You're looking for an edge for your NCAA Basketball picks. The markets are incredibly efficient, especially as we get into mid-February.
You know numbers can help, and you're all over statistics like efficiency, or points per possession.
However, efficiency can sometimes be deceiving. Let me explain.
The Randomness Of 3 Point Defense
The three point shot plays a critical role in basketball. A team like Michigan can get hot and beat almost anyone despite their deficiencies on defense.
However, it's notoriously difficult to predict three pointers.
In an ideal world, we could look at a team's 3 point defense (field goal percentage allowed) at this point in the season and know how they will perform going forward. A simple statistical test can confirm whether this works. Take a team's three point defense during the first part of the season and look at how it correlates with the remainder of the season.
Ken Pomeroy did a study with 3 point field goal rates. He found almost zero correlation between early and late season statistics (my own studies confirm this). This suggests that a team's 3 point defense right now in the season has no bearing on the future.
This analysis suggests you can throw 3 point field goal defense out the window in making prediction. But this is a problem, since 3 point field goal defense has an enormous impact defensive efficiency. Do you throw out those numbers as well before placing your NCAA Basketball picks?
Wait A Minute...
Before we throw out the bedrock statistics of the basketball analytics revolution, let's get away from the numbers and think about basketball.
Some teams play better three point defense than others. If a team has tall, athletic wing players that pay attention, they will most likely defend three point shots better than a team of short perimeter players who only seem interested in driving to the basket.
Don't let the numbers convince you three point defense is not a skill. The numbers also show a lack of correlation between 3 point offense early and late season, and shooting is certainly a skill. I'll take Steph Curry over Dwight Howard in a three point shooting contest any day.
The better conclusion is that randomness plays a huge role in 3 point defense. This implies that teams that have allowed an insanely poor 3 point field goal percentage on defense should expect some regression towards the mean of 35%. We'll look at some examples, like Notre Dame, later.
Getting Help From Other Statistics
Luckily, analytics isn't completely worthless in looking at 3 point defense.
Two point field goal defense correlates with three point defense. Teams that play better defense inside the arc also tend to allow a smaller field goal percentage from 3, which makes sense.
This suggests a simple method for determining which teams have had bad luck in 3 point defense. Look for teams that have strong defense numbers inside the arc (two point field goal defense adjusted for schedule by my algorithm but poor numbers from outside the arc (raw three point field goal percentage on defense).
These teams have gotten unlucky in that other teams have hit a higher fraction of 3 point shots than expected. Going forward, you expect their 3 point defense to get better.
For these unlucky teams, this implies that they will allow fewer points than expected by efficiency numbers, suggesting shading the under. It also suggests to shade the side in favor of this team.
Here are the top 25 teams that have been unlucky with 3 point defense.
3. Notre Dame
4. Western Illinois
6. North Carolina State
7. Southeastern Louisiana
8. Cal State Northridge
9. North Carolina
11. South Carolina
14. Penn State
15. Appalachian State
16. Arkansas State
18. Wichita State
20. South Dakota
22. Bowling Green
23. San Francisco
24. Ohio State
It's important to know the personnel on a team to know whether regression in 3 point defense is likely. Let's look at one example in particular.
The Fighting Irish gave unbeaten Kentucky all they could handle in last year's Elite Eight. However, they fell short and then lost Jerian Grant and Pat Connaughton from last year's team.
Their two point field goal defense hasn't changed much from last season. For a team that seems more interested in shooting 3's than playing defense, Notre Dame ranked 79th last season in two point defense adjusted for strength of schedule. This rank has risen to 41st this season.
However, Notre Dame's three point defense has fallen off drastically. Last year, they allowed opponents to shoot 32.5%, 79th best in the nation. This season, opponents are hitting 39.3%, 338th out of 351 Division I college basketball teams.
Could the loss of Grant and Connaughton affect this three point defense? Maybe, as both were tall perimeter players. However, the numbers suggest some regression in Notre Dame's three point defense.
As another example, consider the Golden Bears, 2nd on the list of unluckiest teams by 3 point defense. Before last night's game, they ranked first in my two point defense adjusted for strength of schedule. However, California has allowed 36% on opposing 3's, 241st in the nation.
My NCAA Basketball predictions, which do not consider luck in 3 point defense, projected a toss up game between California and Oregon last night. The markets, which account for this randomness in 3 point defense to some degree, closed at California -1.5 NCAA Basketball odds board. The extreme 3 point defense numbers for California suggest not betting Oregon because of my points based prediction.
Last night, Oregon shot 28% from three (5-18) and got blown off the court by 20 points.
Teams can also get lucky in 3 point defense. They allow a much better 3 point percentage than their poor two point defense suggests. This implies shading against these teams on the side and the over.
These are the top 25 lucky teams by 3 point defense.
1. Mississippi Valley State
2. North Carolina Asheville
3. Prairie View A&M
5. Alabama A&M
9. Jacksonville State
10. James Madison
13. Southeast Missouri State
14. Kansas State
16. St. Francis (PA)
17. North Carolina A&T
19. Morgan State
20. Delaware State
21. Southern Illinois
23. Grambling State
25. St. Bonaventure
Ed Feng: @thepowerrank