The Hidden “Saturday Advantage” Of Ivy Basketball: A Study On Courts And Dates
Written by Ross Chapman
Home Court Advantage has become an ever more contentious topic for the Ivy League since 2016’s announcement of an Ivy League Basketball Tournament. The men’s and women’s tournaments, which take place at Penn’s Palestra, give a marked advantage to Penn teams as compared to competitions on a neutral court or competitions on another Ivy court. As the League continues to consider the future of the tournament, the impact of Home Court Advantage (HCA) must be researched. In addition, HCA plays an important role in the regular season, especially when teams lower in the Ivy standings play against each other. Finally, the Ivy League is unique in its weekend scheduling for the conference slate. It has been speculated that the back-to-back games and long trips negatively affect players. Is there a “Saturday advantage” not currently accounted for in Ivy League predictions?
Bearing this all in mind, I want to examine how HCA, and how the Ivy weekend schedule, affects the Ivy League at large and Columbia in particular. This article looks at all 560 Men’s Basketball games played between Ivy League opponents from 2009 to 2018. Since every team plays every other team twice every season, evaluating Ivy League season games controls for variables like strength of schedule. For each game, we marked the home and away teams, the score and victor, the relative strengths of the teams as measured by KenPom rankings, and the day of the week on which games were played. We originally looked at only four seasons of data, but had sample sizes too small to make certain conclusions.This study does not examine Women’s Basketball games, non-conference games, or playoff games. Data was gathered from composite schedules from The Ivy League and ESPN.
Data was transcribed by hand and checked for errors. We then split games into two major categories: Day-Before Games, which occurred on most Saturdays, and Non-Day-Before Games, which occurred on Tuesdays, Fridays, and on Saturdays at the beginning of each season. Presumably, teams travelling and playing Day-Before games would be significantly more fatigued due to travel. These games are abbreviated as DB games and NDB games. Within those two categories, as well as in a combined Overall group, we compiled the record of each team, as well as its Home and Away splits. Using the DB and NDB game data, I will introduce a “DB Bonus” or “Saturday Advantage,” which describes the difference in performance of teams when they have to play back-to-back games.
We used additional categories to determine how strong and weak teams perform against each other. We assigned each game a Better KenPom Team and a Worse KenPom Team, based on end-of-season rankings by KenPom. Finally, we created two tiers of teams, a Top Tier and a Low Tier. Top Tier teams, which included all 10-win teams (e.g. 2018 Penn) and two 9-win teams (2014 Columbia and 2018 Princeton), had KenPom rankings of 150 or better. This tiering system reflects my observation that there often one to three top teams in the league who seem to be in very little danger against other, lower tier teams.
If you’d like to follow along with our data, you can check out the first few sheets of this spreadsheet. We identified several major findings in the data:
1. Home Court Advantage matters a lot, especially for Columbia. While sports statisticians often take home court advantage for granted, it ought to be hammered in for those of you not interested in spreadsheets of hundreds of game scores. In the Ivy League, the home team wins 41.3% more games than the away team. The average difference between the home score and away score is 2.05 points, which means that playing at home versus playing on the road creates a 4.1 point swing for an Ivy team. This aligns with the maxim that HCA is worth “about four points” across NCAA basketball. The home team wins 58.6% of the games it plays, which ranks a little bit under the 2016 NCAA-wide average of 60.6%, but still represents a significant advantage.
Columbia has the largest HCA in the Ivy League. Over the last 10 years, Columbia has won 95% more games in Levien than it has on the road. Columbia’s home-away point swing is 5.3 – over a point stronger than the league average. This is less a function of domination at Levien and more a function of severe weakness on the road. The Lions average only 2 road wins per season and 5 road losses. Columbia depends more on the home court than any other team in the league. The Lions cannot afford to drop home games if they aim to compete for the league title.
2. Home teams perform better on Saturdays. (“Saturdays” here refers to Day-Before games, although not every Saturday game is a DB game.) This point touches upon what I believe to be new research in Ivy basketball. During the regular season, teams will play a game on Friday night, drive one to five hours to their next location, and play another game on Saturday. This is viewed as an arduous process, and one that tires out players, especially after they play a strong opponent. The data suggest that this claim holds water. Home teams win 57.4% of their NBD games (Fridays and early-season Saturdays), but they win 60.3% of their BD games (Saturdays during the main season). This 5% increase represents a swing of about 1 game per season across all league games.
The Day-Before distinction is also apparent in home team score margins. As I mentioned earlier, the league-wide home team average margin is 2.05 points. However, this value balloons to 2.92 points during day-before games and falls to only 1.43 points during non-DB games. Two teams who play each other during two Saturdays, once at home and once at the road, will therefore experience a nearly 6 point swing based solely on home court and date. Predicted Score Differentials used by analysts like the Yale Sports Group are often within a 1-point window. I believe that the addition of day-before data may better inform predictive systems moving forward.
3. Worse home teams perform even better in DB games. Teams worse than their opponents win only 29.4% of their NDB games, but win a whole 40% of their DB games. If you want to beat a team who’s better than you, you should (in general) aim to face them at home on a Saturday. Lower-ranked teams experience a 2.45 point DB boost at home, but higher-ranked teams gain a bonus of only 1.15 points. The underdog gets more benefit from the Home DB bonus, the “Saturday Advantage,” then does the favorite.
Individual teams show their own flavors and values of DB bonuses. However, with sample sizes as low as 28 for Home DB games, the findings are hardly conclusive. For what it’s worth, Columbia exhibits a strong penalty when playing Saturday road games. This sounds right – the Lions are undisciplined on the road, and travel fatigue may play a significant part in their woes.
4. Dartmouth is weird. I just wanted to get that out there. The Big Green exhibited, by far, the strongest Away DB penalty and the strongest Home DB bonus. In NDB home games, Dartmouth is a measly 9-31. In DB home games however, when opposing teams are coming off of a game against perennial powerhouse Harvard, Dartmouth totally flips the script and goes 16-14. Of all of the splits I measured, Dartmouth’s only positive record occurred in DB home games. Just remember, folks – when you travel to Harvard and then Dartmouth, you’re likely to lose to the Big Green.
So there you go! I invite other members of the Ivy League community to use this data and see what other conclusions can be reached, and to incorporate it into their predictive models. Any additional questions can be sent to firstname.lastname@example.org. As a disclaimer, many of these assertions do not have large enough sample sizes or large enough effects to be significant at the p<0.05 level. However, I believe them to still be relevant to commenting on the Ivy League basketball scene. Thanks go to Sports Editor Abby Rubel, Cameron Danesh, and Gloriana Lopez for their help with data entry.
Very Friendly Confines image via Columbia University Athletics/Mike McLaughlin