NCAA Basketball: Unprecedented Upsets or Same Old Story?

By Julian Ryan

*Editor’s Note: Written before Week 18 rankings published

Is it really that crazy a college basketball season?

Most college hoops fans would argue that this season has seen an unprecedented amount of upsets. Unlike last year, when Kentucky had clearly established itself by this point as the team to beat come March, no team stands out as the clear favorite. Writers are singling this season out as remarkable, increasingly dramatic and completely unexpected.

Is this notion real or are fans simply recalling the recent dominance of the likes of Kentucky? Using the AP rankings, let us compare the quantity of losses by top five teams this season to those in previous seasons. Considering data back to 2006-07, while this is perhaps an above-average year for top team losses, it is nothing extraordinary.


When compared to the average for the six seasons from 2006-2012, this season’s losses by top five teams gives us a p-value of 0.415, far from statistically significant. This means that we do not have enough evidence to conclude that there is a meaningful difference between the top 5 teams this year and the top 5 teams in previous years in terms of losses while ranked. Even when only looking at losses from week 11 on, when a lot of top teams started dropping off this year, the p-value is still a paltry 0.614, even further away from the typical cutoff of .05 used to determine statistical significance. It is certainly not a completely crazy season by the measure of upsets of the top 5 teams.

An alternative explanation for the public perception of 2013 as “crazy” would be the number of upsets of the number one ranked team. Indiana, who have held on to the top spot from week 15 to the next week were the first team to do so since Duke in Week 9. The five straight losses by top teams from week 10 to week 14 this season is the longest streak in our time period. The next longest were streaks of three from Week 14 to 16 in 2010-11 and from Week 10 to 12 in 2008-09. The inability of the #1 team to hold on to their ranking this year coupled with the media attention bestowed upon that team each week is more than likely enough to convince the average fan of this season’s peculiarity.

However, even considering the number one teams alone, this season does not appear to be abnormal. Of the sixteen number one teams which faced competition during the week of their top ranking, six ended up losing a game that week (including the run of five). Again this is above average, but not extraordinary. From 2006 to 2012, the average number of weeks where a number one team lost was four, and there were six losses by number ones back in the 2008-09 season.

Screen shot 2013-03-06 at 2.53.19 AM
While higher than average, this year’s amount of turnover at the top spot is not far out of line with results over the last seven seasons. If we define a season as unpredictable by the amount of top teams that suffer upsets, as it seems many people do, this year doesn’t look that out of place in comparison to recent seasons. As fans, we tend to have short memories, but it is important to remember that a certain amount of upsets are to be expected in any given year. Even though it seems “crazy” that so many top teams are losing, we are not really seeing anything that new.

About the author


View all posts


  • Just a note that you should probably re-title your graph; if I’m understanding it right it’s more like Total Losses Over Time. Losses by Week makes it look like the Top 5 is losing progressively more games each week – up to numbers which are actually impossible; that is, it’s not clear you’re trying to show total losses accumulated over the entire season to date.

    Fine piece otherwise; I would note that one thing I’ve seen argued in most articles is that the top teams are losing to unranked teams more frequently than in the past, rather than just to lower-ranked teams, etc. Would be interesting to calculate average ratings gap of an upset and see if that might be driving some of the “craziness” claims.

  • Your graph has made it plainly obvious why people are seeing this year as extraordinarily tumultuous, because it actually has been in the second half, which also happens to be right around the time that more people start paying attention to the game because of the end of Football season.

    I’d love to see this sort of study done again, but starting at week 10.

  • I think it seems crazier because more losses are coming vs unranked teams… I think it’s 20 now by top-5 teams after tonight.

  • I agree with Nick. It is not extraordinary for a top 5 team to lose. However a top 5 team losing to an unranked team should be rare. I believe that this season had 22 such upsets. It would be interesting to see how this compares with previous years.

  • Please explain what is meant by “Cumulative Top 5 Losses by Week” means, As Collin and Nick pointed out, for what we would normally assume this means it is just not possible for the plotted values. If it is a cumulative measure of how many teams lost while ranked in the top 5, then there is no way the lines can ever be decreasing. If it is the count of losses in that given week, then it is really odd that the lines are generally increasing; and the average number for week 17 implies that we’re in the neighborhood of 3 losses in that week per top 5 team. In week 17 almost every team is playing only 2 (and sometimes only 1) game during the conference schedule, so it’s just not possible.

    Alternatively it is a sum of the losses that the teams currently ranked in the Top 5 have accumulated, in that case, the data doesn’t really assess the hypothesis you’re trying to test as the data would be telling us about the overall record of the teams ranked in the top 5, to me that is the only data set that seems to make sense for the patterns observed in the graph.

    To me it appears you are either using incorrect data for the hypothesis you intended to test, or your data is inaccurate for the reasons stated in the first paragraph. If I am incorrect in my assessment or you are plotting some other data, please clarify.

  • Pingback: My Homepage

Leave a Reply

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