professional engineering workflow. It is about becoming disciplined enough that you stop judging the change before you have plotted the trend.

The first sub-skill is choosing a metric that answers one question. Start with the decision you need to make. Do not start with every channel you logged. If the question is whether you are braking more consistently, the likely raw channels are speed, longitudinal g, brake pedal position, and brake line pressure. If the question is whether you are becoming more consistent in driver activity, the likely channels are speed, throttle position, steering angle, and brake pedal position. If the question is whether the car is moving toward understeer or oversteer, the source list points you toward speed, throttle position, front lateral g-force, rear lateral g-force, steering angle, and understeer angle. The metric must connect the question to measured behavior.

The second sub-skill is keeping the metric small enough to trust. A good first metric is not a giant score that claims to explain the whole lap. For this lesson, use a metric narrow enough that you can explain it in one sentence. Examples include minimum speed at a selected corner, distance from first brake application to minimum speed in a braking zone, throttle application point after corner exit, peak steering angle in a selected corner, or lap-by-lap delta from a target. Those examples come from the channels available in the corpus, not from a claim that one metric is universally best. The right metric is the one that makes the next decision clearer.

The third sub-skill is plotting in driving order. A run chart loses its value if you strip away sequence. Put lap 1 before lap 2, lap 2 before lap 3, and keep going. If you are comparing runs, keep the sessions in order. Mark any event that could change the interpretation: tire pressure adjustment, setup change, driver swap, traffic, rain, red flag, or a deliberate technique change. The source material supports comparing different laps or runs with previously collected data to reveal the effect of setup changes or driver performance, but you still have to protect the comparison from confusion. If you change the driver and the setup at the same time, the chart may show a change, but it will not cleanly tell you why.

The fourth sub-skill is reading shape before reading the headline number. You are looking for four basic shapes. A steady trend means the metric moves mostly one direction across the run. That can suggest learning, adaptation, tire change, or a vehicle condition that evolves with laps, so you do not stop at the chart. A step change means the metric moves suddenly and then stays there. That can be the signature of a setup change, a driver technique change, or a different traffic situation, depending on what else happened. A spike means one lap is different but the surrounding laps are not. That may be a mistake, a clean lap after traffic, a sensor issue, or a one-time event. A flat but scattered pattern means the average is not moving much, but the driver or car may lack consistency. These shapes are not diagnoses by themselves. They are prompts for better questions.

The fifth sub-skill is checking the chart against other channels. The driver-data process in the corpus is explicit about starting with an overview, looking for incongruencies, digging for details, using other channels if available to check, asking why, comparing where possible, calibrating to your driving, imagining an ideal trace, and setting objectives for the next session. That is a good run-chart workflow. If the run chart says your braking metric improved, check speed, longitudinal g, brake pedal position, and brake pressure. If it says the throttle metric improved, check throttle position and speed. If it says the car balance metric shifted, check steering angle, front lateral g, rear lateral g, and any understeer metric available. A metric that cannot survive a channel check is not yet evidence.

The sixth sub-skill is connecting the trend to what you felt. Logged data forms an objective measurement of vehicle performance, but the source material also says it can be used with the subjective comments of the driver. That pairing is important for intermediate drivers. Your feeling alone is not reliable enough. The chart alone can also mislead if you do not understand the driving context. A good note after a session might be: the car felt lazy on entry after lap 5; the run chart shows the minimum-speed metric dropping after lap 5; the steering channel shows more steering angle for the same corner; the next session objective is to repeat the earlier entry speed while watching brake release discipline. That is a usable analysis loop. It starts with the trend, checks the channels, ties back to feel, and ends with a driving objective.

The seventh sub-skill is resisting the best-lap trap. The best lap matters, but it is not the same thing as the trend. A best lap can be the result of cleaner traffic, a lower-risk push, or one corner executed unusually well. If you only analyze the best lap, you are doing comparative analysis on a narrow slice. The source material supports comparing data from different laps and runs, but the whole reason for metrics and run charts is to accelerate interpretation of large datasets. Let the best lap raise a question, then let the run chart show whether the answer repeats.

Here is the practical method you should use after a session.

  1. Write the question in plain language. Use a question that can be answered by logged channels. For example: did my braking into the selected corner become more consistent across the session, or did the car become harder to rotate after the tire pressure change? If the question cannot be connected to measured channels, it is not ready for a run chart.

  2. Choose one metric. For braking, use speed, longitudinal g, brake pedal position, or brake line pressure. For driver activity, use speed, throttle position, steering angle, and brake pedal position. For vehicle balance, use speed, throttle position, lateral g, steering angle, and understeer-related measurements if available. Do not add channels just because they are interesting.

  3. Plot the metric in sequence. Each lap, run, or stint gets a point. Keep the order visible. If you can annotate the plot with notes, mark the lap where you changed technique, the lap where you caught traffic, the run where the setup changed, or the run where a different driver took the car.

  4. Name the shape before naming the cause. Say steady improvement, steady fade, step change, spike, scattered flatline, or no visible change. This keeps you from jumping straight to a conclusion that the chart has not earned.

  5. Check at least one supporting channel. If your metric is built from brake pressure, check speed or longitudinal g. If it is built from steering angle, check speed and lateral g. If it is built from throttle point, check speed at the same section. The support channel does not have to prove the whole story. It has to tell you whether the first interpretation is plausible.

  6. Compare if you can. Compare your current run with a previous run, another lap, or another driver in the same car. The corpus specifically points to comparing different laps or runs with previously collected data to reveal setup effects or driver performance, and to comparing multiple drivers using the same car. This is where the chart becomes more than a personal diary.

  7. Set one objective for the next session. The source process ends with objectives. Do the same. A run chart without a next-session objective is analysis for entertainment. A good objective is narrow and testable: hold braking start point steady for five clean laps, reduce steering correction after initial turn-in, or repeat the throttle application point without sacrificing exit speed.

Calibration cues matter because the chart can improve before the stopwatch does. If your lap times are similar but your braking metric is less scattered, you may be building repeatability. If your minimum speed through the selected corner rises but your exit speed does not, you may be carrying speed in the wrong phase. If your steering-angle metric climbs while speed stays similar, the car or your line may be demanding more steering input for the same result. Those are not final diagnoses. They are better questions. The improvement is that you are no longer guessing from memory.

A good run-chart review has a particular feel. You are not hunting for a flattering number. You are reducing the data until it answers the question you wrote down. You are using graphics as a reasoning tool, not as decoration. You are letting the order of the session stay visible. You are checking the first interpretation with another channel. You are connecting the chart to your driving notes. You are leaving the review with one objective that can be tested in the next session.

The failure mode is also easy to recognize. You open the data, look at the fastest lap, compare it to another lap, and declare that the change worked. Or you see one point on a chart and immediately assign a cause. Or you create a metric so complicated that you cannot explain what a movement in the chart actually means. Or you skip the channel check because the chart agrees with what you wanted to believe. Each of those errors defeats the purpose of metric-driven analysis. The logger gave you more information than you can interpret raw. The run chart is how you make the information manageable without throwing away the session story.

This lesson sits beside, but does not replace, segment reports. Segment reports help isolate where variance appears around the lap. A run chart helps you understand how a chosen metric behaves across laps, runs, or sessions. Use a segment report when the question is where the time is coming from. Use a run chart when the question is whether the behavior changed, whether the change persisted, and whether the pattern suggests a next-session objective. The strongest workflow uses both, but this lesson is about the run chart discipline: plot the trend before you judge the change.

Worked example: braking consistency across one practice session​

You come in after a session convinced that your braking improved. The feeling was real enough: by the end of the run you were more comfortable pressing the brake pedal harder, and one of the later laps was your best. The mistake would be to compare only that best lap with an early lap and call the improvement proven.

Use a run chart instead. Write the question first: did braking into the selected heavy braking zone become more consistent across the session? Choose one metric from the channels supported by the corpus. A simple option is the distance or time from first meaningful brake application to minimum speed in that zone, using brake pedal position or brake pressure to identify the start and speed to identify the minimum. Then plot that metric lap by lap.

Now read the shape. If the chart steadily tightens and the values settle into a narrow band, the main story may be repeatability. You are doing the same braking job more often. Check longitudinal g and brake pressure to make sure the pattern is not just a sensor oddity or a different definition of brake start. If the chart shows one excellent lap surrounded by ordinary laps, do not build your next-session plan around that one point. Treat it as a clue. Ask what was different about that lap, then check speed, brake pressure, and longitudinal g.

If the chart fades later in the run, the conclusion is not automatically that your braking got worse. The source material points to investigating tire wear and driver consistency, and it also emphasizes using other channels to check. A fade in the braking metric might come from tires, from the driver protecting the car, from traffic, from missed references, or from a pressure change that altered the car. The run chart does not replace judgment. It slows your judgment down long enough to make it useful.

A good next-session objective from this example is specific: repeat the same braking start and pressure build for five clean laps, then review whether the run chart shows a narrower spread. The success criterion is not simply a faster best lap. The success criterion is a trend that supports the skill you intended to practice.

Worked example: comparing two drivers in the same car​

The corpus identifies a strong use case for driver performance analysis: comparing differences in style and performance among multiple drivers, especially when more than one driver uses the same car. That is a perfect run-chart situation because the car is shared and the driver becomes the main variable you are trying to understand.

Imagine two drivers run the same car in separate sessions. You want to know whether one driver is carrying more entry speed or whether the difference is actually in throttle timing. Do not begin with a debate in the paddock. Build a metric. For entry behavior, you might chart minimum speed at the same corner for each lap. For throttle timing, you might chart the point where throttle position begins a committed increase after the corner. The available channels in the corpus support this kind of driver activity review: speed, throttle position, steering angle, and brake pedal position.

Plot each driver in session order. The first useful question is not who has the single best point. The first useful question is whether the patterns are different. One driver may have a higher peak lap but a scattered run chart. Another may be slightly slower but much more repeatable. If the goal is coaching, that difference matters. A driver who is scattered may need a repeatable reference and one next-session objective. A driver who is repeatable but consistently low in a metric may need a technique change.

Then check the supporting channels. If the minimum-speed chart says Driver A is faster into the corner, look at brake pedal position and steering angle. If Driver A carries more entry speed but also uses more steering angle and loses exit speed, the entry metric alone may be flattering the wrong behavior. If Driver B carries less entry speed but reaches throttle earlier and gains on exit, the run chart has revealed a style difference rather than a simple good-versus-bad result.

The teaching point is that run charts let you compare without flattening the drivers into one lap time. You can see whether a difference is persistent, whether it appears only on one lap, and whether it matches the channel evidence. That is more useful than telling one driver to copy the other without knowing which part of the lap actually matters.

Common mistakes​

The first common mistake is judging from the best lap. A best lap is useful, but it is only one point. Good practice is to plot the metric across the run and ask whether the best-lap behavior repeats.

The second mistake is building a metric that is too broad. If your metric tries to score the whole lap, the chart may move without telling you what to do next. Good practice is to choose a metric tied to one question and one phase of driving: braking, throttle application, steering activity, vehicle balance, or consistency.

The third mistake is ignoring the order of the data. A table of lap values can tell you high and low. A run chart tells you whether the change appeared early, late, suddenly, gradually, or only once. Good practice is to preserve lap or run sequence and mark known events that affect interpretation.

The fourth mistake is treating the chart as the whole answer. The corpus process says to use other channels if available to check and to ask why. Good practice is to verify a braking metric with speed, longitudinal g, brake pedal position, or brake pressure; verify a throttle metric with speed and throttle position; and verify a balance metric with steering angle and lateral g channels when available.

The fifth mistake is confusing driver change with car change. Comparative analysis can reveal setup effects or driver performance, but only if you know what changed. If you changed the setup, changed tire pressures, and changed your braking approach in the same run, the chart may show a real change while hiding the cause. Good practice is to mark changes and keep the next test as clean as possible.

The sixth mistake is refusing to act because the chart is imperfect. Club and HPDE data is rarely laboratory clean. The answer is not to pretend the data is perfect; it is to keep learning, keep the question simple, compare where you can, calibrate to your driving, and set one objective for the next session.

Drill: three-run metric trend discipline​

Use this drill at your next event when you have any logger or app that can show lap-by-lap data. The count is three runs, with one metric per run. The duration is one event day or three consecutive sessions if your schedule allows. The success criterion is that you leave each run with one plotted trend, one supporting-channel check, and one next-session objective.

Run 1 is the baseline. Before you go out, choose one question that matters to your driving. For an intermediate driver, braking consistency is usually a good choice because the source channel set is clear: speed, longitudinal g, brake pedal position, and brake line pressure if available. After the run, plot one braking metric lap by lap. Do not judge it yet. Name the shape first. Is it steady, scattered, fading, improving, or dominated by one spike? Then check one supporting channel.

Run 2 is the controlled attempt. Set one objective based on Run 1. If the chart was scattered, your objective is repeatability. If the chart showed a late fade, your objective is to preserve the earlier pattern later in the run. If the chart showed one good spike, your objective is to repeat the conditions that produced it without chasing lap time blindly. After Run 2, plot the same metric again. Compare the shape, not just the best number.

Run 3 is the confirmation run. Keep the same metric unless the previous chart proved it was the wrong metric for the question. You are looking for persistence. A real improvement should be visible as a pattern you can explain and support with at least one other channel. If the trend is not there, that is not failure. It means the change is not yet repeatable or the metric does not match the skill. Your final output from the drill is a short note: metric used, trend shape, supporting channel checked, likely interpretation, and the next objective.

Do not add a second metric during the drill unless the first one is unusable. The discipline is the lesson. You are training yourself to reduce a large set of logged numbers into one meaningful trend and then to test your first conclusion before you believe it.

When this principle breaks down​

Plotting the trend before judging the change is a strong default, but it is not magic. It breaks down when the metric is not connected to the decision, when the channel is not measured correctly, when too many variables changed at once, or when the driver context is missing. The source material notes that usable data must be measured correctly and that basic measurement knowledge matters. If the sensor is wrong, the run chart can be neat and still be misleading.

It also breaks down when you expect the chart to provide a cause by itself. A run chart can show that a metric changed. It can show whether the change was gradual, sudden, persistent, scattered, or isolated. It cannot, by itself, prove why. That is why the review process must include other channels, comparison where available, calibration to your driving, and an objective for the next session.

Finally, it breaks down when you use it as a substitute for driving feel. The corpus supports combining objective logged data with subjective driver comments to evaluate what is happening with the car. Your feel is not the verdict, but it is part of the evidence. The best analysis loop is not data instead of feel. It is data organized well enough that your feel can be tested, corrected, and turned into a better next-session plan.

Author Review​

No quiz questions are attached to this lesson.

Sources​

#DocumentChunkPagesScoreCollection
1Analysis Techniques for Racecar Data Acquisition41138569-fa56-a0a4-38c5-301475e4131a211uio_books_raw_v1
2Analysis Techniques for Racecar Data Acquisition9998c72b-304d-0767-6517-dc3b82cea9fe61uio_books_raw_v1
3Analysis Techniques for Racecar Data Acquisition5eeea298-6191-0fb2-1054-b10fe574a80421uio_books_raw_v1
4Analysis Techniques for Racecar Data Acquisition1d32f116-9b81-52c6-919d-dba1c542c01151uio_books_raw_v1
5Data-for-Drivers-PRINTbbb02386-778f-20ec-ad16-b9c016921743161uio_books_raw_v1
6Analysis Techniques for Racecar Data Acquisitionad559d04-3651-61c2-d02b-5455aba0d7cc71uio_books_raw_v1
7Analysis Techniques for Racecar Data Acquisition15474906-387d-234d-cb57-341d5efc4d3a51uio_books_raw_v1
8Analysis Techniques for Racecar Data Acquisition7ae7b884-5466-cf01-8e1a-333086305e8551uio_books_raw_v1
9Analysis Techniques for Racecar Data Acquisition66088a66-7d06-8e55-03eb-967374239bec61uio_books_raw_v1
10Analysis Techniques for Racecar Data Acquisition7500ea75-aa7b-b200-8b21-aa0a2ca9482c61uio_books_raw_v1
11Analysis Techniques for Racecar Data Acquisition52e7d5ab-412b-acc5-fb49-cb0e8d5511b161uio_books_raw_v1
12Analysis Techniques for Racecar Data Acquisition40c913e8-4a2e-8a0b-994c-961bd15b6592201uio_books_raw_v1
13Data-for-Drivers-PRINTb80dc634-a0a7-d6de-d470-353aed47e2a6171uio_books_raw_v1
14Data for Drivers27ec1aea-60bb-f052-9a1a-294b72597f55171uio_books_raw_v1