Friday, June 28, 2013

Video Previewing Kentucky Race

Post-Race Update: How did my Kentucky favorites do?

Expected Winning Percentage
17% Jimmie Johnson (led 182 laps and finished 9th after losing the lead with 24 laps to go)
11% Matt Kenseth (led 38 laps and WON the race)
10% Tony Stewart (finished 20th)
10% Kyle Busch (finished 5th)
  8% Kurt Busch (finished 6th)
  7% Brad Keselowski (early accident caused a 33rd place finish)
  7% Martin Truex, Jr. (finished 7th)

Thursday, June 27, 2013

Kentucky is Sprint Cup's newest track. So how long do we have to wait until the next one?

(This is a cross-post with BSports Stats Insights)

This weekend's race in Kentucky will be only the third at that track since NASCAR's Sprint Cup started hosting races there in 2011.

There are only 36 races each season, and it's extremely difficult to add a new track to the schedule, because another track has to lose a date for that to happen. Simply swapping race dates within the same tracks is a difficult challenge.

Here is a listing of the current Sprint Cup tracks, and when they hosted their first race:

As you can see, there were no new tracks for 10 years prior to Kentucky making an appearance.  If you just looked at that number, perhaps you’d guess that the next new track would come around in 2021.

However, notice some patterns:

All our current tracks came in three waves (circled above):

1949-1961: eight new tracks in 13 years
Then an 8-year gap
1969-1974: four new tracks in 6 years
Then a 14-year gap
1988-2001: ten new tracks in 14 years
Then a 10-year gap
2011: Kentucky

That would suggest Kentucky started a new wave of track additions. NASCAR has shown a pattern of big pauses in between periods of rapid growth. Is Kentucky the first track in the newest wave, one we can't even see yet?

Below are the same tracks, now color-coded to show the separate waves:

The first set of tracks was in the South. Then the second wave moved north (Dover, Michigan, Pocono), and the third wave went out west (Arizona, California, Nevada, Texas).

Where is Kentucky leading us? Perhaps back to NASCAR's Southern roots?

Adding new tracks wasn't always a rare event.

In the 1950s and 1960s, races were held on many different tracks: Then NASCAR's "modern era" began in 1972, and most of the tracks were eliminated.

Getting back to 2013: If new tracks were added to the Sprint Cup circuit, which ones would be the candidates?

As an easy starting place, we can look to the Nationwide and Truck Series for inspiration, considering tracks there as available for promotion to Sprint Cup. (Remember that Kentucky speedway held Nationwide races for several years before making it to Sprint Cup.)

  • Iowa
  • Mid-Ohio
  • Eldora
  • Bowmanville

Plus two tracks that used to hold Cup races:

  • Elkhart Lake (one Cup race in 1956, currently used by Nationwide)
  • Rockingham (annual Cup races 1965-2004, currently used by Trucks)

Another option of course, would be to build a brand new race track and use it for Sprint Cup right away. For example, Kansas and Chicago were added to the Sprint Cup schedule without any prior history of Nationwide or Truck racing.

Kentucky is this weekend. It may be Sprint Cup's newest track, but that fact may not be true for long. If we see a new wave of tracks come on board, don't be surprised, it might just be that fourth wave happening.

Consider also the effects of the economy: if another spurt in economic growth happens, new tracks may be built (like in the New York City area, where NASCAR has previously tried to build a track). And, as more tracks join the circuit, they may affect the overall standings. For example, different drivers would benefit if more road courses were added, or more superspeedways, or short tracks. Obviously none of these things are going to change this season, but as you watch Kentucky this weekend, keep these ideas in the back of your mind.

Tuesday, June 25, 2013

Here is TNT using my stats again: Sonoma edition

Martin Truex had a 67% chance of winning the race, based on his position in the race and the number of laps left to go.

He did in fact win.

Saturday, June 22, 2013

BSports Video Previewing Sonoma and Discussing Road Course Ringers

Sonoma Post-Race Prediction Check

Sonoma Favorites:

  • Jeff Gordon (Finished 2nd and led 4 laps after a pit road penalty)
  • Kurt Busch (Finished 4th and led 15 laps after two speeding penalties)
  • Kyle Busch (Finished 35th after a few incidents)

Threats for First Sonoma Win:

  • Martin Truex, Jr. (WON the race and led 51 laps)
  • Marcos Ambrose (Finished 7th and led 18 laps)
  • Ryan Newman (Finished 15th)

Thursday, June 20, 2013

The Real Road Course Ringers: They're Not Who You Think They Are

(This is a cross-post with BSports Stats Insights)

As we go into Sonoma this weekend, for the first of two road-course races this year, there is always talk about the road course ringers who are specialists at turning left and right, and dominate the other drivers when not on our usual oval tracks.

Look at this list:
Left and right turns this weekend at Sonoma

Brad Keselowski
Carl Edwards
Clint Bowyer
Denny Hamlin
Greg Biffle
Jeff Gordon
Jimmie Johnson
Juan Pablo Montoya
Kevin Harvick
Kurt Busch
Kyle Busch
Marcos Ambrose
Robby Gordon
Ron Fellows
Scott Pruett
Tony Stewart

Lots of big-time "oval" names in there, right? Yet, here's the surprise: these are the drivers who have performed the best on road courses in the past 10 years.

If you apply the Pure Speed Ratio of top 5 finishes to lead lap finishes to the drivers on this list, you see this ranking:

(The data includes only drivers with 5 or more lead lap finishes at road course events in the last 10 years)

What do we observe?
  • Marcos Ambrose is truly dominant on road courses, and is virtually guaranteed a Top 5 finish, barring any major problems
  • Scott Pruett (a noted road course expert) is also excellent, proving his reputation in the numbers
  • Ron Fellows (another ringer) is at the bottom of this list, not where you'd expect him to be, below many typical oval drivers
  • Other famous ringers like Boris Said (just a single Top 5 in 14 lead-lap-finishes), don't even make this list.
Just to give you some context about how rare it is for a ringer to actually succeed in Cup, here are the names of several other road-course experts:

Andrew Ranger
Andy Lally
Andy Pilgrim
Anthony Lazzaro
Brandon Ash
Brian Simo
Butch Leitzinger
Max Papis
Patrick Carpentier
P.J. Jones
Patrick Long
Tom Hubert
Tomy Drissi
Tony Ave

The total combined number of top 5 finishes among all these drivers is 0. Think about that. None of these drivers have ever scored a top 5 finish on a road course.

Let's go back to the original list of drivers, now with some color-coding:

I've highlighted them in three colors:
  • Beige is for the true ringers, who are clearly outperforming compared to their performance on oval tracks. Notice that every single one of these drivers came from another series. None of them are native NASCAR veterans– for example, Pruett, Gordon, and Montoya all ran IndyCar.
  • Blue is for any Sprint Cup Series Champions. Did you see how many of them show up here?
  • Gray is for any driver who has been close to a title. Look at these names again: Bowyer, Harvick, Busch, Edwards, Hamlin, Biffle. All of them have come very close to winning the title (many second and third places in the final season standings) 
Those three categories cover every single driver on this list. Notice you don't see any "mediocre" driver on the list. We see only the guys who have all won titles, or come the absolute closest to winning them. This list of drivers is responsible for 74% of all the top 5 finishes at road courses in the last 10 years.

So what's my point? In general, minus a couple of extreme exceptions:

The best road course drivers are in fact the best oval drivers.

Consider the names that don't make this list:

Dale Earnhardt, Jr.
Jamie McMurray
Jeff Burton
Joey Logano
Kasey Kahne
Martin Truex, Jr.
Paul Menard
Ryan Newman

Etc. Etc. 

These guys are all good drivers, but haven't shown the type of full-season dominance that the elite drivers above have all shown.

What can we learn from this?
  • Race car driving really is a skill: The best NASCAR drivers are the best no matter where the track is.
  • Contending for titles means you need to be good everywhere
  • The true ringers are the same guys we see every weekend: You don't need road course experts if you already have a top-level driver

Thursday, June 13, 2013

BSports Video about Hamlin's 42%, and Michigan Preview

Post-Michigan Update: How did my picks do?


  • Greg Biffle (led 48 laps and WON the race)
  • Carl Edwards (led 16 laps and finished 8th)
  • Jeff Gordon (finished 39th after crashing early)

Threats for First Michigan Wins:

  • Martin Truex, Jr. (finished 3rd)
  • Brad Keselowski (finshed 12th)
  • Jimmie Johnson (led 18 laps and finished 28th after hitting the wall from 2nd place with 3 laps to go)

Note that Johnson still has never won a race at Michigan. As we discussed in the video, he's the worst performer in the final 10% of Michigan races.

Here is TNT Using My Stats for Bloomberg Sports

Based on Jimmie Johnson's position with 20 laps to go, we calculated him to be a heavy favorite to win the race. He in fact did the win race.

Denny Hamlin's 42% Chance of Getting 2 Wins Before the Chase

(This is a cross-post with BSports Stats Insights)

Based on my models, the cutoff place for 20th place after 26 races will be about 670 points. Denny Hamlin has 285 points right now.  He would need 385 more points in the next 12 races. That's an average of 32.1 points per race (or an average finish of 11.9 or better in the next 12 races). You can review myprevious article about predicting Chase cutoffs to see how I approach this.

Just achieving this 11.9 average finish will be a challenge for Hamlin: his average finish so far this year is 16.2. Hamlin will need to step his game up and avoid crashes to get there. Needless to say, he doesn't have much room for error. That’s just the first step.

If he does get into the top 20, he will need to be among the two drivers with the most wins (of drivers ranked 11th through 20th).

In the past years of this format, a driver with 2 or more wins has always been guaranteed a spot in the Chase. Drivers with 1 win may or may not make it. Right now, of drivers ranked 11 through 20, Tony Stewart has a single win. The other nine drivers have 0 wins.

All this data suggests that Hamlin will need two wins to make it. What are the probabilities of him getting that pair of wins?

Here's what we know:
  • We have completed 14 races so far this year.
  • There are 12 more before the Chase: races through 15 through 26.
  • In the past 5 years, if you only look at races numbered 15 through 26, Hamlin has won 7 out of those 60 total races, for a winning percentage of 11.7%.
Note that 11.7% is the winning percentage for each individual race. The real question is how do we figure out his ability to get multiple wins in this same time period?

Our tool for answering this question is the binomial distribution. You can use the binomial distribution to answer questions like "If a coin flip comes up heads 50% of the time, and I flip my coin 12 times, what's the chance I get 2 heads in those 12 flips?"

Or…we can ask the exact same question, but applied to winning races: "If Denny Hamlin wins a race 11.7% of the time, and he races 12 times, what's the chance he gets 2 wins in those 12 races?"

Here's what the binomial distribution tells us, given Denny’s historical winning percentage:

What’s the probability of Denny winning 2 or more races? It’s simply the sum of the probabilities for the numbers of 2 or more wins. That’s how we arrive at a 42% chance.

Now take the same approach, and apply that to everybody else outside the top 10.

If you look at this table, you’ll see that Denny has significantly better prospects than *anyone* ahead of him. It bodes well for Hamlin that none of the drivers ahead of time are strong threats to win multiple races before the Chase starts.  (Remember that Tony Stewart is halfway there with his first win. He has a 56% chance of winning at least one more time before the Chase starts, to get to two total wins.)

There you have it: using some math and historical data, we can see that Denny Hamlin has a 42% chance of getting those two wins.

If anything, given his racing performance this year, getting the 2 wins might be more doable than getting into the top 20 in points. We’ll definitely be keeping an eye on Denny to see how he does over this final stretch.

Tuesday, June 11, 2013

Jimmie Johnson Won at Pocono (and how have my other predictions done)

It's been a month since our last prediction check.  Let's see how I've done.

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.

All-Star Race:
Jimmie Johnson (WON the race and led 10 laps)
Kyle Busch (led 29 laps and finished 3rd)
Ryan Newman (finished 13th)

Jeff Gordon (finished 3rd)
Kyle Busch (led 150 laps and finished 4th)
Jimmie Johnson (led 143 laps and was leading before a restart penalty, finished 17th)

Potential First-Time Dover Winners
Clint Bowyer (finished 6th)
Kevin Harvick (finished 8th and led one lap)
Aric Almirola (finished 18th)

Jimmie Johnson (WON the race and led 128 of 160 laps)
Denny Hamlin (finished 8th)
Jeff Gordon (finished 12th)

Potential First-Time Pocono Winners
Dale Earnhardt, Jr. (finished 3rd)
Clint Bowyer (finished 15th)
Mark Martin (finished 19th)

Wednesday, June 5, 2013

Moving on Up: The Surprising List of Drivers Who Gain Spots During a Race

(This is a cross-post with BSports StatsInsights)

This week, we take a look at a very simple stat to identify which drivers actually improve during a race. Let's call it "Positions Gained in Race".

Our method is simple: take the top 35 in driver points for 2013, and compare each driver’s average starting position with their average finishing position. This shows us who moves up during the race, and who loses spots.

First, here is the raw data, sorted by average finish in Figure 1:

What can we infer? It’s hard to visualize in tabular form, so let’s plot the data in Figure 2.

Generally speaking, if you start towards the back, you’ll finish towards the back.

But now here is the twist: when you compare the difference between the starting and finishing positions, and sort accordingly in Figure 3, a few things stand out:

  1. Toyota fields 5 of the bottom 6 cars on this list. Specifically, all 3 Joe Gibbs Racing cars are in this bottom 6.
  2. Danica Patrick, despite her overall struggles this year, is doing a good job moving ahead during a race. She is a better racer than qualifier.
  3. Cars that start up front tend to lose spots during a race, and cars that start in the back tend to gain spots during a race.
  4. The exception to #3 is Jimmie Johnson, who starts very well (11.3) and then still improves nearly 3 spots from there. Amazing.

Figure 4 is a visualization of the table above:

From above, point #3 is worth exploring: Do cars that start up front generally lose that advantage during a race?

Refer to the Figure 5 below.  The bottom left corner shows those cars which start well but lose that advantage during the race. Notice they have a starting position around 10th or better, but generally lose spots during the race.

Look at the upper right corner: these cars start poorly (in the high 20s and 30s), but gain spots during the race.

Today's Lesson

There is a tendency for mean-reversion during a race: finish position is generally more centered than starting position. When we say mean-reverting, that suggests drivers who start up front tend to fall back during the race. Drivers that start poorly find their way higher during the race. Both the best starters and the worst starters will move toward the middle by the finish.