Friday, April 11, 2014

The Chances of Seeing 16 Unique Winners Reach the Chase

With Joey Logano's win at Texas, we now have 7 unique winners through 7 races. This matters a lot in 2014, because the eligibility for the 16 Chase spots is primarily based on wins, and secondarily on points. If there are at least 16 unique winners after 26 races, then all winless drivers (except possibly the overall points leader) will be eliminated from joining the Chase. Right now that list of winless drivers still includes champions like Jeff Gordon, Matt Kenseth, Jimmie Johnson, and Tony Stewart.

Specifically, the number of winners will affect how teams strategize in the next 19 races: if there will end up being exactly 16 winners, you'll need to gamble to get a win. If there won't be enough winners, you'll need to be consistent to stay at the top of the points. If there are 17 or more winners, you'll need to gamble for a win and be high enough in points to qualify. That's why the topic is important: the number of winners -- or projected winners -- will dictate team strategy as we approach the Chase.

We've seen this topic already debated heavily in the press, for example Nate Ryan at USA Today, Bob Pockrass at the Sporting News, and David Caraviello at

Today we look at the probability of seeing 16 unique winners by the time the checkered flag waves at Richmond. First, we use our base assumptions of historical winning percentages during the 2009-2013 seasons:

The final column uses the binomial distribution to calculate the probability of each driver winning one race out of the next 19, given their historical win rates. For example, drivers like Jeff Gordon and Kasey Kahne win 4% of their races, which translates to a 53% chance of winning one race out of 19. Compare that to Matt Kenseth, whose 8% historical average translates to an 85% chance of winning one race in the next 19.

The highlighted drivers are the 2014 winners, and the cutoff line shown is after the 16th driver, indicating that Dale Earnhardt, Jr. and Joey Logano have taken spots that mathematically were more likely available to others. Based on that data, we can see the likelihood of unique winners throughout the season. Here's what it was at the start of the season, before the Daytona 500:

The outcomes are based on 1 million trials of race outcomes: at the start of the year, the chance of 16 unique winners was small: only about 5 percent, with an additional 2% for all totals above 16.

After the first six races saw six unique winners, the percentages obviously increased. Heading into our most recent at Texas, the likelihood of ten additional (or 16 total) winners before the Chase had shifted closer to 20%. Remember the calculations take into account the specific drivers who are still without wins: it knows that drivers like Johnson are definitely due to get one. 

After Logano's win at Texas, the numbers moved even further towards the 16+ camp. We now have a 17% chance of seeing exactly 9 more winners (to reach 16 total), with an additional 10 percent for all numbers above that. In total, there is now a 27% chance of seeing at least 16 unique winners before the Chase starts.

That means the likelihood of having at least 16 winners is now a reasonable one. It's not the most probable scenario for now, but it does suggest that Chase spots are filling up fast. Worrying about points should not be the most important thing on the minds of teams.

In the most likely scenario, we will probably end up with 7-8 more winners, reaching a total of 14-15 uniques, leaving only 1-2 spots for drivers to get in on points. For almost all drivers, getting one win through good luck or a smart gamble will be easier than keeping the consistency necessary to be a top-2 non-winner in the points standings.

Sunday, March 16, 2014

An Interview with Lesa France Kennedy

This week, published my interview with NASCAR vice-chair Lesa France Kennedy. Here is the complete article.

Here is our weekly BSports video:

Moving on to this week’s picks:

Bristol Favorites
Kyle Busch
Jeff Gordon
Brad Keselowski

Threats for First Bristol Win
Greg Biffle
Clint Bowyer
Ryan Newman

How did last week’s picks fare:

Las Vegas Favorites
Jimmie Johnson (led 34 laps and finished 6th)
Matt Kenseth (led 2 laps and finished 10th)
Tony Stewart (finished 33rd)

Threats for First Las Vegas Win
Dale Earnhardt, Jr. (led 51 laps and finished 2nd, running out of gas on last lap)
Kasey Kahne (finished 8th)
Greg Biffle (finished 22nd)

Saturday, March 8, 2014

New Chase Rules Creating a Spiral of Aggression

Two races down, and two drivers are already in the Chase. Only 24 more chances to get in. The most important lesson to take away from Phoenix is that teams will do whatever they can to win. Gambling on fuel strategy, tires, track position, even just plain old aggressive driving are all in play. This trend will only continue: each week that goes by is one less chance to qualify for the Chase. For the drivers who don't have wins, they will need to gamble increasingly out of desperation to secure a Chase spot. And for drivers who do have wins, like Earnhardt and Harvick, gambling will be the smart way to go to capitalize on their advantage, to earn additional Chase bonus points.

As we discussed last month, the new rules would create much more uncertainty, randomness, and volatility. Certainly that has proven to be true, especially at Phoenix when we saw many drivers try different approaches at getting that win. Harvick was just too strong for them. It’s my prediction that this dynamic will create an increasing spiral of aggression. Teams without wins will do whatever is necessary to get one; and once they do get a win, they can gamble even more because they have nothing to lose.

A few weeks ago, I calculated expected percentages for the drivers that would most likely earn a win. Both our 2014 winners were in the top-16 of most likely winners. And with wins by Dale Jr, and Kevin Harvick, two of our expected 'in-the-Chase' drivers have already been checked off the list.

Winners, Losers, Gamblers, and Helpers

With Daytona specifically behind us, that leaves only four more "wild card" races left before the Chase starts: two road course and two plate races. These are the four best chances for non-Tier 1 drivers to earn a spot in the Chase.

As the year continues, we will be better able to analyze if teams are taking bigger risks – and succeeding. By analyzing loop data like laps led, percentage of laps in top 15, and average running position, we can see if the winners in 2014 are racing with a materially different style than in past years. An increase in gambling frequency and success might reveal itself if we see more winners who perhaps led fewer laps than in the past, or had a worse average running position than historical averages.

Also, we need to pay attention to team orders allowing for even distribution of wins. For example, a few weeks from now, if Harvick is leading a race late, and second place is a winless teammate, will he find a reason to pull over and let the teammate by? Their overall team would be better off if both drivers each had one win, rather than two for Harvick and none for the teammate. It's likely that teams have already planned for this scenario.

Weekend Forecast

Here are the picks for this weekend’s race:

Las Vegas Favorites
Matt Kenseth
Tony Stewart
Jimmie Johnson

According to my models, these three former champions are by far the best-performing drivers at Las Vegas, based on their combination of laps led and consistent finishes. They are in a cluster by themselves, fully separated from everybody else.

Threats for First Las Vegas Win
Dale Earnhardt, Jr.
Kasey Kahne
Greg Biffle

Here's a good shot for Earnhardt to pick up a second win. Both races this year have been won by Hendrick engines -- a good opportunity for somebody like Kahne to get his first win.

Monday, February 24, 2014

How Did My Daytona 500 Picks Do?

It's time to review how my Daytona 500 picks fared in yesterday's race:

Daytona 500 Favorites
Dale Earnhardt, Jr. (WON the race and led 54 laps)
Kyle Busch (led 19 laps and finished 19th)
Tony Stewart (finished 35th - engine problems)

Threats for First Daytona Win
Denny Hamlin (finished second and led 16 laps)
Kurt Busch (finished 21st and led 15 laps - had a late spin)
Clint Bowyer (finished 42nd - engine failure)

Going 1-2 is not bad, given how many problems affected big name drivers.

Thursday, February 20, 2014

Race Car Drivers are Made, not Born

Malcom Gladwell, in his 2008 book Outliers, tried to explain why most elite hockey players in Canada were born in the first half of the year. While growing up through the levels of youth hockey, the slightly older children were bigger, faster, and stronger – and as a result were deemed “more talented.” From there things snowballed, as these “more talented” players were singled out for additional development: better coaching, access, encouragement, and confidence. When they grew up to reach the pros, the slightly older players were represented in much higher proportions. Because of the way that system worked, being born later in the year hurt a player's chances of reaching an elite level.

In 2006, the Freakonomics guys also found a similar birth-month effect in World Cup soccer players. It’s also been looked at for corporate CEOs. Birth months also seem to affect issues around learning and health.
What about in racing? Does a birth month effect hold true? Are elite drivers born in a certain part of the year? Do drivers in the first half of the year have the same accumulation of development that happened with Canadian hockey players? Or is racing different enough, both in terms of the skills needed to compete and in how youth racing is organized, that birth month will be irrelevant?

Using Racing Reference’s birthday data on over 850 active drivers, we can see whether elite drivers are more likely to be born in specific months. We split the drivers into three sets, to see if there is a difference in America’s highest level of competition compared to lower levels or other countries:

1) Americans in Sprint Cup
2) Americans in every other series
3) Foreign drivers

Here is what we found:

In all three charts, there is no meaningful difference in births the first half of the year versus the second half. A breakdown by quarters of the year also shows no specific advantage in one time period. The charts suggest drivers from any month of the year have an equal chance to show and develop their talent. What are some reasons for this? For one, youth racing is not a nationally organized league with the same rules in every region, and it's not as stubborn in dividing kids by specific birth years. Unlike other sports, there is not a different racing level just for 6-year-olds, and just for 7-year-olds, 8-year-olds, etc. Racing divisions may have drivers of many ages competing against each other. The difference of a few months would not matter as much to development, as it might in a sport like hockey, where physical size and strength makes more of a distinction between competitors.

The good news for aspiring drivers is: it doesn't matter when you’re born; everybody has an equal chance.

Friday, February 14, 2014

Who’s Making the Chase This Year?

As we saw last week, the new Chase format is practically win-and-you’re-in: drivers with a win in the first 26 races will have a spot in the 16-driver Chase field. So who’s making the Chase? We can answer that question by figuring out which drivers are most likely to win a race.

We’ll use each driver’s career winning percentage to give an expectation for what’s possible in 26 races. By using each driver’s career total of wins and race starts, we get their career winning percentage. With that number, the binomial distribution tells us how likely it is for each driver to win 1 or more races.

The table above ranks drivers by their career winning percentage. The percentage numbers colored in red highlight the most likely number of wins per driver through 26 races.

The drivers at the top of the list should be expected to win at least one race. But if we look at the data, there are some misleading pieces. For example, Jeff Gordon has recently not won as many races as his career average would suggest, but he shows up in the second spot on the list. In the last few years, he’s won only about 3% of his races – putting him in that Clint Bowyer / Dale Jr area, which may be a more reasonable place for him. Speaking of Dale Jr, he also hasn’t won many races over the past few years, so he’d be even further down the list. That leads us to think about other who don’t win – this list doesn’t contain those drivers who have never won a race (Danica Patrick, Ricky Stenhouse, Kyle Larson, Aric Almirola, Austin Dillon).

But overall, the table gives us a sense for what should be expected for most drivers. Kyle Busch, for example, has an 11% chance of winning 4 races, but also a 10% chance of winning no races. Kevin Harvick has a 10% chance of getting 3 wins, but a 27% chance of winning nothing before the Chase starts.

We can summarize the data to see the one thing that matters: will a driver win at least one race or not? By taking the table above and combining the probabilities for any amount of race wins, we get this:

The most interesting names are at the bubble, around the 50% area. Look at contenders like Dale Jr, Clint Bowyer, and Jamie McMurray. Whether they get a win or not is almost a coin flip. These guys don’t win a lot of races historically, so trying to get just 1 win in 26 chances is anything but a guarantee.

Moving further down the list, drivers like Brian Vickers, Marcos Ambrose, and David Ragan have shown success on road courses and restrictor plate tracks, giving them a decent chance at sneaking their way in.

With an expanded range of 16 names, the biggest names in the sport – Jimmie Johnson, Tony Stewart, Carl Edwards for example – should have strong odds of getting in. But it’s less certain for those drivers on the cusp.

Thursday, February 6, 2014

All of a Sudden, Everything Got a Lot Less Predictable

NASCAR announced last week that the championship would be determined almost entirely on race wins, rather than on accumulated points. With that giant change from the sports past, it will be much more difficult to make predictions. NASCAR Chairman Brian France apparently thinks the enemy was “math”. In a sport with more engineers and numbers-minded participants than any other, where fuel mileage still gets calculated by hand rather than a dashboard gauge, the basic arithmetic required to add points was apparently becoming too much for people to deal with.

"No math. No bonus points. It's as simple as it gets," France said. He goes on to say "The avid fans like it because they don't particularly care for points racing, even though they understand it. The casual fans don't understand points racing ... often, with all the mathematicals, you've got to have a computer next to you to figure out who is in and who is out at a given moment. (This) clears all that off and then emphasizes winning, which everybody understands."

So with that logic, we move forward into 2014, trying to find a smarter way to analyze the best strategies. Looking at data between 2001 and 2013 (every year with 36-race seasons), we see that a driver's end-of-season points ranking was tied most closely to his average finish and the number of top-10 finishes:

 Points rankings had a low correlation with race wins. Drivers could do well in the season without any wins, and conversely they could do poorly despite many wins, because we know the difference between first and second place is often very random. On the contrary, accumulating a high average finish, or many top 10s, is not random over the course of a full season.

Now that everything has changed without "all the mathematicals", and wins become the primary factor, it’s important to see how wins correlate with these same variables:

The chart shows that Laps Led and Top-5s are the two most important places to focus. At the very bottom, Laps Completed and Races at the Finish are the two least relevant factors. This suggests that drivers and teams consider gambling much more often. High risk, high reward bets for track position will be the name of the game in 2014. Consistency doesn’t matter anymore, as the new playoff system invites much more risk-taking.

The eventual champion could be very random and unpredictable. In the chart below, the top row shows that Average Finish and Top 10s were very tightly correlated to points rank. There is a very dense pattern of points along the line, and no outliers. Either of those numbers could have told you almost everything you needed to know about where a driver would finish in the standings.

However, the bottom row shows that wins have little relationship with either Top 10s or Average Finish. The datapoints are scattered all over, including many outliers. The drivers with the most wins are not the ones with the best average finish or top 10s. 

So what do we learn from this? We should expect the upcoming season to be a very interesting year. There will be many unpredictable results. Among the first 26 races, we will have three plate races and two road course races: five high-volatility opportunities for mediocre drivers to gamble their way into the playoffs.

Thursday, January 23, 2014

Chase Volatility: Who Gets In and Who Gets Left Behind

As NASCAR continues to tweak its Chase format, expanding from 10 drivers to 12 to 13 -- to possibly 16 now -- we should all remember they are never going to be able to get what they want: a perfect solution to answer all their problems, guaranteeing that stars are in contention until the end, while maintaining the value of the regular season.

The volatility of who makes the Chase – and more importantly, who doesn’t – has been baked in as a natural occurrence of racing since they started doing this a decade ago.

There are always several drivers who can make the Chase one year but miss it the next.

This natural rotation of Chase drivers is a fact of life. As we see in the table above, at least two drivers miss the Chase the following year, with that number going as high as seven in 2005. The average is consistently around four. Each year you can safely assume that on average, four Chase drivers from the previous year won’t make it back, leaving four spots open for those looking to get in. 

Four sounds right going into 2014. Who are some likely candidates? Assuming NASCAR leaves the Chase field at 12, we can expect Brad Keselowski and Denny Hamlin to make their way back in. Tony Stewart returns from injury and should get in as well. Perhaps Ricky Stenhouse will improve over his rookie season to become a contender. Martin Truex Jr might perform well as he takes over Kurt Busch’s Furniture Row ride. We’ll also see talented rookies Kyle Larson and Austin Dillon both have a shot at Chase contention.

Not all of these drivers will get into the Chase, but it’s reasonable to see about 4 of them making it in, displacing 4-5 Chase drivers from 2013 (since last year had a 13th driver).

The approximately four drivers who rotate out of the Chase on an annual basis usually do it by barely missing the Chase. They don’t typically fall far out of the Top-12 in the years they miss. One great example of this is between 2005 and 2006: Tony Stewart, Greg Biffle, and Carl Edwards finished 1-2-3 in 2005, but were the three best non-Chase drivers in 2006 (finishing 11-12-13). 

The average finish in the points standings for drivers in the Chase is 6.5, and in the year following it is 10.0. A drop of 3.5 spots on average highlights the fact that if they miss the Chase, it’s because they narrowly missed it.

Their performance doesn’t drop off radically, but just enough for them to get beaten by a few other drivers. For instance, consider 2013 Chasers like Ryan Newman, Kevin Harvick, and Kurt Busch – all might have problems integrating with new teams in 2014. And Dale Earnhardt Jr. will have to cope with the eventual departure of his crew chief. If these drivers end up missing the Chase Top 12, you would most likely see them very nearby, in that 13-16 range. 

If NASCAR does in fact expand the Chase to 16 drivers, it might be enough to save everybody from last year’s group, making it possible for every 2013 Chase driver to return. With the exception of missing races due to illness or injury, Chase drivers almost always finish in the top-20 the following year.  With an expanded field of 16, we could probably name all those drivers right now, with 90% accuracy.

By expanding the Chase, NASCAR would actually be diluting the field, and ruining their first 26 races. It’s very predictable who gets in the top 16. The top 12 is less predictable, and the top 10 is even less so. If the Chase only included 10 drivers, you could see half the field change yearly. That would enhance the value of the first part of the season. By expanding to a 16-driver playoff, the value of the first 26 races would be diminished significantly.

Thursday, January 16, 2014

Forecasting Austin Dillon’s Rookie Season

With the new season fast approaching, one of the major story lines will be the return of the famed #3 car. Rookie Austin Dillon will have a giant spotlight on him all year: in addition to driving the #3, he has to fill the shoes of outgoing driver Kevin Harvick. Dillon also has to prove himself worthy of his Cup ride, to silence the criticisms that his grandfather and car owner, Richard Childress, paved a smooth road for him in racing. Dillon is coming off a fresh 2013 Nationwide Series title. He did it without any wins, but through strong consistency: 22 top-10 finishes. In this article, we consider Dillon’s chances in Sprint Cup for 2014 and beyond.

Because Dillon has raced only a handful of Cup races, the simplest way to project his Cup performance is by comparing him to other drivers after a similar level of experience in the Nationwide Series. Dillon has competed in 77 career Nationwide races, allowing us to compare the start of his career with other top drivers. We used this approach last year to accurately project Danica Patrick’s rookie performance.

The chart below plots every finish for drivers in their first 77 Nationwide races (or less than 77 if they haven’t raced that many). Dillon is ranked along with every driver who finished Top-30 in Sprint Cup last year. The blue box represents the majority of finishes for each driver. The red line in the middle is each driver’s median finish. The list is sorted by median finish, with the best drivers at the top.

Dillon’s performance puts him in the top echelon of drivers, with a median finish matching that of Matt Kenseth, Greg Biffle, Carl Edwards, Dale Earnhardt Jr., and Joey Logano. These are all drivers who annually contend for the Chase, a strong sign suggesting Dillon will be a successful Cup driver over the long haul. 

Another way to project Dillon’s performance is by focusing just on his best races, not his average. By only looking at the top 25% of each driver’s finishes, we can highlight pure talent and upside potential. Winning ability is found early in drivers’ careers, and my past research suggests that upside potential is hard to teach: drivers either have it or they don’t. It’s easy to teach a driver to improve their bad finishes as they develop, but it’s harder to improve their good finishes. By comparing only the top 25% of a driver’s finishes, we can focus on who has the most upside potential.

In the chart below, each driver is sorted by the left-edge of the box, representing their top 25% finishes. Again, Dillon finds himself around top-notch company, like Matt Kenseth and Clint Bowyer. Both drivers regularly contend for titles, and it would be fair to expect Dillon to do the same in the next decade.

In summary, the prospects for Dillon’s Cup Career look good. He should be at least as good as last year’s rookie of the year, Ricky Stenhouse, Jr., who finished 19th in points. We should expect Dillon to do better than 19th in 2014 - in fact, Dillon could reasonably earn a top-15 finish every season. If the data above continues to have accuracy in predicting Cup performance, he might even make the Chase this year as a rookie.