The Path To Remaining Unbeaten In The NCAA This Year

Like most basketball fans, I enjoy close games. A contest featuring teams separated by one or two possessions late makes for enthralling theater. At the end of a tight game, every good decision and crucial mistake is amplified. Even the crowd feels like it wields more influence with time winding down. Part of the reason these games and moments appeal to us is because the stakes are high. Near the end of games we can see clearly how each possession has a significant impact on the outcome of the game. Traditionally, we can identify these types of contests by looking at things like amount of lead changes in the game, margin of victory, or largest leads held. How do we identify other places during the game where stakes were high, though, even when the score doesn't indicate as such?

One way to look at stakes at a given point in a game is by examining the win probability graphs on kenpom.com, noting the amount of leverage on each possession.

From Ken Pomeroy's introduction to win probability graphs on his site in May 2010:

[Leverage] measures how much is at stake on a particular possession. The cut-offs for the five categories are fairly arbitrary at this point. You can think of it as a proxy for the watchability of a game at that point.

The colors range from blue, where win probability is largely unaffected by the potential outcome of a possession, to yellow, where the outcome of a possession can have significant impact on the win probability (more precisely, at least a 10% swing between a 2-point possession and zero points). Leverage is not based on what happened during the possession, but is the range of win probability based on what could have happened.

By categorizing possessions within a game by leverage, we can distill the substance of a close game to its most granular level: the range of outcomes of a single possession. It is on these graphs that we are able to see which possessions were most important, even if they did not seem so at the time. 10 percent swings in win probability are momentous and exciting to imagine, especially when they're happening at various, seemingly more tranquil points in the game.

The chart below includes the amount of seconds every remaining unbeaten team has accumulated in each of the five leverage categories. For instance, Arizona has played 4,547 game seconds at high leverage and Syracuse has played 12,247 game seconds at low leverage. The number in parentheses refers to the percentage of total game seconds that have been played at that category of leverage.

               Low                      Medium                     High
Arizona    14740/38.4%   9135/23.8%   4016/10.5%   5962/15.5%   4547/11.8%
Wisconsin  13630/35.5%   7239/18.9%   7116/18.5%   5736/14.9%   4679/12.2%
Syracuse   12242/31.9%   9253/24.1%   6621/17.2%   5153/13.4%   5131/13.4%
Wichita St. 5290/15.6%  10151/29.9%   5877/17.3%   5331/15.7%   7251/21.4%

We can see from the chart that Arizona, Wisconsin, and Syracuse have played significantly fewer high-leverage possessions than has Wichita State. The three major-conference teams all have similar leverage profiles but can be expected to play much more high-leverage basketball as conference play progresses.

Now, I'm not directly addressing strength of schedule in these observations. Arizona could have had slightly fewer high-leverage possessions than Wisconsin in part because the Wildcats have the 59th strongest SOS and the Badgers have the 27th strongest. However, because Ken's win probability system calibrates well for team strength, the SOS is partly baked into the initial and in-game win probabilities.

Furthermore, this analysis also assigns no difference to the low and high-leverage possessions that are included when an undefeated team has trailed this season. It's for this reason that we have to be careful to assume less high-leverage time means that a team has been strictly dominant. They could have fallen behind early to a good team and the low leverage time that results counts the same as if the positions were switched.

Let's take a closer look at the four unbeaten teams through the leverage lens.

Arizona (17-0)

Minimum win probability this season: 8.5%, 72-70 victory over Michigan (trailed 58-50, 7:19 2nd half)
Game with largest percentage of high-leverage game time: 72-66 victory over Duke (1,289 seconds of 2,400 total seconds; 53.7%)

Among undefeated teams, Arizona has thus far had the least amount of high leverage game time. Less than 12 percent of their total game seconds have been at high-leverage. Based on the leverage measurement, the three most riveting games the Wildcats have played this season were against Duke, Drexel, and Michigan, all of which were away from home. The Arizona-Duke game had nearly 20 straight game minutes of high-leverage possessions, meaning again that possible outcomes could have swayed the win probability at least 10 percent on each trip during that time.

Wisconsin (16-0)

Minimum win probability this season: 21.8%, 59-53 victory over Florida (trailed 16-4, 11:37 1st half)
Game with largest percentage of high-leverage game time: 75-71 victory over Iowa (1,182 seconds of 2,400 total seconds; 49.3%)

Wisconsin is the lone team here to have not played a game where there was at least 20 minutes of high-leverage possessions. The Badgers have also impressively not been under 21 percent win probability at any point this season. Considering these two facts, one could argue Bo Ryan's squad has been the most consistently dominant team in the nation this season, all while playing the country's 27th-strongest schedule.

Syracuse (16-0)

Minimum win probability this season: 13.7%, 78-62 victory over Villanova (trailed 25-7, 11:00 1st half)
Game with largest percentage of high-leverage game time: 75-67 victory over Minnesota (1,283 seconds of 2,400 total seconds; 53.5%)

Syracuse is close behind Arizona and Wisconsin in terms of amount of high-leverage game time accumulated. Their matchup this season with the highest leverage was on a neutral floor against Minnesota in November, when they started with an initial win probability of 46.5%. That contest featured over 21 minutes of high-leverage basketball. Because of the incredibly lopsided runs both ways in the Orange's victory over Villanova last month, only seven of the forty minutes in that game registered as high-leverage. Syracuse has played the weakest schedule thus far of the undefeated squads, though that will change as ACC play continues.

Wichita State (17-0)

Minimum win probability this season: 7.1%, 72-69 (OT) victory over Missouri State (trailed 54-35, 11:48 2nd half)
Game with largest percentage of high-leverage game time: 70-65 victory over Saint Louis (1,568 seconds of 2,400 total seconds; 65.3%)

Wichita State appears to be the outlier of the group, as more than 20 percent of its game time has been played at high-leverage. That's not to say the Shockers should have lost by now: it's just they've played in games where outcomes are more frequently in doubt. This also obviously doesn't mean Wichita State won't be the last unbeaten team to lose; because of their remaining schedule Ken's system still gives them by far the best chance of going unblemished (13.7 percent). But it does provide insight into how many tight situations they've been in and gives us reason to believe that come NCAA tournament time they will have the toughest time of winning six in a row.

Note: We do not have win probability graphs for Wichita State's victories over Western Kentucky and Southern Illinois. The Shockers won both games comfortably and thus they are not quite as far behind the other three as the data above shows.

Also read:

On St. John's and Its Talent

Fairness in Quality Wins

Conference Race Simulations, Part I

Nic Reiner is a contributor to KenPom.com, which covers the analytics of college basketball.