Mythbuster #2: “Defense wins championships.”

This year’s Super Bowl appeared to reconfirm the old axiom that defense trumps offense in championship games—at least on the gridiron. A few readers recently asked if this applied to the NCAA tourney and I promised to do a mythbuster on it.

I’ve done this analysis before in a more basic way, but I figured I’d look at the question this time from a few different angles. This first angle compares the performance of the top 10 defensively efficient and offensively efficient teams for the last decade (that’s as far back as my KenPom stats go).

The 100 most defensively efficient teams since 2004 have reached the Final Four 19 times, won seven championships and own a +.091 PASE. Those numbers are slightly better than they are for the 100 most offensively efficient teams, which have one fewer Final Four contender, one less champion and a PASE barely above expectations (+.007 PASE). Score one for defense.

If you’re the kind that likes to look at raw scoring output, defense continues to post better PASE numbers—but notches fewer championships. The top 104 lowest scored upon 1-6 seeds (allowing fewer than 63.1 points per game) have 19 Final Fours, five championships and a +.096 PASE. Meanwhile the 103 highest scoring 1-6 seeds have the same number of Final Fours, but three more tourney crowns, with a PASE of +.028. For this analysis, I’d say the results are mixed.

The next angle I took on the offense/defense debate was to track the efficiency numbers of tournament advancers. The average tourney team comes into the dance scoring 111.1 and allowing 93.4 points per 100 possessions. With each succeeding round, the survivors own better efficiency numbers on both offense and defense, but the offensive numbers improve at a slightly higher rate. Check out this chart:


From the first round to the Final Four, offensive efficiency climbs at a slightly higher rate than defensive efficiency declines—about 1.6% per round versus 1.4%. But then the last two combats and the ultimate champion show more dramatic improvements in offense over defense. The two finals combatants are 1.1% more offensively efficient than Final Four contenders, yet only 0.8% better on defense. And the difference between the last two teams and the ultimate champion is even wider. Champs have 1.7% better offenses and only 0.7% better defenses than Final Four survivors. So if you’re just evaluating which side of the ball is more important in the championship game, these numbers lean toward the offense.

The last angle I took to assess the “defense wins championship” axiom is perhaps the most telling. I examined the value of offensive and defensively efficiency as well as raw scoring in predicting the outcome of head-to-head toss-up games. These are the match-ups in which opponents are within three seed positions of one another. Here’s what I found:

  • Teams with better offensive efficiency numbers are 125-92 (.576) and 7-1 in championship games.
  • Teams with better defensive efficiency numbers are 105-108 (.493) and just 3-5 in championship games.
  • The higher scoring team in toss-up games is 343-268 overall (.561) and 16-6 in championship games.
  • The lower scored-upon team in toss-up games is 297-316 (.485) and 8-14 in championship games.

By every head-to-head comparison, offense is a better predictor of close-seed tourney match-ups than defense. And that’s particularly true in the ultimate final game.

Verdict: There’s a reasonable case to be made that offense trumps defense in the NCAA tourney—at least when it comes to crowning the champion. The numbers say that you need to play solid defense. But that will only take you so far. When it comes to the big prize, you better be proficient and prolific at scoring.

Of course, the best teams in the dance are strong on both ends of the floor. For each of the analytical approaches above, the teams possessing high numbers in both offense and defense categories are far and away the top achievers:

  • The 22 teams with top 10 offense and defense efficiency rankings have been to 10 Final Fours, won five championships and posted a whopping +.593 PASE
  • The 18 teams among the highest scoring (>75.9 ppg) and lowest scored-upon (<63.1 ppg) have been to six Final Fours, cut down four nets and posted a +.449 PASE
  • Teams with higher overall Pythag values are 130-86 (.602) in toss-up games and 7-1 in championship games.

Right now, two teams closest to having top 10 offenses and defenses are Louisville and Florida. Are they the 2014 tourney favorites?

This entry was posted in Key Factors, Mythbuster, Tourney Trends. Bookmark the permalink.

One Response to Mythbuster #2: “Defense wins championships.”

  1. larry k says:

    Hi Peter et al,
    After reading this post, it got me off on a tangent thinking about issues more related to ‘Game Theory’ or possible strategies to use when filling out our brackets. After all, i think the vast majority of us seek to use the data formulated by Peter (and others) to successfully construct, and profit from, our brackets. I saw no other place on the site to construct such a thread, so i decided to just comment on the latest post and hope that suffices…..
    Since most of Peter’s data is team data, focused on attributes compiled to aid us in our actual game matchups, i thought i’d branch out a bit and investigate such game theory topics as:
    – given what we know about a particular tourney (2014 for example), what would be close to the optimal number of ‘bracket entries’ (percentage wise) to submit for any given pool, to optimize our chances to profit or to outright win? what would be the optimal breakdown of how many seeds in any particular seed group should we advance? and how far?
    i’d like to discuss these topics generically at first, meaning irrespective of the teams that actually occupy the seeds.
    The questions composed are indeed complex ones, and by no means do i intend the logical answers to simply arise from this one post, but would like to encourage widespread discussion on them with the hopes that we call can learn something to help us.
    To start, i thought i’d analyze the data i have from two pools that i currently enter, and have entered over the past 8 years. unfortunately, i do not have the data for them going back further because i was not entered prior to 2006.
    i have only analyzed the data from one of the pools to this point, because i do not have direct access to the other pool without intervention by the pool organizer, who has yet to get back to me.
    So, here is what i have found from the first pool i analyzed.

    Pool dynamics…. Consistently between 50 and 59 ‘brackets’ submitted each year from 2006 to present. Of those number, the total number of individuals submitting the ‘sheets’ was between 40 and 49, obviously due to the fact of certain entrants, as i did myself, submitting multiple entries during any particular year.
    Scoring.. Like most bracket pools, points awarded per win increase each round. the pool creator uses a modified Fibonacci ™ sequence, .i.e. 1,2,3,5,8,14. So as you can guess, the game from the elite eight on are more valuable, as with standard pools… there are no ‘bonuses’ for taking a lower seed. Tiebraker is total number of wins for the entrant. Entry fee, because of uniformity as it relates to percentage of entire pool, is irrelevant. there is no vig or takeout to damage the pools possible profitability.


    8 years of data examined.
    430 Total entries
    366 of those entries had a NUMBER ONE seed selected as the ultimate NCCA champion
    40 of those entires had a number TWO seed selected as the ultimate NCCA champion
    24 of those had a seed 3 or lower selected as the ultimate NCAA champ

    6 of the 8 years, the actual NCAA champion was a number ONE seed
    NONE of the 8 years had a number 2 seed as a champ

    Translated: 85.1 percent of ALL entries submitted had a number ONE seed selected as the ultimate champ
    9 percent had a number TWO selected as the champ

    This, in itself surprised me… because the sample is small and only comprised of ONE pool’s entrants, i am not sure what to make of it.. what i do know is that it seems skewed… but then again, when i look at the fact that 75% of the time (in the last 8 years), a number one seed DID win the championship, maybe it goes more in line with the thinking of Peter’s YourMom pool….

    The first idea this left me with is, if this year’s potential top seeds are almost negligibly better than their 2 and 3 seed counterparts, why on earth would i ever select a ONE seed as champion??

    i would be very interested to hear anyone’s thoughts on these findings, and even moreover, if you believe similar numbers exist in the pools’ you actually play in…

    i’ll do my best to dig up the numbers for the other pool i play in within a few days… that might shed a bit more light, because there are typically 150+ entrants in that pool… thanks to a Louisville-Michigan final last year, and much much help from Peter’s numbers, i was able to take down 2nd place in that pool last year… and not to be a commercial for peter’s work, but since joining the membership here, i have cashed 2 years in a row, as opposed to being out of the money for the 4 years prior. Kudos Peter.

    Back to the NFLcombine,
    Larry K

Leave a Reply

Your email address will not be published. Required fields are marked *

* Copy This Password *

* Type Or Paste Password Here *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>