When it comes time to fill out your bracket, you’ll have 63 match-ups to decide. I’m here to set your mind at ease: you probably shouldn’t even think twice about 16 of them. The stats say that picking any first-round upset lower than a 5v12 makes little sense.

I’m not suggesting that you’ll get all 16 of these picks right, mind you. I’m just saying that chasing longshot upsets is a good way to destroy your bracket early. Consider this: if you had automatically advanced the top four seeds in every one of the 28 dances since the tourney expanded to 64 teams, you’d average getting 14.4 picks right per tourney. That’s 89.7% accuracy. And 14 or 15 out of 16 isn’t going to bust any bracket.

The seed advancement probabilities posted below should be enough to convince you a the folly of chasing a 3v14 or 4v13 upset. Heck, using those probability percentages, there’s a 16.5% chance that all 16 of the top four seeds will advance to round two. Over the course of the 28 tourneys, that would mean you’d get a perfect score four or five times. In fact, reality reflects these percentages. Here are the number of times that you’d score 16, 15, 14 and 13 since 1985 (that’s right: this strategy has never scored worse than 13):

- 16 – four times (1994, 2000, 2004 and 2007)
- 15 – seven times
- 14 – 12 times
- 13 – five times

Don’t get me wrong: this strategy probably isn’t going to win you the ESPN Tourney Challenge, where you’re competing against 10 million brackets. But it’s more likely to win you your tourney pool, where there are just 100 other players to beat.

If you’re still thinking about tabbing a 13 or 14 seed to spring an upset, consider the impact beyond round one. Only seven of these 224 teams (3.1%) have won their next game. Meanwhile, that three or four seed you just cut out of your bracket has a 47.3% (106 of 224) chance of advancing to the Sweet 16

Here’s a little demonstration of the pitfalls of picking a first-round longshot upset. There were five dances since 1985 where 15 of the one through four seeds advanced in round one and the lone upset was a 13 knocking off a four. Let’s say you decided to chase that 13 seed Cinderella in each of those tourneys, picking one 4v13 upset. Odds say that you’d get one of those five picks right. If that were the case, you’d have a perfect 16 score in one of the five dances, but a 14 in the remaining four dances—three picks worse than automatically advancing the high seeds.

To better your average score of 15 correct picks per 16 games, you’d have to be 60% (three for five) correct in your choice of the right 4v13 upset. That’s 2.4 times better than a random pick of the four games. Not going to happen consistently. And never mind that you’ve done all of this for the sake of improving your bracket by a single point in round one. Remember: with each round, you get more points for correct bracket picks—in some pools, the score increments by a single point; in others, it doubles. So for the long odds of scoring one extra point, you would be sacrificing the much greater possibility that the high seed you eliminate will amass more points in later rounds. Don’t do it.

Interesting analysis. How does this change if one considers underdog points (i.e. 7 points for a 12 over 5 seed)?

Good question, Jack. My database doesn’t track point spreads. But I can tell you that the average scoring margin for 5′s over 12′s in round 1 is 4.6ppg. So on balance, I’d take a seven-point cushion. Coming soon to the *Tips section will be my 2013 Seed Guide. That will have the records of every seed match-up that has ever occurred in the tourney, along with the average scoring margin. Interestingly, the average margin of 4′s over 13′s in 112 games is exactly 9.0ppg. Pick ‘em.

Pete,

I think the original question is related to in a tourney pool, sometimes upsets are awarded bonus points (underdog points) in the scoring. A 12 vs 5 round 1 victory is worth the seed differential, thus 7 points vs just the 1 point for the round 1 it occurs in.

Gary

I see your point. I guess I might consider a 4v13 in that case. A 21.4% chance at nine points is good odds I suppose. Think of it this way: 13 seeds win 3.14 of the first round games. Four seeds in .86. Multiply that by 9 and you get an average of 7.74 points to 3.14. Heck, most years it would pay to take all 13 seeds.

Pete,

Thanks for the thoughts. It allcomes down to looking at how your pool is scored and then looking for “opportunities” to take the right chances on lower seeds to pull upsets and/OR identifying higher seeds that are less likely to make a deeper run in the tourney. THAT IS WHERE BRACKETSCIENCE somes in!!!!!!!

Gary