Bubble Bashers: The Worst Ever?

-Jim Root

In the wide world of college basketball, one of the few things more consistent than Jon Rothstein’s tweets is the “worst bubble ever!” torch-and-pitchfork mob emerging around this time of the year.  Somehow, every single season manages to outdo the prior one, with nationwide alarm sirens warning us that we’ll never find 36 worthy at-large candidates. This has always driven me bananas because of how reactionary/”prisoner of the moment” it feels, so I set out to prove that the bubble is always terrible around this point on the calendar. Many eventual bubble teams are still piecing together a resume, hunting for wins through February and early March (plus conference tournament time), so it strikes me as perfectly normal that many current candidates will lack true substance.

For this task, I took two approaches:

  1. Direct comparison of teams between years, a “micro” approach - I took a quick and dirty sample of bubble teams from 2017, 2018, and 2019 and compared them to each other.

  2. More of a data-driven, “macro” approach - focusing mostly on quality wins, I compared the three years to each other to see if any significant outliers emerged.

Before we start, some remedial at-large/bubble context: overall, we’re going to have around 46-48 “at-large caliber teams” make the field: the 36 actual at-larges, plus the champions of the 10-12 best leagues (Power 6, American, plus likely Gonzaga / Nevada / Buffalo, etc.). That number can grow or shrink based on a few variables, but for simplicity, let’s say 47 “at-large caliber” teams will make it (this will matter more in the data-driven pieces).

Team Comparisons

The most obvious way of looking at this is directly comparing teams, so that’s where I started. Using the Wayback Machine, I pulled data from bracketmatrix.com (another gem) for 2/3/17 and 2/10/18 to find the four teams closest to the cutline - last two in, first two out from all three years as of that date and compared to the current bubble. I then took those 12 teams and made the lovely below table, which shows the number of Q1/Top 50 wins, Q1+Q2/Top 100 wins, Q3/101-200 losses, and Q4/200+ losses:

There’s really not a lot of distinction to be made between any of those teams - they would all be similarly on the bubble in any other year (although that Illinois State / Clemson divide looks really bizarre, in hindsight). This is basically what I expected, and it can be difficult to get a grasp on the wider trends of the year when looking at specific teams, so let’s quickly move to the more data-driven analysis.

Good Wins (Q1 + Q2)

A little added detail on this strategy: I pulled Warren Nolan data (elite website) from the Wayback Machine from the past two seasons at a similar date (2/1/17, 2/3/18) to compare to this year (2/5/19), with the basic determination that I’ll sort by amount of “good wins” and “great wins.” For simplicity, I’m equating the following:

  1. “Top 100” wins from 2017

  2. “Q1 + Q2” wins from last year’s RPI era

  3. “Q1 + Q2 wins” from this year’s NET formula

These shall be known as “good wins,” as each was the criteria for evaluation at the time. This seems reasonable to me, by and large, even if they aren’t exactly apples-to-apples. I made a similar distinction for “great wins”: Top 50 in 2017, Q1 in 2018 and 2019.  

I also limited the sample to only the top 90 teams each year (RPI in 2017/2018, NET in 2019), as that stretches to the fringe of the at-large consideration pool. Rarely, a team may make a major run down the stretch to force itself into the tournament picture, but I feel pretty comfortable that my sample captures the teams most likely to obtain at-large bids.

One small thing to note about that distinction – there were a lot more total wins that fit this description in 2017 than in the past two years:

My quick theory on that (fairly confident in this) is that because there was no distinction for location in 2017, it was easier in general to win games in these classifications (i.e., beating #76-100 at home qualified in 2017, but it no longer does with the addition of quadrants). Regardless, we’ll be looking at the data in ways that largely renders that spike irrelevant, so no worries.

One more note: I understand that classifying teams only by good wins is an oversimplification – excludes SOS, bad losses, etc., but it’s a solid barometer on a macro-level (and those other factors would likely follow similar patterns), so I think the analysis has merit.

Alright, let’s get this started with a simple distribution. Of the 90 teams examined in each season, the following chart shows how many teams had a given number of wins in that season:

That’s a little busy, so to put it differently, here’s a simpler breakdown, broken down into tiers:

This is more indicative of the size of the bubble each year, but that’s a solid place to start.

In 2019, 29 teams have already earned 7+ good wins, which likely puts them on track for an NCAA Tournament berth, compared to 33 in 2017 and 24 last year. Combine that with the numbers for “True Bubble,” and you get a picture of the competition for the tournament: 20 this year, 25 last year, and 16 in 2017. In every season, that gets us to exactly 49 teams total, just about lining up with that 47 mentioned earlier.

Obviously, there’s room for some of the “In Contention” teams to make a movement if they can collect a few significant scalps down the stretch, and a few of the “True Bubble” teams aren’t actually that close to the field (someone like Providence has 4 Q1/Q2 wins but is pretty far from the cutline right now). The point here is more high-level: the bubble looks about the same as it always does, in terms of quantity. 

But Jim, we didn’t say smallest bubble ever! We said worst!

Right you are, ol’ chap. One way to examine the “strength” of the bubble teams is to see what percentage they possess of the overall available pie of good wins:

This year’s “True Bubble” teams possess 24% of the available needle-movers, down from 2018 (30%) but up from 2017 (18%). That supports the notion that this year’s bubble is merely “average,” and nothing exceptionally better or worse than last year.

As stated in the intro, the timing of this analysis (both mine and by Joe Bubble) is the key. The remaining schedule holds a tremendous amount of additional “good win” opportunities for everyone, and plenty of teams who haven’t exactly distinguished themselves yet will likely do so in the next 40 days. As long as bubble teams continue win their share of these key games – as they’ve done so far this year up to this point – Selection Sunday will offer a similar challenge to most years.

Great Wins (Q1 Only)

I decided to dig even further into the detail, zeroing in on what I’ll call “Great” Wins – so Top 50 in 2017, Q1 in 2018 & 2019 – just in case that wrinkle would unveil something further. Let’s start with the same kind of tier distribution:

This is where it starts to get at least a little bit interesting. Forty-four of our 90 teams have yet to amass a second difference-making victory, up 4 from last year and up 8 from 2017. That’s where you start to have teams like San Francisco (0), UCF (1), Butler (1), Clemson (0), Creighton (0), etc. competing for the last couple of bids, despite not really having anything remarkable atop their ledgers. Two Q1 wins isn’t a magic number by any means; this year’s Penn State and Oregon are both in our “top 90” sample, have multiple Quad-1 wins, and wouldn’t sniff the tournament if the draw was held today. It is, however, something that Belmont, Arizona, and Georgetown (among others) will cling to as a differentiator, with the Hoyas the most likely to add to that number.

However, before we go condemning this year’s bubble due to a lack of Q1 wins: the percentage of the available pie is exactly the same as last year, as the 3+ win teams combine to have 72% of the wins, compared to 28% for those residing more in bubble territory:

They’re distributed among more teams (rather than concentrated on the teams with 2 wins), so I can see how that’s being perceived as a weaker bubble. It’s a small difference, but one with some merit, showing that fewer teams are making convincing arguments for inclusion via great wins at this point in 2019.

So, What Did We Learn?

I feel fairly confident in saying that this year’s bubble is roughly equal to that of years’ past. Perhaps it’s a little softer at the bottom, but it’s nothing dramatic. There’s also a valid argument to be made that earning Q1/Q2 wins is more difficult when using NET as the organization tool: it does a far better job of measuring who the best teams are than the defunct RPI, so there are fewer chances to steal “big wins” against teams that don’t actually belong in the country’s upper echelon. But that is an entirely different article, and I won’t dive into that now!

Of course, I understand that it’s fun to be hyperbolic when talking hoops, and proclaiming “the worst bubble ever” is an easy lightning rod to stir up chatter. It derides the teams on the bubble, hopefully spurring them to bigger efforts in must-win games (I saw you LIVE Tuesday night, St. John’s!), and creates buzz around the sport as many football fans turn their focus to hoops. So while I think saying “worst bubble ever” is almost always wrong, I won’t make any grand call to action to eradicate the phrase. Instead, let’s just be aware of its exaggerated nature and hope that like most years, the bubble eventually filters down to a few borderline teams by the time Selection Sunday rolls around.