The NCAA Men's Basketball Tournament is just around the corner and that means it will be time to fill out your brackets. As we all know, the winner of the pool is not always the person who follows college basketball closely. In fact, the winner of the office pool always seems to be a person who hasn't watched a game all year and picks winners based on uniform color and thinks Pepperdine is a children's cough syrup. The bottom line is that anyone can win the NCAA college basketball pool with a little luck and that's what makes it so interesting. The goal of the following analysis is to examine the factors that affect who wins and who loses those key NCAA Tournament matchups. Seeding remains the most important indicator of a team's chances to advance, but there are many other factors that can help determine who will win these games. I have analyzed 14 years worth of tournament results, team records and team statistics in an attempt to find out which factors matter the most. Here are my findings.

Disclaimer

Let me start by saying that this site does not support or encourage gambling. I have provided this information for the amusement of college basketball fans and stat junkies like myself. If this helps you fill out your brackets, that's great, but I would never advise anyone to gamble with money that they cannot afford to lose. The beauty of the NCAA pool is that it is very low risk with a possible high reward and the tournament is much more interesting when you have someone to root for. The following information is based on trends in the tournament over the past 14 years. There is no guarantee that these trends will continue this season. More importantly, even if these trends do hold true, they give you only a slight edge. Most of what happens in the tourney is pure chance and it takes a lot of luck to do well in your NCAA pool. Please feel free to email me and let me know what you think about the results. However, if you complain to me because you used these trends and they did not help you choose winners, I will not listen. Use this information at your own risk.

Rule #1 - Know Your Seeds

The first step in filling out your bracket is to understand the importance of seeding. Unless you are a complete novice to the religion that is March Madness, you know that in general the better the seed, the better the team. There are of course exceptions to this rule. At times the committee loses its mind, but in general the seeding is a fairly accurate representation of the quality of the teams. Here is how the seeds have performed on a round by round basis since 1997.

First Round

The following are the won-loss records of the better seeds in Round One (1997-2010):

• Number one seeds are 56-0 over the past 14 years and a perfect 104-0 since the tournament switched to 64 teams in 1985
• Number two seeds are nearly as strong, winning 96% of first round games since 1997
• The #13 and #14 seeds have had some success in the first round, winning 18 of 112 games (16%)
• Interestingly, #12 seeds have performed better than #11 seeds (37% vs 30%)
• The #7 and #8 seeds are 57-55 against the #9 and #10 seeds since 1997

Second Round

Some interesting patterns also emerged in Round 2.

• #1 seeds have advanced to the Sweet Sixteen 49 of 56 times since 1997 (88%). #1 seeds are 23-1 in the second round over the past six years.
• #2 seeds are 20-9 (69%) against #7 seeds but only 12-13 (48%) against #10 seeds
• #3 seeds are 12-4 (75%) against #11 seeds and 18-16 (53%) against #6 seeds
• #4 seeds are 12-16 (43%) against #5 seeds and 11-5 (69%) against #12 seeds
• #5 seeds are 5-2 (71%) against #13 seeds
• Since 1997, only one #14 seed and three #13 seeds have advanced to the Sweet Sixteen (one of those was a #13 seed that beat a #12)

Third Round

By the time the third round (also known as the Sweet Sixteen) is complete, most of the major upset specials have left the tournament.

• Since 1997, a total of 41 #1 seeds have advanced to the Final Eight (just under three per year)
• #2 seeds account for 23 Elite Eight appearances (1.6 per year)
• #3 seeds were next with 17 appearances (1.2 per year)
• The #4 and #5 seeds each advanced to the Elite Eight five times between 1997 and 2010
• #6 seeds have advanced to the final eight six times in the past 14 years
• Seeds 7, 8 and 10 have each reached the Regional Finals four times since 1995
• Between 1997 and 2010, no #9 seeds reached the Elite Eight
• In 2002, Missouri became the first team seeded 12 or higher to reach the Elite Eight since the tournament expanded to 64 teams in 1985. Teams seeded 12 to 14 are 1-19 in the third round during that time.
• Between 2007 and 2009, 23 of 24 teams reaching the Elite Eight were seeded either 1, 2 or 3 but two #5 seeds and one #6 seed were among the final eight last year.

Final Four

It has been my experience that it is very difficult to win a large office pool without correctly picking at least three of the Final Four participants. This is because most pools allocate a greater amount of points to the later rounds. Although it is tempting to pick several upsets in hopes of stealing some key bracket points, the truth is that it is rare for more than two teams seeded lower than a #2 seed to reach the Final Four. The obvious exception to this rule was 2000 when two #8's and a #5 joined #1 Michigan State. In 2006, #2 UCLA, #3 Florida, #4 LSU and #11 George Mason reached the Final Four. In 2010, two #5 seeds (Butler and Michigan State) were in the Final Four. Since 1997, 25 of the 56 Final Four participants (and 9 of 14 champions) were #1 seeds. The five non #1 seeds to win since 1995 are #3 Florida in 2006, #2 Connecticut in 2004, #3 Syracuse in 2003, #2 Kentucky in 1998 and #4 Arizona in 1997.

Seed Differential

As we have seen, seeding is the greatest predictor in determining who will win tournament games. The #1 seeds have performed significantly better than #2 seeds and #2 seeds have performed much better than seeds 3 through 6. However, these numbers have also shown us that when seed differential (the difference between the seeds of the participants in a given game) is small, the advantage for the team with the better seed diminishes. Here are the breakdowns by seed differential since 1997:

Analytical Dataset

As you can see, when the two teams have a seed differential of two or less, the team with the better seed (ie. the lower seed number) won just 50.3% of the time. This tells me that in matchups between teams with a seed differential of less than three, seed is not the primary factor in determining the winner. In past years, I've focused my analysis on the matchups between teams with seed differentials of 0, 1 or 2. Adding all of the games between teams with a seed differential of 3 would not make sense because the better seed wins 56.2% of the time (and my goal is to analyze games where the two teams have an equal chance of winning the game). When I excluded games between #1 and #4 seeds (#1 seeds were 18-6 against #4 seeds) the team with the seed three slots better than the opponent was only 50-47 (51.5%) so I included these "3 seed differential" games to my latest analysis. So my final analysis dataset includes (a) all games where the seed differential is 2 or less and (b) all games where the seed differential is 3, excluding games between #1 and #4 seeds. This gave me 274 games to analyze where the team with the better seed was 134-130 (50.8%). Ten games involved teams with the same seed.

Road and Neutral Court Record

Another of the major criteria used by the Selection Committee in determining who is invited to the NCAA Tournament is a team's road/neutral court record. The conventional wisdom states that tournament games (which are of course played on neutral courts) are a lot more like road games than home games, so the teams that have played well on the road will, in theory, be better prepared for the tournament. In this analysis, I compared teams with road/neutral winning percentages 20% better or worse than their NCAA Tournament opponents. I started with the analytical dataset described above and eliminated all games in which the difference between each team's road/neutral record was less than 20%. In other words, if Boston College won 60% of their road/neutral games and played Indiana who had won 50% away from home, the game would not be analyzed because the difference is only 10% (60%-50%). If Indiana had won only 40% of their road/neutral games, the game would be part of the analysis. Eliminating games without the 20% gap, I ended up with 74 games where the team with the better road/neutral winning percentage was 41-33 (55.4%).

When I increased the minimum road/neutral winning percentage gap to 30%, the team with the better away-from-home performance was 17-8 (68%). This can be taken two ways: you could argue that the data indicates that road/neutral winning percentage is a good predictor of success because teams with the 30% advantage won more than two-thirds of the time or you could argue that road/neutral is not a good predictor because teams with a 20%-29% advantage were only 24-25. I do feel that this lends some strength to the argument that teams who have shown the ability to win on the road have a better chance to win against equally strong teams that have not performed as well away from home but the results are not as strong as I'd hoped.

Record Against the RPI Top 50

A long suspected barometer for potential success in the tournament has been the team's record against teams in the RPI Top 50. Most of the teams in the RPI Top 50 wind up in the NCAA Tournament so this measure shows how each team has performed against tournament caliber competition. With seed differentials of 3 or less (excluding 1 vs 4 matchups) and the aforementioned 20% gap between the competing teams' winning percentage against the RPI Top 50, the team with the better RPI Top 50 winning percentage was only 40-39. When I eliminated games in which at least one team had played less than three games against the RPI Top 50, the results were a little better for the team with the better winning percentage against the RPI Top 50: 38-34 (52.8%). I do feel that a team's performance against the top RPI teams is a useful measure of potential March Madness success, but one would be advised to examine each team's schedule because many tournament teams are very different in March - for better or worse - than they were in November and December when many of the nation's top teams meet in non-conference battles.

Record in the Last 12 Games

One of the major criteria used by the NCAA Selection Committee to determine who gets a tournament invitation is a team's record over its last twelve games. At one point, the Committee was using a "last ten games" metric so my analysis is actually a mix of "last 10" and "last 12". I analyzed the instances where one team had at least two more wins than their tournament opponent over the last 10 or 12 games. In other words, if Team A won of 10 of 12 games heading into the tournament and Team B came in winning only 8 of 12, there would be a 2 win gap and that matchup would be included in the analysis for this category. I found that the team with the better "last 10/12" record won 55.6% of the time (69-55).

This indicates that a team's record in the last 12 games is a good predictor of tournament success. Not so fast, I altered my analysis to look at matchups between teams with at least a three win gap. I found that the team with the better record down the stretch was only 25-28 (47%). So, when the gap was 2, the team with the better record an outstanding 44-27 (62%) but that advantage disappears when the gap is larger. This is exactly the opposite of what I expected. The results are disappointing but not entirely surprising because a team's record down the stretch is dependent on the strength of their conference. For example, a team that finished 7-3 in the Big East is probably playing better basketball than a team that finished 8-2 in the West Coast Conference.

Experience

It have always felt that another key factor in determining NCAA Tournament success was experience. Teams playing in their first NCAA game often have that "deer in the headlights" look. Meanwhile, experienced teams seem to be able to hold their composure in any situation. Not only are Duke, North Carolina, Kentucky and Michigan State great teams but they have been exposed to so many tourney games that they probably feel little additional pressure. Smaller schools like Gonzaga and Butler have been able to do well, presumably having learned from their initial tournament experiences. I tested this theory by looking at all tournament games that matched one team that played in the previous year's NCAA Tournament and one team that did not. I found that teams in my analytical dataset with tournament experience in the prior year were 66-42 (61.1%) against teams who were not in the tournament in the prior year. This clearly supports the theory that prior year experience is a predictor of tournament success.

Conference Tournament Fatigue

It has always been a theory of mine that teams who play three times in three days in their conference tournaments (or four times in four days) will perform poorly the following week in the NCAA Tournament, especially in the second round as the fatigue caused by the previous week sets in. To test my theory, I pulled conference tournament results from the past 14 years for the ACC, Big East, Big Ten, Big 12, Pac Ten and SEC. I then compared the number of conference tournament games played with each team's success during the following weekend in the first two rounds of the NCAA Tournament. The table below shows mixed results.

In the first round, the #1 and #2 seeds from the six major conferences who played 1 or 2 conference tournament games where 29-2 while the teams who played 3 conference tournament games (none played four) were 61-0. This indicates that #1 and #2 seeds are not adversely affected by the extra games. This probably makes sense because the top two seeds in each region are likely to be teams with fairly deep benches and therefore less susceptible to fatigue. Teams seeded #3 and #4, however, were 42-6 (88%) when playing 1-2 conference tournament games but 31-10 (76%) when playing 3-4 games. Furthermore, #5 and #6 seeds were 38-16 (70%) in Round 1 after playing 1-2 conference tourney games the prior weekend as opposed to a slightly weaker 20-11 (65%) when playing 3 or 4 times. Strangely enough, teams seeded 7-10 were 42-51 (45%) when playing 1-2 times in the conference tournament but 21-11 (66%) when playing 3-4 times. The performance of the 7-10 seeds is very hard to explain. My best guess is that teams who (a) are seeded 7-10 and (b) played 3 or 4 conference tournament games were probably on the NCAA Tournament "bubble" and played their way into the NCAA Tournament with a nice conference tournament run. They are playing well (they probably upset one or two teams the prior weekend) and have already been playing under the pressure of fighting for their postseason lives. The confidence they gained in the conference tournament may have carried over into the first round of the NCAA Tournament. Teams seeded 7-10 who played just one or two conference games were probably not on the bubble so the first round of the NCAA Tournament might be their first "do or die" situation. I realize this is a weak theory but it's my best attempt to explain the strange data.

The results from the second round were equally strange. #1 seeds who played 3-4 conference tournament games the previous week were 30-4 in the second round whole those who played 1-2 conference tournament games were 13-0. That doesn't tell me much. One of the most bizarre stats I've seen in my analysis is the combination of the 6-10 record for #2 seeds in the major conferences who played 1-2 conference tournament games and the 21-6 record for #2 seeds in those conferences who played 3-4 games in the conference tourney. This would indicate that those extra games were beneficial, not detrimental. The results were mixed for seeds 3-10 with 3-6 seeds performing a little worse when playing 3 or 4 games the previous week. Teams who played exactly four conference tournament games were 9-8 in the first round and 2-7 in the second round. Don't read too much into this stat because 7 of 9 teams who advanced to the second round had the lower (higher number) seed.

Because of the strange results (#7-10 seeds in the first round, #2 seeds in the second round) I wouldn't put too much stock in conference tournament fatigue. In the past, I used conference tournament data going back to 1990 but this year, I limited the analysis to 1997-2010 to be consistent with the other variables. The conference tournaments seem to be less impactful on NCAA Tournament success in recent years. Maybe the players are in better shape than they were 20 years ago. Maybe the reduced travel resulting from the "pod" format has helped. I wouldn't completely dismiss the possibility of a team wearing down in the tournament but, as is the case with all of these stats, I would analyze on a team by team basis. Do they have a deep bench? Did the starters play big minutes in the conference tournament games? Knowing those details will help you determine whether or not fatigue might become a factor.

First Round Upset Profiles

Perhaps the greatest lure of the NCAA Tournament is its unpredictability. The single-elimination, "do-or-die" format creates an excitement that is rarely matched in professional sports. The beauty of March Madness is that a tiny, unknown college from the middle of nowhere can beat a top ten program on a national stage. For one day, a school like Weber State or Hampton could be the lead story on ESPN. Based on past results, we can expect one or two major first round upsets each year (wins by small conference teams seeded 13 or above). This leaves the "bracketeer" with two choices: (A) Play it safe and take the favorites, knowing that you will probably absorb one or two losses but will not fall too far behind the competition or (B) Try to pick one or two major upsets and attempt to get an early lead in your bracket pool (assuming that you don't miss too many picks among seeds 5-12). I suppose there is always strategy C. An example of strategy C is the person who picks a dozen upsets just so they can boast about picking a team like Hampton in 2001. This is bracket suicide. No one will be impressed when this person finishes dead last in their pool. My goal in this particular analysis was to find out why these upsets occurred. I wanted to know if there is a profile for the Davids that won and the Goliaths that lost.

I found that teams seeded 2, 3 or 4 who lost in the first round (there have been 20 since 1997) had slightly worse road/neutral records, records in their last 10-12 games and records against the RPI Top 50. So it would seem that some teams are slightly more likely to be upset than others. However, the "giant-killers" (teams seeded 13, 14 or 15 who won in the first round) had very similar road/neutral and last 10-12 records to the teams that lost (I omitted the RPI Top 50 analysis because these teams play so few games against the RPI Top 50). The giant-killers, however, had a better average RPI (75 vs 90) than the teams seeded 13 to 15 who lost. Prior year tournament experience didn't seem to be a major factor as 21% of upset winners were in the tourney in the prior year as compared to 24% of underdogs who failed to pull off the upset.

Conclusions

As I stated earlier, seeding coupled with pure chance are still the overwhelming factors in deciding who will advance in an NCAA Tournament game. There are factors like injuries, foul trouble, unusually poor shooting, unusually hot shooting, officiating and more that will impact a game. In short, the better team does not always win. My only goal in doing this analysis was to give myself a slight edge on the competition and to prove or disprove my usual logic in filling out my brackets. Based on what I discovered, I believe that a team's road/neutral performance and experience are the two most important factors other than seeding. Teams playing long into their conference tournaments obviously had trouble getting out of the second round but this should be examined on an individual basis because teams with deep benches will be far less susceptible to this problem. Record against the RPI Top 50 should be considered but taken with a grain of salt if the wins or losses were early in the season (let's not forget that Ball State beat UCLA and Kansas to start the season in 2001, then proceeded to lose five games to teams with RPI rankings above 100). I hope you have found this analysis to be informative. You can find breakdowns of the tourney teams on many of the large sports websites such as ESPN.com and CBSSportline.com. The best resource for team record, schedule and RPI information is collegerpi.com. I also recently discovered a great site called statsheet.com.

More Analysis

I have also updated my analysis of tournament performance based on team statistics (free throw shooting, field goal percentage, turnovers, etc). For this analysis, I used data from 1998 to 2010. March Madness Trends Analysis Part II is available now.

Good luck with your brackets.
Patrick Reilly
pjreilly@dublin.com