Busted. Obliterated. Decimated.
That’s how I thought my March Madness bracket would look after the tournament’s first weekend, even with the assist that Sciencing’s motherlode of data provided while making my picks. After all, the one predictable thing about March Madness is that it’s unpredictable.
Instead, everything went…well, OK. The good news is that I got 13 Sweet Sixteen teams correct. The bad news is that my rate of successful first-round picks wasn’t exactly scintillating. Out of 32 first-round games, I picked 23 correctly. That’s not awful exactly, but it’s certainly not grand.
The first round is where I really relied on Sciencing’s data to guide my picks. And that’s where it gets interesting. Who’s to blame here, me or Sciencing? Did the data lead me astray? Or was it human error that canceled out my data-driven advantage?
Let’s do a bit of posthumous forensic bracket analysis to investigate.
Sciencing data indicated it was likely we’d see one upset by a 13-seed over a four-seed, so I picked UC Irvine over Kansas State. And boom! UC Irvine beat Kansas State to register the first round’s biggest upset. That’s exactly how it’s supposed to work, folks. The perfect marriage of man and machine, with data telling what’s likely to happen then a final human touch to bring home the bacon.
It all goes downhill from there though.
Sciencing data said it was likely we’d see between one and two upsets by a 12-seed over a five-seed. Here, I erred on the side of caution and picked just one: Murray State over Marquette. Boom! Correct again. But, improbably enough, two more 12-seeds (Oregon and Liberty) pulled off first-round upsets as well, meaning I got just one of the five-12 games correct. Womp womp.
I picked one 11-seed to beat a six-seed, again guided by what Sciencing data indicated was likely to happen. And indeed, the data did not lead me wrong! One 11-seed did beat a six-seed…it was just the wrong upset. I picked Saint Mary’s over Villanova as my upset here, but Ohio State taking down Iowa State was the one that actually played out.
The seven-seed versus 10-seed and eight-seed versus nine-seed matchups are where I really got hosed. Three 10-seeds won; I picked two to advance and only got one of those two correct in Minnesota. Meanwhile, all four nine-seeds advanced; I picked two of them correctly in Washington and Central Florida.
Now back to this week’s Sweet Sixteen.
As I mentioned earlier, I’m actually in pretty good shape, with 13 of the 16 teams correctly prognosticated. Will my good luck continue? Again, let’s consult the data.
I’ve got one-seeds Gonzaga and Duke beating four-seeds Florida State and Virginia Tech, respectively. Sciencing says one-seeds win these matchups 71-percent of the time, so I’m feeling pretty good about these picks holding up. Meanwhile, I’ve got one-seed North Carolina taking down five-seed Auburn. Sciencing says one-seeds win these matchups 83-percent of the time, so I feel confident here as well.
Elsewhere, I have three-seed LSU beating two-seed Michigan State, two-seed Tennessee beating three-seed Purdue and two-seed Kentucky beating three-seed Houston. Sciencing says two-seeds win these matchups 63-percent of the time, which is almost exactly in line with how I picked these three games playing out.
If the first round taught me anything, it’s that data can only take you so far. Sciencing’s information was certainly helpful giving a general picture of how the first round was likely to unfold. The tricky part comes in figuring out just what to do with that information. For example, data can tell you that it’s likely one of the four 11-seeds will upset a six-seed. But you still need to correctly divine which 11-seed will shock the world. And therein lies the rub.
All of which is to say that, while I’m generally confident in my Sweet Sixteen picks, I wouldn’t be surprised at all to see my bracket go completely haywire from here.
We’ll check back next week to see where things stand.