Friday, May 31, 2013

Video About Raw Speed Article, and Previewing Dover

Wednesday, May 29, 2013

Video about Warm-up Racing and Recap of Charlotte

Tuesday, May 28, 2013

A Simple Way to Measure Raw Speed

(This is a cross-post with BSports StatsInsights)

In this article, I want to come up with a simple way to look at which drivers are the fastest, and to do so in a way that is accessible to a general audience.

What is our metric? It turns out that an effective way to gauge raw speed is to look at the ratio of a driver’s total top 5 finishes to the number of finishes they’ve had on the lead lap.

Why is this a meaningful measure? This ratio tells you, when a driver finishes on the lead lap, what is the probability that they finished in the top 5. A higher ratio implies more pure speed: the driver is faster than the competition and able to convert that speed into results.

This ratio excludes poor performances that result in finishes outside the lead lap, which are races often mired by bad equipment, bad luck, and bad circumstances. Note that because we exclude the worst races, it’s best to consider this ratio as an added dimension to other basic stats. For example, when paired with overall points rank, it's a great way to identify drivers who have a greater potential to outperform their peers. 

Look at Figure 1 below to see who has the best pure speed ratio so far in 2013.

Notice Kyle Busch at the top of the list. After 12 races, he’s had 8 lead lap finishes, and 5 top 5s.  That 5/8 ratio is the 63% you see above.

Denny Hamlin has 3 top-5 finishes in 5 lead-lap-finish races, or a ratio of 60%.

This metric shows how effective Busch and Hamlin have been throughout the season.

Note that this statistic doesn't tell us the complete story. For example, drivers like Paul Menard and Aric Almirola, despite both being in the top 12 in points, are at 0%. 

What does this mean? Menard and Almirola have achieved their rank in the standings not due to pure speed, but rather due to consistency and smart racing.

At this moment, Kurt Busch is 18th in points. But in the pure speed ratio, he ranks 5th overall at 50%, up there with Carl Edwards and Brad Keselowski. This tells us that when Busch has his car setup together, he is doing fantastic. Anybody watching the races this year has known this, but now we have a way to quantify it. Busch is 18th in points not because he lacks speed, but because he lacks consistent finishes. On the flip side, Ricky Stenhouse is 17th in in points because he has finished every race this year, helping to offset his lack of speed, as you see he has a 0% in the pure speed ratio.

What can we do with this new information? And who can use it?
  • Fantasy league participants can find high-potential drivers who can convert their speed into wins. 
  • Team owners looking for young talent can see which drivers have the ability to make it at the next level
  • The media can better identify which drivers are showing pure speed, in a way that filters out distractors such as crashes and bad luck.

Monday, May 20, 2013

Why Dale Earnhardt Jr. and Denny Hamlin Should Race More Often on Saturdays

(This is a cross-post with StatsInsights)

There has been much debate about whether Sprint Cup drivers should also be running in the support series races (Nationwide and Truck series). We've heard terms like “Buschwhackers” or "claim jumpers" to refer to drivers who consistently do the weekend double. We also have seen rule changes to deter veteran Cup drivers from racing in the support series, thereby opening up more spots for younger drivers.

Does running a race the day before help or hurt the performance for Sprint Cup drivers in their Cup races? Do these "warm-up" races earlier in the weekend help a driver when it matters most? Those are the questions we explore in this article.

One school of thought is that the warm-up races are helpful. Drivers stay sharp with extra seat time, and are happiest racing. Besides, it's better than sitting around doing nothing. Proponents of this philosophy also believe that more time on the track serves as a learning lesson for what will happen the next day, giving them a leg up over the Cup competition.

Then there is the opposite theory: that drivers need to stay focused on just their Sprint Cup duties. Advocates of this position argue that any other racing is a distraction from a full commitment to your Cup car. Drivers should spend the weekend improving their car, communicating with their crew chief, and fine tuning the setup. Trying to handle multiple races in the same weekend can be overwhelming physically, mentally, and emotionally. Cup performance will suffer as a result. After all, it’s better to do one thing well than to do two things poorly.

So which theory is right?

As usual, we turn to the data.

I looked at all races run in the big 3 series, starting in the year 2000. I filtered out only the drivers who had a minimum of 15 Sprint Cup races with, and 15 races without, a "warm-up" race. For our experiment, we define a "warm-up" as racing the day immediately prior to a Sprint Cup race.

A first glance at Figure 1 indicates that, for the most part, the warm-up races do not affect average finish in the Cup series. The diagonal line is where the warm-up races have no effect. You can see that most drivers are very near that line.

But let’s take a closer look: this overall effect is not true for all drivers. Some are significantly above or below the diagonal line.

Two things stand out:

1)    Most drivers benefit from having a warm-up race (we can claim this because a majority of the points are above the line)
2)    The drivers that are hurt by warm-up races are hurt a lot (the points below the line are generally much further away from the line than the ones above)

It turns out that most drivers do slightly better with a warm-up race, and other drivers do worse.  The ones that do worse can often do significantly worse.

Figure 2 breaks down these differences. Here we focus on the bigger names in the sport, filtering in drivers with above-average results:

The bars in Figure 2 show the improvement a driver has in average finish when they run a warm-up race the previous day.

Denny Hamlin and Dale Earnhardt Jr. improve about 3 spots in their average finish with a warm-up race already under their belt.

In contrast, Kurt Busch does worse by 8 spots, and Brad Keselowski is not far behind with a drop of 6 spots.

What have we learned today?
This analysis reveals several interesting ideas.

1)    Going back to our pair of competing philosophies around warm-up races: it turns out that both are right; depending on the driver, warm-up races can be a good thing or a bad thing.
2)    The effect is asymmetric: most drivers benefit slightly, and some drivers are hurt in a big way
3)    Denny Hamlin and Dale Jr. should be getting into more Nationwide races, to help their Sprint Cup efforts. Kurt and Brad, on the other hand, might want to consider throwing in more rest days.

Tuesday, May 14, 2013

Kyle Busch and Kasey Kahne are NASCAR's Most Volatile Drivers

(This is a cross-post with StatsInsights)

A driver's average finish is often discussed in the media as a benchmark for how well that driver performs. However, this statistic only gives part of the picture; to really get a complete sense of a driver's performance, we also need to consider the volatility of those finishes.

What do we mean by volatility? Let's start with an example.

If driver A finishes 10th and 10th in two races, that's an average finish of 10.  But if driver B finishes 1st and 19th in two races, that's also an average of finish of 10. However, driver B shows a lot more volatility in getting to the same average.

Drivers with higher volatility are more likely to have extreme finishes. Lower volatility means the drivers consistently finish around the same spot. The statistical parameter that we'll use to help us make this assessment is the standard deviation of a driver's race results.

Looking at the Data

Figure 1 shows all the drivers who have run at least 100 races since 2005 (when Loop Data statistics became available) and who have had an overall average finish better than 22 (to discard teams with consistently bad results, start-and-park teams, etc).

We see here that Kyle Busch, Kasey Kahne, and Kurt Busch are the three most volatile drivers. They have a history of extreme finishes, which means that they are more likely than other drivers to either finish way up front or way in the back.

Notice that championship drivers are generally in the upper half of the volatility scale. Jimmie Johnson, Kurt Busch, Jeff Gordon, Tony Stewart, and Brad Keselowski all show above-average volatility compared to other drivers. This suggests that volatility is a good thing, not a bad thing: In order to win championships, you need some volatility (but not too much).

Figure 2 shows the most volatile individual driver seasons since 2005:

What observations can we make?

Notice we see many of the top drivers on this list.  In fact, drivers like Matt Kenseth, who don't normally seem to have extreme finishes, still appear on this list.

Also notice in Figure 2 we see Mark Martin's 2009 season. This season is interesting because he was 34th in points after 4 races, but got all the way to second in the final points standings. That's because of high volatility: engine failures and crashes early in the year, followed by a string of wins later in the year.

Jimmie Johnson has three seasons on this list, but in two of them, he didn't win the title. And in 2010, he had to come from behind in the final race of the year to win the cup. Other than Johnson in 2010, there are no championship seasons shown in Figure 2.

This data helps reinforce the idea that there's an "ideal" level of volatility, an above-average amount that can help a driver have a great season.


How can people use this info? If you are a fantasy league player or team owner, it's important to look beyond just the average finish when you pick which drivers you think are going to do well. Look at their volatility too, because that suggests the ability to get race wins (but also an increased tendency to crash out of races). High volatility drivers give you more risk and more upside. Low volatility drivers are consistent, but the numbers suggest they probably won't win you many races.

Look at the example below in Figure 3. You see that Kyle Busch has about the same average finish as Mark Martin, Greg Biffle, and Clint Bowyer (similar sized magenta bars). But if you were betting on who would win a race, you'd want to go with Busch because he has a much higher volatility.

The key is adding this second datapoint for every driver to get a better picture for how they are driving. A driver needs to have some volatility to have a great season. Not too much, but just the right amount.

Sunday, May 12, 2013

Looking Back on My Recent Predictions

It's time for another prediction review. We are due for a refresh, as the last time I did this was two month ago, after the Bristol race.

Each week, when I film a video for BSports, I provide the names of a few drivers who I believe will do well at that upcoming track, based on the stats of their past performances.

Remember, these are the goals:

  • Highlight drivers with strengths at the specific track that weekend.
  • Identify drivers who, based on my statistical models, have a greater-than-expected chance to shine each week
  • Predict an overall winner
Let's see how I've done.

Denny Hamlin (finished 2nd)
Jeff Gordon (finished 3rd and led 16 laps)
Greg Biffle (finished 13th)

Potential First Time Darlington Winners
Ryan Newman (finished 10th)
Jamie McMurray (finished 16th)
Kasey Kahne (finished 17th after spinning out while leading the race with 35 laps to go)

Dale Earnhardt, Jr. (finished 17th after restarting 7th with 2 laps to go)

Dark Horses
Ricky Stenhouse (finished 13th, led 2 laps, and almost won the race if they had ended the race at at the first rain caution. Restarted 5th with two laps to go)
Brian Vickers (took over for Denny Hamlin at the first caution, but was involved in a wreck early)

Kyle Busch (finished 24th and led 40 laps.  Kyle was leading late before getting involved in a wreck)

Dark Horses
Jeff Burton (finished 5th and led 7 laps, was leading with 6 laps to go)
Carl Edwards (finished 6th and nearly took the lead from Kyle Busch late in the race)

Jimmie Johnson (finished 3rd and led 9 laps)
Greg Biffle (finished 19th)
Tony Stewart (finished 21st)

Potential First-Time Kansas Winners
Martin Truex, Jr. (finished 4th and led 46 laps)
Aric Almirola (finished 8th)
Kurt Busch (finished 15th)

Greg Biffle (finished 4th)
Matt Kenseth (finished 12th)
Tony Stewart (finished 21st)

Potential First-Time Texas Winners
Kyle Busch (WON the race and led 171 laps)
Brad Keselowski (finished 9th)
Martin Truex, Jr. (finished 2nd and led 142 laps)