How to make accurate football predictions with linear regression

As a smart football fan, you would like to identify overrated college football teams. This is a difficult task, as half of the top 5 teams in the preseason AP poll have made the College Football Playoff the past 4 seasons.

However, analytics has a nifty trick for identifying teams that don’t belong near the top of the preseason polls.

In addition, this trick lets you look at the statistics on any major media site and identify teams playing above their skill level. In a similar fashion, you can find teams that are better than their record.

The trick relies on regression to the mean.

When you hear the word regression, you most likely think of how extreme performance during an earlier period most likely gets closer to average during a later period. It’s difficult to sustain an outlier performance.

This intuitive idea of reversion to the mean is based on linear regression, a simple yet powerful data science method. It powers my preseason college football model that has predicted almost 70% of game winners the past 3 seasons.

The regression model also powers my preseason analysis over on SB Nation.  In the past 3 years, I haven’t been wrong about any of 9 overrated teams (7 correct, 2 pushes).

Linear regression might seem scary, as quants throw around terms like “R squared value,” not the most interesting conversation at cocktail parties.  However, you can understand linear regression through pictures.

Let me explain.

1. The 4 Minute Data Scientist

To understand the basics behind regression, consider a simple question: how does a quantity measured during an earlier period predict the same quantity measured during a later period?

In football, this quantity could measure team strength, the holy grail for computer team rankings. It could also be turnover margin or win percentage in one score games.

Again, consider this question:

How does a quantity in an earlier period predict the same quantity during a later period?

Some quantities persist from the early to later period, which makes a prediction possible. For other quantities, measurements during the earlier period have no relationship to the later period. You might as well guess the mean, which corresponds to our intuitive idea of regression.

To show this in pictures, let’s look at 3 data points from a football example. I plot the quantity during the 2016 season on the x-axis, while the quantity during the 2017 season appears as the y value.

If the quantity during the earlier period were a perfect predictor of the later period, the data points would line up on the diagonal line (the x and y values are equal along the dashed line).  

In this example, the points do not line up along the diagonal line.  There is an error in predicting the 2017 quantity by guessing the 2016 value.  This error is the distance of the vertical line from a data point to the diagonal line.

For the error, it doesn’t matter whether the point lies above or below the line.  It makes sense to multiple the error by itself, or take the square of the error.  This square is always a positive number, and its value is the area of the blue boxes in this next picture.

The area of the blue boxes is the mean squared error.

In the previous example, we looked at the mean squared error for guessing the early period as the perfect predictor of the later period.  Now let’s look at the opposite extreme: the early period has zero predictive ability.  For each data point, the later period is predicted by the mean of all vales in the later period.

This prediction corresponds to a horizontal line with the y value at the mean.  This visual shows the prediction, and the blue boxes correspond to the mean squared error.

The area of these boxes is a visual representation of the variance of the y values of the data points.  Also, this horizontal line with its y value at the mean gives the minimum area of the boxes. You can show that every other choice of horizontal line would give three boxes with a larger total area.

Regression requires finding the line that minimizes the squared error, or the area of the boxes.  This line is called the best fit line, and the next visual shows the best fit line along with the corresponding minimum mean squared error.

In trying to scare off normal people, the quants will thumb up their nose and say things like “the best fit line explains 70% of the variance.”  Even worse, they might call this the “R squared” value.

You can understand this statement through the pictures.  The best fit line explains 70% of the variance means that the total area of the red boxes is 70% less than the original blue boxes of the horizontal line.

In this example, the best fit line causes a significant reduction in the area of the box corresponding to the left data point.  The box gets larger for the middle point (the blue box is obscured by the red box corresponding to the best fit line). But overall, the area of the red boxes are 70% less than the blue boxes.

You can also think about the error the best fit line doesn’t explain.  The area of the red boxes is 30% of the area of the blue boxes. This corresponds to 1 minus the R squared value.

The higher the R squared value, the smaller the red boxes of the best fit line.  The line explains the data very well.

In contrast, the lower the R squared value, the larger the red boxes of the best fit line, which will be more horizontal.  It doesn’t do much better than the horizontal line of the average.

Get my best football predictions

The Power Rank specializes in making accurate football and March Madness predictions.

To get a sample of football predictions usually reserved for paying members of the site, enter your best email and click on “Sign up now!”









 

 

 

 

 

 

 
 

2. Persistence versus regression to the mean

The data above comes from my team ratings in college football.  To develop these numbers, I take margin of victory in games over a season and adjust for strength of schedule through my ranking algorithm.  The rating gives an expected margin of victory against an average team on a neutral site.

For the 2014 through 2016 season, here is how the team rating for the season predicted the next season.

The data hugs the best fit line, and ratings from the previous season explain more than half of the variance in ratings the following season (54.1%).

Compare this to the same plot with turnover margin, or take aways minus give aways.

For turnover margin, the best fit line is flat.  Turnover margin in one season explains 2.6% of the variance in turnover margin the following season.

From these two plots, we can make two statements:

  1. Team strength in college football as measured by adjusted margin of victory tends to persist from year to year.  This will be useful in making a preseason college football model.
  2. Turnover margin regresses to a mean of zero from year to year.  This implies that turnover margin last season has almost zero ability to predict turnover margin this season.

When most people talk about regression, they usually mean the strong type we see in turnover margin.  To see this from a different perspective, let’s consider the wins and losses of Coach Average.  His results come from flipping of a coin with an equal chance for a win and loss.

For this visual, I generated this data once with 8 lines of Python code.  In no way did I search for a sequence with streaks. However, streaks almost always appear in the sequence.

In this simple experiment, the flipping of any one coin has no impact on the outcome of the next coin.  The code makes each game for Coach Average independent of all other games.

Regression to the mean implies that despite a hot 8-2 start for Coach Average, he should still expect to win half of the next 10 games.  In fact, he wins 6 of the next 10 games.

Coach Average also expects to regress to .500 after 9 straight losses starting on game 19.  However, he loses 6 of the next 8 games. 

3. Skill versus luck

Phenomena in the real world are not as simple as this coin flipping experiment, and we need to be cautious in making statements about sports. When a quantity like turnover margin has no ability to predict future turnover margin, it doesn’t imply the lack of skill in preventing or forcing turnovers.

While the analytics community doesn’t have a complete picture of turnovers, a few key insights have started to emerge.  The game situations matters in turnovers. Over the past decade in college football, teams in the lead have committed only 14% of all turnovers.  

This would help explain why Alabama has posted +68 turnover margin the past 8 seasons.  The Crimson Tide failed to have a positive turnover margin only in 2014, when they gave away the ball 2 more times than they took it away.

The dependence of turnovers on game situation makes sense.  Teams in the lead tend to run the ball, especially later in the game.  Turnovers happen at lower rate on running than passing plays.  If a team faces a deficit, they need to throw the ball to get back into the game.

For another example of how regression doesn’t necessarily imply a lack of skill, let’s turn to college basketball.  Ken Pomeroy wondered how much control teams have over three point shots. He asked how a team’s 3 point shooting percentage in the first of the conference season predicted the second half.

The visual shows his results.


The left panel shows 3 point percentage on offense.  The first half of the conference season has almost no ability to predict the same quantity later in the season.  Does that mean shooting is not a skill? Tell that to Steph Curry.

The visual also shows the strong regression for 3 point field goal defense. This suggests a lack of skill in defending the 3.  To confirm this, Pomeroy performed a more detailed study on 3 point defense for teams over a 5 year period.  He concluded there is some skill, but randomness plays a bigger role than anyone expects.

When a quantity regresses to the mean like turnover margin or 3 point shooting percentage, it doesn’t necessarily imply a lack a skill.  Fumbles may regress to the mean, but that doesn’t mean a running back isn’t fumble prone if he palms the ball with one hand while running through the line of scrimmage.

However, it is safe to say that randomness plays a large role in turnovers and 3 point shots.

4. How to make preseason college football predictions

USC had high expectations heading into 2017.  Sam Darnold took over the starting QB job the previous season and led the Trojans to a 9-0 finish, which included a dramatic win over Penn State in the Rose Bowl.

To start the 2017 season, the pollsters put USC 4th in the preseason poll (both AP and Coaches).  This made Clay Helton’s team the favorite to make the College Football Playoff out of the Pac-12 conference.

In contrast, no one knew quite what to expect from Georgia.  Just like USC, 2017 was their second under a young coach, Kirby Smart.  But in contrast to USC, they struggled in 2016.

Georgia started true freshman Jacob Eason at quarterback, who delivered a mediocre 55% completion rate.  They ranked 81st in my adjusted yards per attempt.

Georgia went 8-5 in 2016, a record acceptable only for new coaches in Athens.  To start 2017, they landed at 15th in the preseason AP poll.

So what does regression say about these two teams?  Each year, I put together a preseason college football model that uses regression on many variables.

In college football, team performance tends to persist from year to year. Programs like Alabama have financial resources and traditions that Rice will never have.  These teams will not swap places in the college football hierarchy.

Hence, my preseason model uses the past 4 years of team performance to predict the next season.  This part of the model says that team is most likely to perform as some combination of their last 4 years, with recent years weighted more.  This makes the model cautious about an outlier season or 9 games.

The preseason model also considers turnover margin.  Turnovers can impact the scoreboard, as a key fumble halts a game winning drive, or an interception returned for a touchdown turns a tight game into a laugher.

However, turnover margin regresses to the mean of zero from year to year in college football.  Hence, the model uses turnover margin in each of the past 4 seasons. This holds back the excitement for a team that made a huge jump in rating with a +25 in turnover margin.

Last, the model considers returning starters.  More experience implies better performance for college football teams.

Over the past 3 season (2015-17), this regression model for college football has predicted 69.8% of game winners.  This rate doesn’t include easier to predict cupcake games with FBS teams facing inferior FCS teams.  The model only makes predictions for games with two FBS teams.  

Heading into 2017, the preseason college football model had USC 16th.  In the previous 4 seasons, USC had never finished the season higher than 14th in my college football rankings.  Despite their impressive 9-0 finish in 2016, they only rose to 14th because of a poor start.

The model doesn’t distinguish between returning starters at different positions.  The quarterback has an outsized impact on a football team, and Darnold’s status as a top NFL prospect could convince you of USC as higher than 16th.  However, 4th in the AP poll seemed too optimistic.

In contrast, the regression model agreed with the AP poll on Georgia.  The model had the Bulldogs at 18th while the polls had them at 15th.  Analytics and polls agreed on Georgia as a solid top 25 team but not a playoff contender.

During the actual 2017 season, USC did not live up to their top 4 ranking.  They dropped an early road game at Washington State. Then their playoff hopes ended when Notre Dame stomped their defense for 8.4 yards per carry in a 49-14 win.

Georgia’s season got off to a distressing start as Eason sustained an injury in the first game.  However, this turned into a blessing, as true freshman Jake Fromm took over and had a brilliant season. Georgia’s pass offense finished 10th in my adjusted yards per attempt.

The defense also improved, as they jumped from a solid 28th in my adjusted yards per play in 2016 to an elite 3rd in 2017.  This unit had only one bad game when they allowed 40 points at Auburn. However, they atoned for this blip by holding Auburn to 7 points in the SEC title game win a few weeks later.

Georgia made the College Football Playoff, and they took Alabama to overtime in the championship game before losing.

Here’s the take home message about college football preseason predictions: It’s much easier to predict regression for a team like USC than a sudden rise for Georgia.  

In each of the past 3 seasons, I’ve written about 3 overrated college football teams in the preseason polls on Football Study Hall, an SB Nation site (2015, 2016, 2017).  This analysis combines my regression model with knowledge of programs.

Looking back on these predictions, I’ve been right about 7 of 9 teams.  For the other two teams, Penn State and Oklahoma State finished lower in the final poll than the preseason poll in 2017.  However, I’ll call these predictions a push as they both performed better than I predicted.

There’s a good reason I have never written a corresponding 3 underrated teams article.  A regression model will not identify teams like Georgia in 2017.  Their true freshman quarterback worked out, and their defense made a leap.  

There is a random element to teams that make a sudden rise.  These predictions are more difficult than finding an overrated team by a regression model.

5. How to make accurate predictions with regression

I want you to take home two main points.

First, quantities in football like turnover margin show very little persistence from an earlier period to a later period.  These quantities regress to the mean, as your best guess for the later period is the average.

To find football teams who might not be as strong as their record suggests, look for teams with large, positive turnover margin.  In contrast, teams with large, negative turnover margin might be better than their record. Use these links for data on college football and the NFL.

As another example, consider an NFL team’s record in close games from year to year.  Based on data from the 2012 through 2017 season, the visual shows the one year explains 1.3% of the variance the next year.

A team like Oakland, who went 9-2 in one score games in 2016, should not expect the same good fortune again.  The Raiders went 4-3 in one score games in 2017. However, regression to the mean isn’t a perfect predictor, as Cleveland has a 2-16 record in one score games the past 3 seasons.

Second, some quantities like team strength in college football tends to persist from year to year.  This allows for predictive models based on linear regression.

Even with this persistence, the models still predict regression for outlier performances, both good and poor.  The 9-0 stretch for USC to end 2016 serves an example. However, regression models can not predict teams that jump from ordinary to the outlier, like Georgia in 2017.

These ideas apply for both my preseason regression model at The Power Rank and Bill Connelly’s S&P+ numbers.  Use these rankings as a guide to find overrated teams near the top of the polls.

The secret that no one tells you about in predicting Georgia vs Alabama

Alabama and Georgia meet in the college football championship game, and my member numbers predict Alabama by 4.2. I often make this prediction with a certainty that the most likely result is an Alabama win by 4.

However, posting a single number is deceiving. There’s uncertainty in making a prediction, whether by analytics or the markets, and no one talks about it.

To illustrate this uncertainty, let’s look at Alabama.

Alabama versus Clemson year to year

To close the 2016, Alabama met Clemson in the college football championship game. The markets closed with Alabama as a 6.5 point favorite.

Alabama lost a close contest to Clemson, but that didn’t dampen the enthusiasm for the Crimson Tide heading into 2017. The hype for this dynasty reached a fever pitch when they destroyed Vanderbilt 59-0.

In November of 2017, the markets made Alabama better than even odds to win the college football playoff. Think about this. If you ask yourself whether to take Alabama or any other team to win the playoff, the markets favored Alabama. It didn’t even matter that they had yet to make the 4 team field for the playoff.

Then the Auburn game happened. Alabama lost 26-14 on the road to their arch rival.

This one game changed everything. With the loss, Alabama didn’t win the SEC West. Instead, Auburn won the division and played Georgia in the SEC title game.

Alabama no longer controlled their playoff destiny, and they sat at home sweating it out on championship weekend.

Lucky for the Crimson Tide, two of the top 4 teams lost on championship weekend. The debate about the last playoff spot came down to one loss Alabama that didn’t win their division against two loss Ohio State that won their conference.

The committee picked Alabama, and they were slated to face Clemson again in the playoff semi-final game. The markets closed with Alabama as a 3.5 point favorite.

Remember, Alabama was a 6.5 point favorite just a year before. I personally find this swing in point spread insane.

Alabama improved from 2016 to 2017. By my adjusted yards per play, they had the top defense both seasons. But QB Jalen Hurts developed as a sophomore in 2017. The offense jumped from 12th in 2016 to 8th in 2017 in my adjusted yards per play.

On the contrary, Clemson isn’t the same team as last year. Their defense has arguably improved, but they lost a generational talent in QB Deshaun Watson.

What was the real cause of Alabama going from a 6.5 to 3.5 point favorite? In 2016, Alabama scored 15 non-offensive touchdowns. This made them seem invincible both to the eyes as well as points based analytics.

This preseason, I wrote about how this scoring production from defense and special teams was irreproducible. In 2017, Alabama had 2 non-offensive touchdowns, and one came on an interception against Clemson.

The second factor in this point spread swing? The loss to Auburn. Alabama was undefeated in 2016, but had a blemish in 2017.

The difference in Alabama versus Clemson suggests at least a 3 point uncertainty in making predictions.

The impact of one game

Now, how about Alabama versus Georgia? The day of the game, Alabama is a 3.5 point favorite, with a few sports books at 4. This is close to my member prediction Alabama by 4.2, which suggests no value in this game.

But, if Alabama had beaten Auburn, they would have played Georgia in the SEC championship game. In fact, they would have played on the exact same field in Atlanta at which they will meet for the playoff championship game.

If Alabama had played Georgia in the SEC title game, Matthew Holt of CG Technologies, the guy who sets the line in Vegas, says Alabama would have been a 7.5 to 9 point favorite.

This suggests one game against Auburn shifts the spread by up to 6.5 points.

Auburn was a big data point for Alabama, as the Crimson Tide lost to the best team on their regular season schedule. However, this loss should not imply a shift of 6 or more points in the Georgia prediction.

There is uncertainty in making football predictions.

Podcast: John Ewing on NFL Wildcard Playoff, CFB title game

On this episode of The Football Analytics Show, I talk with John Ewing, data scientist at BetLabs and The Action Network. Among other topics, we discuss:

  • The true reason behind his trends, such as NFL teams coming off a 20 point loss to a good team are 62.1% against the spread
  • The NFL playoff game this weekend in which our predictions straddle the markets
  • Two Super Bowl futures that John likes
  • Whether Philadelphia could be a home underdog next weekend (remember, Carson Wentz is out)
  • The hidden factor that affects the total in the Georgia vs Alabama championship game

I really enjoy all of John’s insight on Twitter. To follow him, click here.

After the interview, I have my own segment on my spread prediction for Alabama vs Georgia. But it’s really a story about the uncertainty in making these predictions.

I use an insight from Matthew Holt of CG Technologies, who tells me what the spread would have been for an SEC championship game between these two teams. Listen at 34:38.

To listen on iTunes, click here.

To listen here, click on the right pointing triangle.

Podcast: How to win your college football bowl pool

On this episode of The Football Analytics Show, I discuss my research on how to win your bowl pool. This includes:

  • The value in entering bowl pools based on public picking behavior
  • The kind of pools that reduces the randomness that prevents you from winning
  • Contrarian strategies work, but only if you get in the right kind of pool

This is the audio version of this article.

I also discuss my market based predictions that adjust for NFL QB injuries. Listen at 14:27.

To lisen on iTunes, click here.

To listen here, click on the right pointing triangle.

How to win your college football bowl pool

You want to win your college football bowl pool. There’s no better way to kick off the New Year than collecting the winner’s money and talking some trash to your friends.

Entering a bowl pool is easy enough, as you pick the winner in each bowl game straight up. Some pools require confidence points that define the points earned for picking the winner in each game while others weight each game equally.

However, bowl games are hard to pick. These games usually feature only strong teams, and the markets pick 63% of the game winners straight up.

This article will take you from novice that doesn’t know the difference between Ole and Southern Miss to expert that optimizes your chance to win a bowl pool.

This starts with recognizing the value in bowl pools, as the data on how the public picks games reveals inefficiencies. Entering a bowl pool isn’t like betting spreads.

Then we’ll get into the type of pool that you should enter. The six digit prize in that thousand person pool might seem enticing, but we need to get real about the effects of randomness.

Finally, I’ll discuss the contrarian strategies you will need to win intermediate sized pools. You will identify the mistakes of the public and pick against them to maximize your chances to win.

However, these contrarian strategies only work for certain kinds of pools. Confidence points or no? A big aspect of winning your pool is identifying the right one. Hopefully, I’ve caught you before you start getting all those bowl emails from your friends.

Let’s dig into my research on how to win your bowl pool.

You can listen to the audio version of this article here.

Is there any value in bowl pools?

If bowl pools were efficient, entrants would pick big favorites at a high rate. For more closely contest games, the teams would get picked closer to 50-50.

We can check the efficiency of bowl pools by digging through the pick distribution on Yahoo over the past 4 years. Millions of people pick bowl games on this site, so it gives a good picture of public behavior.

In the visual below, I show the fraction of the public that picked the favorite as a function of the closing point spread. The data includes all bowl games from 2013 to 2016.

For big favorites, the majority of the public picks these teams, as expected. The line shows the expected win percentage based the point spread, and the public pick rate roughly matches this expectation for spreads greater than 11.

However, the public falters on games with smaller point spreads. The visual shows a wide scatter of points around the market expectation for games with point spreads less than 10.

For example, consider the Outback Bowl at the end of the 2015 season. 8-4 Tennessee played 10-2 Northwestern, and the public favored the ten win Wildcats. 55% of pool entrants picked Northwestern.

However, Tennessee closed as a 9.5 point favorite. The markets had optimism about Butch Jones as he closed out his 3rd year at Tennessee, a short two years before he got fired.

Tennessee destroyed Northwestern 45-6, providing a nice win for anyone who took the markets and analytics more seriously than the public.

As another example, let’s look at a predicted tight game. In 2013, 9-4 Utah State played 12-1 Northern Illinois in the Poinsettia Bowl. While the record favored Northern Illinois, the markets closed with Utah State as a 2 point favorite.

The public seemed to look at record again, as 86% of entrants picked Northern Illinois. Utah State won 21-14.

Later in the article, we’ll see how these discrepancies between pick rate and point spread can be used in a contrarian strategy.

Do you make this mistake in selecting pools?

With the recognition of value in bowl pools, the next step is finding the right pool. While the thousand person pool with a six digit prize pot might seem enticing, I recommend avoiding these contests. To understand why, consider this scenario.

Steph Curry walks into a gym full of basketball players hoping to make an NBA team. Feeling kind of cocky, he challenges the players to a three point shooting contest. Beat Steph, and get an NBA rookie contract.

Will any player beat Curry, the greatest shooter to ever grace a basketball court? It depends on the number of players that compete.

Let assume Steph has a 95% chance to win each contest. It’s unlikely any one player beat him.

If Steph only takes on two players, there’s a (0.95 * 0.95), or 90.3% chance that he wins both contests. He has pretty good odds even against players that have excelled in college.

However, the more players that challenge him, the more likely someone catches him in a cold streak and beats him. For three contests, the probability he wins them all is 0.95 raised to the third power (equivalent to 0.95 * 0.95 * 0.95), or 81.5%.

For 13 contests, Steph has a roughly 50-50 chance to win them all. If he lasts 50 contests, he has a 7.6% chance to win them all. His probability to remain undefeated goes down exponentially, as the math folks like to say.

If you use analytics and market data in your bowl pool, you’re like Steph Curry. You have a very good probability to beat an arbitrary pool entrant picking like the general public according to Yahoo.

However, a problem arises if you get in a pool too large. The visual shows your win probability as a function of pool size in 2016.

I’ve assumed that each game get weighted equally, and you pick the market favorite in each game. Others brackets get picked according to the Yahoo public distribution. Each data point requires 10,000 simulations of bowl season according to the win probability from the markets.

Despite playing a solid entry with all favorites, your odds to win a pool drop drastically with the size of the pool. Don’t get in a large pool.

Let me repeat that: don’t get in a large pool. No analytics or market data can help you defeat the randomness that comes with these large contests.

Contrarian strategies for winning your bowl pool

The last section explained why you don’t want to get in a large pool. Stick with a small pool for the best odds. For example, picking favorites in a 30 person pool gives you a 28.5% chance to win in 2016, a great return on investment.

For intermediate sized pools, you might ask whether you can increase your win probability with contrarian strategies.

For example, consider a March Madness bracket in which the choice of champion gets the most points. In my book How to Win Your NCAA Tournament Pool, I discuss how you make a contrarian choice by picking a champion that has a decent win probability but is not getting picked by others.

For example, Duke was the contrarian choice in 2015. They had a 12% chance to win the tourney by my numbers, but almost no one was picking them as champion in their pools. When they won, you got a large number of points that no one else got. Picking favorites, or chalk, in earlier rounds would earn you enough points to beat the few others that had Duke as champion.

By picking a contrarian champion, you reduce the expected value of points in your bracket. Duke’s 12% win probability was small compared to the 36% of Kentucky. However, you increase the variance in points your bracket produces. If Duke wins, you get a positive bump not many others will have.

The contrarian choice is a lower expected value, higher variance strategy.

In applying this idea to a bowl pool, you need to find a favorite getting picked by the vast majority of the public. The idea is to find a large difference between the public pick percentage and a team’s win probability.

For example, in 2016, Oklahoma had a 57.9% win probability over Auburn but 83% of the public were picking the Sooners. While chalk says to pick Oklahoma, the contrarian choice says take Auburn and their 42.1% win probability.

The visual shows the win probability for the favorites vs contrarian strategies for pool with no confidence points. The favorites curve is the same as in the last section. I’ve added a contrarian entry that picks Auburn instead of the favorite Oklahoma and Northwestern instead of the favorite Pittsburgh.

The contrarian strategy lowers the win probability for this pool in which each game gets weighted equally. What’s going on here?

Let’s think of each game as a coin flip with a weighted coin. In the math jargon, this is a Bernoulli random variable with a parameter p that gives the win probability for the favorite.

The next visual shows the variance for a Bernoulli random variable. For the math inclined folks, the formula is p(1-p).

Notice the symmetry of this variance around the value 0.5. The function reflects like a mirror about this line.

For the Pittsburgh game, p=0.609 as the win probability for favorite Pittsburgh. When we made the contrarian choice, we picked Northwestern with p=0.391. However, the variance in this game outcome didn’t change. Pittsburgh’s win probability 0.609 is 0.109 from the plane of symmetry, just as Northwestern’s win probability 0.391 is also 0.109 from this plane.

In this pool with no confidence points, you are lowering your expected value but not changing the variance. This is why the win probability decreases.

Here’s the take home message: you can’t apply contrarian strategies in pools where all games have the same value, or pools with no confidence points.

For a pool with confidence points, the variance for a game is the number of points times p(1-p). You can change the variance through the confidence points, so the contrarian strategies will work on these pools.

The top 3 tips for college football bowl pools

Here are the 3 main points:

  • There is value in college football bowl pools.
  • Please don’t get in a large pool. Someone will get lucky and beat you.
  • Contrarian strategies don’t work in pools without confidence points. Get in a pool with confidence points.

This advice also applies to weekly NFL pools in which entrants pick the winners of games straight up. To check out the public odds on Yahoo, click here. And remember to know your pool size.

If you need a more in depth information on how to win your bowl pool, The Power Rank offers a guide to bowl season that gives confidence points and suggest contrarian games.

To check out this package, click here.