USA Today

Don’t pick Duke!
Play to win, not to not lose…so don’t pick Duke!

While tomorrow much of America will be ignoring work to watch basketball, today many will be ignoring work to figure out who to pick in their work pool. I am here to help!

Don’t be That Guy!

There are two main mistakes I have seen people make over the years. One, they pick too many favorites to make the Final Four. Two, they pick the same teams everyone else picks. In the investing business, we call this being long crowded stocks.

Owning crowded stocks is usually NOT a good thing! It means you are consensus which means, even if you’re right, the stock probably won’t move as much as you expect because there are no incremental buyers.

In bracket picking, being long crowded teams has a similar effect. You don’t win a pool by getting the championship game right. You win a pool by getting the most points in your pool. While getting the champ right helps a lot, it doesn’t guarantee anything. If you picked the most popular team to win and they do indeed win, you are probably NOT winning your pool.

The Duke Paradox

I call this the Duke Paradox. In the ESPN Tournament Challenge, 40% of people have picked Duke to win the title. FORTY!!! That means if you pick Duke and are right and your pool has 100 people, you have to beat the other 39 in getting all the other games right to win your pool. That is really hard.

You might argue with me that since getting the champ right pays so many points in your pool that you MUST pick Duke. No, you mustn’t. There is no GUARANTEE that Duke wins. What if they don’t? Then, someone who understood the odds better and picked a different one seed will win.

This is a lot like people who pick their family members’ birthdays in the lottery because “there is an equal chance of any number being pulled”. True, but there is not an equal chance of whether you have to split the pot if you win and if you pick lucky numbers, you are much more likely to split the pot! Picking Duke is like picking lucky numbers!

For context, the 538 model has Duke as the favorite to win the title. At 19%. Not 40%! They have Virginia at 17% and Gonzaga at 15%. In other words, there is no clear favorite.

Let’s say they are too conservative and Duke really has a 25% chance to win. OK, well 75% of the time you’re wrong and of the 25% you’re right, remember you have to beat those other 39% of your pool in all the non-Duke games. In other words, you probably have to get at least two of the other final four teams right, if not all three.

Let’s look at the ESPN consensus again. Carolina is second at 15%. Let’s toss them out. Gonzaga is next at 8% followed by UVA at 7%.

If someone told you the real odds were Duke 19% UVA 17% and Gonzaga at 15% but that 40% of the people bet on Duke, 8% on Gonzaga, and 7% on UVA, what logically should you do? I would say pick either Virginia or the Zags because, if you’re right, you only have to beat a small group of people in all the other games to take the trophy. Maybe even pick a UVA-Gonzaga final? That combo is probably at 3-4% of all entries.

If You’re Not First, You’re Last

The key thing to remember is nobody remembers if you finished 5th or 50th. There is no consolation prize. You need to avoid the crowded picks to increase your chance of winning. That means you have a greater chance of finishing near the bottom. So? If you finish say 10th in your pool, you probably aren’t doing it right. You were too conservative. You need to embrace volatility. Don’t be afraid to finish last!

The Ian’s Blog Bracket Challenge

OK, that’s the lesson. Now, let’s get into some of the data. I’ve put five brackets up on ESPN that you can track over the next few weeks. Each is a different approach to selecting.


First, we have the chalk bracket. It’s very simple. Pick the lower seed in every matchup. You have four 1’s in the Final Four and a Duke-Virginia final. This is essentially our control group.


Next, we have the consensus group. This is essentially the crowded picks. I used the ESPN popularity index for these. There is clearly some judgment involved but I picked first round upsets for any higher seed that was over 50% or for 5 or 6 seeds < 60%. The Sweet 16 was nearly all chalk other than knocking out Purdue who was only at 50%. Then it went all chalk from there with four 1’s in the Final Four and Duke winning.


Then we have the Analytics bracket which was a combination of 538’s predictions and betting lines for the first round. I also used historic results (note: it’s a gambling site if you can’t access it, blame your firewall) for each seed to try to have a representative spread of seeds at each point in time.

So, for example, typically two 3s and two 4s make the sweet 16 so that’s what I did. The Final Four is usually one or two 1’s , one two, and one 3 or 4, so I did two 1s, a two, and a three.

Typically, one 4 loses in round one and I picked that one by choosing the 4 that had the lowest odds in 538 and was the smallest favorite. This knocked out Kansas State. Other first round upsets were done similarly. Note, I came very close to knocking out LSU as a 3 in the first round but decided not to as most years all 3’s survive but they are the greatest risk if you want a 3-14 upset. Duke still does come out the winner here given the 19% odds are the highest in the 538 model.


Next, comes the contrarian model. This was a little tougher to put together as rather than start in round one, I started at the finals and worked backwards. As discussed above, I went with Gonzaga and Virginia in the final since they are under-owned by the public. Because I have the two least picked one seeds, I felt no need to take chances with the other two regions so went with both two seeds, Kentucky and Michigan State.

Also, Tennessee was the 2 seed most often picked for the Final Four (because they’re in Virginia’s region) so I was willing to eliminate them early. My last 8 were the four 1s, three 2s, and one 3. Again, no need to go out on a limb with bigger upsets given I only have to beat very few others if I get the final right.

My Sweet 16 does have an 11 and 13 though they largely fell out of the process as, in contrast to the Analytics approach where the favorite with the lowest odds was eliminated, in this case I eliminated the most popular 3-6 seeds early and the 3 & 6 were in the same subregion as were the 4 and 5.

Similarly in the 7-10 and 8-9 games, I picked the least chosen teams to win. In other words, I will do poorly on the first weekend if the “expected upsets” like Oregon and Murray State happen, but will do exceedingly well if the less likely double digit seeds win.

Stock Pickers

Finally, I will put up my own bracket tonight which I am calling the Stock Pickers bracket to try to strike a blow for active management!

To see the results of all this, you can view my bracket group at ESPN. If you want to enter your own picks, you are welcome to as well as the group is open. I will post updates throughout the tournament!