The Tom Izzo Bermuda Triangle: 2026 update

Prior to the overtime thriller over the Illinois on Saturday night, the Spartans had lost two game in a row and had not played well for three straight games. To fans, it feels like Michigan State does this almost every year. It is the very nature of college basketball that good teams sometimes take bad losses, such as a road loss at Minnesota. It happens. But, it is the timing of the skid that feels so familiar.

Several years ago in some the corners of MSU Spartans sports internet, someone pointed out that Michigan State's winning percentage frequently took a noticeable dip in late January and early February. This was time of year was christened as the "Tom Izzo Bermuda Triangle."

I decided to also dig into this phenomena from a few different angles. I have a good database of MSU game result and spread data back to the 2006 season. I decided to study this effect using three factors: raw win percentage, performance against the spread, and performance relative to the Vegas spread expectation, also known as PAVE.

Let's start with the raw win percentage and the performance against the spread data in Figures 1 and 2.

Figure 1: Michigan State's raw win percentage for the 2006-2025 seasons as a function of time (days from the beginning of season).

Figure 1: Michigan State's performance against the final Vegas line (right panel) for the 2006-2025 seasons as a function of time (days from the beginning of season).

In these two figures, I am calculating the seven-day average of the Spartans' performance for the 2006 to the 2025 season. Over this span, Michigan State has won slightly more than 70 percent of their games straight up and is just over 50 percent against the spread. These average values are shown as the horizontal line on each graph for reference.

While the data in both graphs fluctuates, there is only one part of each graph where there are several data points in a row that deviate noticeably from the average. The area of each graph is highlighted with the dashed oval and it spans a roughly three-week period from mid January to early February. 

During this time, on average, the Spartan's raw win percentage and performance against the spread dip by about 10 percentage points. This is the "Tom Izzo Bermuda Triangle."

But raw win percentage and covering the spread (or not) is perhaps not the best metric. The former does not account for the strength of the opponent and the latter does not account for winning or losing. Both metrics also also either "on or off", which makes quantification more challenging.

Therefore, I decided to run the numbers one more way that combines the best of both worlds. For each MSU game back to 2006, it is possible to calculate the odds that MSU should have won each of those games based on the final Vegas line. 

This "performance (a win or a loss) against this Vegas expectation" (or PAVE) is a scaled metric that accounts for winning and losing and the strength of the opponent. In my periodic odds update, this is essentially equivalent to the metric that I refer to as "luck."

In order to understand the PAVE metric, consider a situation where Michigan State is a 13-point favorite in a game. The odds of a Spartan win are 90 percent, based on historical data. If the Spartans win that game, it is just slightly better than expected. 

Mathematically, I would credit Michigan State with +0.1 of a win (or 1.0 minus 0.9) over expectation. However, if the Spartans were to lose that game, it would be a big hit. In this case the Spartans' would lose 0.9 of a win (or 0.0 minus 0.9). 

In cases where the Spartans are a big underdog, the situation is reversed. In a game where Michigan State is a 13-point underdog, an upset win would earn the Green and White +0.9 of a win, but a loss would only subtract 0.1 of a win from the total. Games that are near toss-ups simply add or subtract roughly 0.5 of a win, depending on the exact spread.

Using this method, it is possible to track the Spartans trajectory, relative to expectation, throughout the course of each season. Figure 3 below shows the average PAVE for Michigan State divided up over the same seven-day periods as Figures 1 and 2.

Figure 3: Michigan State's performance against the spread (PAVE) average over seven day averages for the 2006-2025 seasons as a function of time (days from the beginning of season).

Once again, Figure 3 contains several fluctuations over the course of the year. But the only portion of the figure where there are several data points under the green line are in the same period of time from late January to early February.

The best representation of the data that I could think of to visualize the Tom Izzo Bermuda Triangle was to track the cumulative average PAVE metric as a function of the number of days from the beginning of the season. That data is shown below in Figure 2, and it is striking.

Figure 4: Cumulative average PAVE for Michigan State for the 2006-2025 seasons as a function of days from the start of the season.

The graph above appears to accurately trace the trajectory of an "average" Michigan State season. The first few games can be a bit rough, but then momentum (and wins) starts to build into the start of the new year. 

The metric holds steady in early January, but in late January or early February, there is a crash. Then, there is then a strong surge where the Spartans tend to overachieve for the rest of the season, right into March Madness.

Sounds familiar, right?

Granted, there are bumps along the road. There seems to be a dip right before Christmas and another dip right at the end of February, but the overall shape of the curve is very telling. There is also a lot of variance in the data from year to year. Some years don't follow this pattern at all, but more than half do. 

In looking at the data for the individual years, the mid-season dip starts on average on Jan. 25 and ends around Feb. 7. In other words, the Tom Izzo Bermuda Triangle lasts for a period of about two weeks in the final week of January and the first week of February.

In situations like these, it is always hard to separate cause, effect, and mere coincidence. That said, my scientific training tells me that there is a trend here and that there is likely something behind it. My working hypothesis is that it relates to the way the Izzo structures his program. 

The beginning stage of the season is when Izzo and the staff just try to understand what pieces that they have and how they fit together. In mid-season, Izzo goes into the lab and starts to tinker. At this point he understands his team and he is trying to figure out how to get the most out of them.

Izzo is trying to tighten the screws in order optimize his team's output. This is when the mid-season dip tends to happen. It is a mid-season set of growing pains.

Once February hits in earnest, Izzo shifts into full post-season mode. He knows what he has and he has pushed his players to be the best possible versions of themselves for the given year. At this point, it is time to just turn them loose. It doesn't always work, but the trophy case in Breslin Center is a pretty good indicator of the success of this strategy. 

If nothing else, Figure 4 gives a very quantitative measure of this success. Over the past 20 seasons, the Spartans have won a total of 11 games more than the Vegas lines projected from Feb. 9 to the end of the year. I doubt any other team or coach in the country could boast such an accomplishment.

But what does this analysis tell us about the remainder of the 2026 season?

Last Saturday night's win over Illinois on the heals of three straight poor performances is a good indicator that the Spartans has passed through to the other side of the Tom Izzo Bermuda Triangle. The game also happens to have occurred on precisely the average final day of the mid-season dip.

If the historical trends hold true, Michigan State is primed for a strong finish to Big Ten play. Friday night's road trip to Madison to face the Badgers will be an excellent test to this hypothesis. The subsequent homes games against UCLA and Ohio State will provide a similar test. 

A typical Tom Izzo team of caliber of the 2026 squad is expected to win all three of these games.

However, Figure 4 also suggests that a small dip at the end of February or early March is also very possible. Based on the schedule, the back-to-back road trips to Purdue and Indiana as well as the regular season finale at Michigan all fall into this timeframe. 

A loss or two in the final four games of the regular season would certainly not be a surprises or a historical anomaly.

Once March starts, as everyone knows, few coaches have had as much as success as Tom Izzo. While probably is not destiny, the odds are strong for another deep run in the post season.