One of the biggest problems for an undersized Seattle University team this season has been their offensive rebounding. While it's something that might be obvious to anyone who's watched the team, statistics help us to see that if we cannot see the team. Photo via jlindstr.smugmug.com
"As the economist Paul Krugman has noted, anyone who has seen how economic statistics are constructed knows that they are a sub-genre of science fiction. This science fiction has been steadily employed to camouflage our economic decline." - journalist Chris Hedges in a talk at the New School in New York.
Whether applied to economics, education, health care, or basketball, statistics have as much capacity to illuminate hidden insights as they do to take us to fanciful worlds as enchanting as James Cameron’s visually stunning rendering of Pandora.
The risk hardly makes using statistics useless.
In women’s college basketball in particular, given that it’s impossible to see every single game, statistics hold the capacity to provide us with insights about what’s going on across the country that we cannot otherwise obtain. The challenge is to find ways of "constructing" statistical narratives that tell a more complete story than the standard box score without drifting too far into a science fiction reality. It’s satisficing, but even in doing so, it’s better than the alternative of looking at point differentials or box scores alone especially when the level of competition can vary so greatly.
Among the greatest insights that statistics can provide us at the team level is the ability to compare one team’s performance to another using standards that don’t eliminate the chaotic beauty of basketball, but make it easier to meaningfully process and digest.
Q&A: StatSheet's Robbie Allen - Blogger So Dear
Traditional boxscore stats are rarely the best stats to look at when analyzing a game. In basketball, tempo-free stats help to compare apples to apples by leveling the playing field from a possession standpoint. Some teams get more possessions than others so naturally they’ll have more field goal attempts, points, etc. Tempo free stats allow you can compare a team that runs the Princeton offense with an fast-paced team.
At this point, there are enough validated player and team ratings out there that there is little need to create new systems until we gain a better grasp of exactly what the existing metrics do and do not tell us. Accepting that a college player rating system is challenging because of the wide variation in strength of competition and overall schedule, I’ve focused instead on identifying ways that we can describe team performance.
Over the last few days, I’ve described a few of the statistics that I have found most useful in my effort to better understand women’s basketball: Model-Estimated Value as a means to measure the quality of a team’s performance in terms of the collective value of its players contributions and synergy as a means to describe a team’s style on a continuum based on ball movement.
Having looked at the ways to describe the style and value of a team’s performance, today I’m going to move on to describing the substance of a team’s strengths and weaknesses that contribute to success (or failure), defensively and offensively. Furthermore, once a team’s strengths and weaknesses are established, it’s also possible to look at each individual’s contribution to those strengths and weaknesses. That not only helps us evaluate player performance, but also look at the relative importance of specific player contributions to team success as described by the Four Factors.
As with the previous analyses, it’s not a deterministic explanation, but a way to describe the character of a team in more nuanced ways than describing final outcomes.
Describing Four Factors: Team Factor ratings
One of the major advances in statistics is the use of possession data to approximate how well a team performs on a micro level during the course of the game. While there are multiple ways to look at possession level production and efficiency, Dean Oliver’s Four Factors give us a level of detail – factors that determine success – that most other means of analysis don’t.
The Four Factors to Winning at Basketball - StatSheet.com ChangeLog
Dean has identified four factors that are the most important determinants of basketball success. They are:
1. Shooting the Ball Well, which is measured by effective field goal percentage (eFG%). eFG% is like field goal percentage except that it gives 50% more credit for made three-pointers (since it accounts for more points). The calculation is (0.5*3PTM + FGM) / FGA.
2. Taking Care of the Ball, which is measured by turnover percentage (TO%). TO% is a pace-independent way to measure ball security. TO% = Turnovers / Possessions.
3. Offensive Rebounding indicates a team's ability to get second chance shots, which dramatically improves efficiency. This is measured by offensive rebounding percentage (OR%). OR% = Offensive Rebounds / (Offensive Rebounds + Opponent Defensive Rebounds).
4. Getting to the Free Throw Line is measured by Free Throw Rate (FT Rate). This isn't just a measure of how many free throws a team makes, but the frequency in which they go to the line. FT Rate = Free Throws Attempted / Field Goals Attempted.
Basketball Reference has a simple description of the formulas that is worth looking at if you’re not already familiar with Four Factors. Essentially, each factor can be thought of in terms of possessions: offensive rebounding as extending possessions, turnover percentage as maintaining possessions, and effective field goal percentage and free throw rate the ability to capitalize on possessions by scoring points. When you think of measuring a team’s level of "continuity", the ability to manage possessions in these ways – and disrupt opponent possessions – is essential.
While looking at each factor individually can be helpful to analyzing games or teams, putting the factors into a linear weights formula actually explains team success as well.
Oliver’s recommended multipliers described below have been developed for NBA basketball, but work well enough for women’s basketball to approximate team success:
Calculating those numbers for a team and their opponent will pretty reliably explain victories in offensive and defensive terms – the strength of a team’s defense determined by their opponents team factor rating.
However, what I’m more interested in is the character of a team – their strengths and weaknesses not only relative to other teams, but relative to their own success. One way to approximate that is by determining the weighted differentials between a team’s offensive and "defensive" performance.
Team factor ratings case study: Seattle University
SeattleU is an undersized team that is transitioning into Division I. As such, it should come as no surprise that one of their biggest deficiencies relative to their opponents is offensive rebounding. Here are there numbers (along with other numbers I’ve detailed in previous posts):
If we take the differentials between each team factor and apply the multiplier to them, we get a sense of which are the biggest contributors to success or failure.
Based on weighted differentials, their rebounding is by far the biggest problem.
However, something else we’re able to do with this team factors breakdown is begin to assign player credit to team performance in each area.
How can we determine individual contributions to specific team strengths and weaknesses?
The simplest way to approximate individual contributions to winning is to determine a minutes weighted contribution that each player makes to each team factor. So in other words, the production of players who play more minutes is weighted more heavily in terms of how much they contribute to each factor. This is team relative, not league relative, so even a bad offensive rebounding team like SeattleU will have top contributors. But if we accept the team factors rating as a valid measure of team success, then we’re essentially determining how specifically across the factors each player contributes to the team’s overall success.
Here are SeattleU’s numbers through 15 games (to see their season stats, click here):
Ashley Brown, who has the biggest contribution to the team’s eFg%, is a good example to illustrate how this works. Brown does not have the best eFg% on the team at 49%. However, weighted by minutes, her shooting percentage – and especially her three point shooting – contribute more to the team’s overall eFg% than other players who have higher percentages, but play less minutes (Johnson, Nevi, Kerfoot shooting 55%, 57%, and 50% respectively). In other words, it’s not just identifying who the best shooter is, but whose shooting in the minutes given is most important.
In effect, this strategy of dividing player credit helps to strike a balance between the usage and efficiency debate: a poor volume shooter who plays big minutes is not contributing as much to the team’s overall eFg% as a player who is more efficient in slightly less minutes. Conversely, a player who is efficient in reserve minutes is not as big a contributor to eFg%.
This is a descriptive approach to player credit – it describes the relative contribution of a player’s production in a specific area to team success rather than explaining overall importance to the team. For dissecting game and season data, it helps to identify key players beyond the standard points, rebounds, and assists per game. However, it’s not the best way for determining individual contributions to success and I’ll get to those tomorrow.
Obviously, there is no magic formula to winning – two equally successful teams could have different balances of factors
Ultimately, the use of team factors is a way to describe what makes a team successful than a means to predict success – things such as matchups, game strategy, and other quirks that occur during the course of a basketball game can also influence team success.
Moreover, it’s also important to remember that as much a team can benefit by maximizing their strengths, a game is sometimes won or lost simply by one team putting it all together while their opponent flounders. It’s something that University of Washington coach Tia Jackson commented on earlier this season:
Washington Still Focusing On Themselves With DeHaan, Michigan State Looming - Swish Appeal
"We’re learning still," said coach Tia Jackson. "I think right now we’re a team that’s put a defensive game together, we’ve put a rebounding game together, we’ve put an offensive game together, and now we’re trying to put it all in one game. Using a Regina Rogers is part of that equation and it’s something that we haven’t had before. You know we’ve got veteran players who haven’t played with her in their two years here. So we’re still getting used to it."
A five minute stretch of near perfect play, where things are clicking on all (four) cylinders can sometimes be all a team needs to establish a big enough lead to get the win. Conversely, as has been the case for Washington in a few games this year, erratically doing one thing at a time can lead to a pile of losses.
Nevertheless, what the team factors analysis tells us is general trends of what a team does well over the course of an entire 40 (or 48) minute time period and how those things contribute to winning.
Tomorrow: Applying all this statistical stuff to a Pac-10 preview and assigning individual credit directly to team success
- People using this type of analysis at the NBA level have suggested determining a team’s strengths and weaknesses by looking at their performance relative to league average. While that might work in the NBA, the competition in the NCAA varies so much that establishing reliable "league averages" would be difficult. For that reason, I’ve started by determining a team’s average but will try using league averages for the Pac-10 after a few games into conference play.
- One of the most aggravating uses of statistics to me is in education – the annual yearly progress metric (essentially five percent improvement across each student demographic attending a school in significant numbers) makes a complete mockery of the public education project. If a student in bottom 20th percentile in reading in the state even makes a 20 percentile jump in, say, 10th grade, not only is s/he still below average – thus jeopardizing future educational attainment relative to his peers – but depending on the school he attends, s/he’s likely a better test taker than actual reader…meaning failure in the subject that matters most: the ability to actually participate in society. Statistics in education are great for describing trends, but not for determining whether a child has expanded their ability to participate in our society...