NCAA Tournament March Madness pools are as common these days as reality television shows. Seemingly everyone is entered into one and follow their brackets like gospel. However, most people neglect to take into the account the importance of the proper scoring system.

Yahoo, ESPN, and other sites cop out to appease the masses and use the “doubling points system.” In this system, people that really don’t watch or care about basketball are rewarded because having the correct champion is worth as much as the entire first round combined. You read that correctly, 1 game = 32 games.

I’ve yet to hear a convincing argument on why this is a good system. In fact, I contacted Yahoo and ESPN asking them why they use that particular system. The response that I received was that: “it’s easy to understand” and “allows people that don’t follow basketball to be more competitive.” The second response more or less baffled me. It’s a skill competition, why dumb it down so a 4 year old who knows nothing about basketball can simply pick two one seeds, put them in the finals and rack up the points. Yes, I know the finals don’t always pit one-seeds against each other, but it’s the most common situation and outcome.

The scoring system that I have used in my pool is a modified progressive system that rewards correct picks by round in the following fashion: 1 point for a correct first round prediction, 2 for a second, 4 for a third, 6 for a fourth, 8^{th} for a fifth, and 10 for having the champion. While the system that I use isn’t necessarily mathematically perfect, it’s close, correctly rewards the prognosticator who correctly predicts the champion, and negates the following scenario that is possible under a 1, 2, 4, 8, 16, 32 system from occurring:

**Player A who has:**

29/32 correct in rd. 1

15/16 correct in rd. 2

4/8 correct in rd. 3

2/4 correct in rd. 4

1/2 correct in rd. 5

0/1 correct in rd. 6

for 107 points,

**would lose to Player B who has:**

10/32 correct in rd. 1

7/16 correct in rd. 2

5/8 correct in rd. 3

2/4 correct in rd. 4

1/2 correct in rd. 5

1/1 correct in rd. 6

for 108 points.

Player A would have 51 wins out of 63 and lose to player B who has 26 wins out of 63 with 63 of their 108 points coming from 1 team. If you can only get 40% of your games right you shouldn’t beat out someone who got 81% right. Essentially, in a normal year, player B just put 4 #1 seeds in the elite 8, 2 in the finals, and guessed the correct one to win and beat someone who hit more than 4/5ths of their picks and had the team that lost in the finals.

I’ve encountered the argument that bracket pools are just about picking the winner of the tournament. If that were the case, it wouldn’t be a bracket pool; it would be a question… which team is going to win? People that watch and study the teams shouldn’t be punished for their knowledge and analysis.

Another argument that I’ve read is that hitting 80% of the games is nearly impossible. This is a fallacious argument, but also good argument for NOT using the doubling points system, since if a player is accomplishing something that is “nearly impossible,” they should certainly beat a player that has more losses than the Clippers.

To take it a step further, and using history as a guide, here is a breakdown of just picking the better seed (using an average of 6.2 lower seed 1st round upsets as a guide). Obviously year to year it is a fluid machine, and this accounts for rounding.

26/32 higher seeds win round 1

12/16 higher seeds win round 2

5/8 higher seeds win round 3

2/4 higher seeds win round 4

1/2 higher seeds win round 5

1/1 higher seeds win round 6

47/63 (75%)

There are better studies based on more advanced logarithmic permutations and calculus, but those really aren’t necessary to belabor a point that is painfully obvious to see.

In my opinion the only way where a grossly over-weighted system should be used is if they are using seed multipliers. Seed multipliers provide a fun twist on bracket pools, but in reality shifts the matchup breakdown away from true technical accuracy and turns it into a best available bad decision analysis.

To drive home the point in another way, using the 1, 2, 4, 8, 16, 32 system simply having the winner and runner-up of the tournament correct, and every other game either blank or incorrect would give player B a win over player A that has a pretty good tournament sheet, nails nearly 2/3rds of their games, and has the runner-up in the finals, see below

**Player A**

21/32 – 21 points

12/16 – 24 points

4/8 – 16 points

2/4 – 16 points

1/2 – 16 points

0/1 – 0

43/63 (63.5%) = 93 points

**Player B**

2/32 – 2 points

2/16 – 4 points

2/8 – 8 points

2/4 – 16 points

2/2 – 32 points

1/2 – 32 points

11/63 (17.5%) = 94 points

Regardless of anyone’s take on the situation, it should be easily agreed upon that someone with 40/63 correct and the runner up is more deserving of a win than someone that left their sheet blank aside from the winner and runner up and had 11/63 correct.

Whether you are a seasoned veteran of bracket pools or a first-timer, make sure you check the points system and plan your picks accordingly.

Best of luck with your brackets!

I just filled out a bracket in Black Sharpe Marker, and decided to Eat It after reaidng this article…twice.

RISE AND FIRE….COUNT IT.

Jason, I hear you. The one thing you want in a scoring system is balance. The biggest complaint with the typical multiplier system is that you usually have to pick the national champion to win, making the first-round games basically meaningless. The first round should mean something. After all, it’s half the tournament!

Someone who does extremely well in the first and second rounds (3/4 of the tournament) but who stumbles thereafter should nonetheless stand a chance against someone who did only so-so early on but who did well late. A balanced, robust system would allow that.

So why not make it simple? The most games correct wins. The reason, I think, is that not all games are equal. And this is another complaint with the typical multiplier system: it treats all the games within a round the same. You get 1 point for No. 1 beating No. 16, and you get the same for No. 13 beating No. 4. We know that’s not right. The first was a no-brainer, the other took onions, or wishful thinking. Reward the player who gambled and won.

But if an upset is worth more than chalk, it could mean that the player with the winning bracket did not pick the most games correctly. That’s okay, so long as the player who did was not far behind. You don’t want to encourage contestants to pick the upsets, and you don’t want to encourage them to pick the favorites. There should be no gaming the system. Pick the most games correctly, and you should be in the money most years if not the winner’s circle.

How do you go about designing a system to achieve these goals? Once you realize that not all games are equal, that each matchup is different, you’ll want the points for each game to reflect that difference. One way is to use the seeds.

In each region, the teams are seeded 1 to 16. What’s the easiest game to pick correctly? The 1 over 16, no doubt. No insight or luck needed there. It’s the first thing you pick in your brackets, and you never give it a second thought. So expect a minimum reward for going out on that limb!

What’s the hardest? You might say the 8-9 matchup. It’s certainly a much harder decision, a toss-up, one of those games that takes a long time to decide, and you should be rewarded more for getting that one right.

But the hardest game to pick correctly is 16 over 1. Although it has almost happened a few times, no one in their right mind would sacrifice a possible national champion for such a monumental upset. But you know, if you did pick that upset correctly, your score should reflect the monumental risk.

You can rank all the regional matchups from easiest to hardest based on seed. A 1 over 16 scores 1 point, a 1 over 15 scores 2, a 1 over 14 scores 3, and so forth. A 1 over 2 scores 15. If a 1 seed meets another 1 seed in the Final Four or national championship game, it’s a battle of equals, theoretically a little tougher game than a 1 meeting a 2, so score 16. Indeed, any battle of equals should be worth the same.

On the upset side, the process continues. A 2 over 1 should be worth more than a 1 over 1. It’s an upset, barely, so it scores 17. A 3 over a 1 scores 18, a 4 over a 1 score 19, and so forth. A 16 over a 1 scores the max, 31.

It doesn’t have to be linear – the strength of the teams certainly isn’t – but it’s manageable, and we can toy with that later if we want. The idea here is to recognize the different matchups and to reflect that difference in the points for each game.

You can set up an Excel matrix with labels 1 through 16 across the top (the losing seeds) and 1 through 16 down the left side (the winning seeds) and fill in all the scores in linear fashion. Or follow the formula: Score 16 minus the difference of the seeds for correctly picking the favorite, and 16 plus the difference of the seeds for correctly picking the upset. Score 16 for like-seeded teams. You’ll quickly see a linear pattern emerge in the matrix, such that you won’t even need the formula to complete it.

No multipliers. This isn’t a round-based system that skews the weight toward the later rounds. It’s a game-by-game matchup system. Correctly picking the champion will still accumulate a lot of points, but not nearly so much as the typical multiplier system. Indeed, it almost never amounts to 20 percent of your total, which is still a really healthy but not necessarily decisive share.

Last year, UConn accumulated 75 points out of ~400. But you could have scored the same by picking the three biggest upsets, including No. 11 VCU over No. 1 Kansas (26 points). So you can make your points elsewhere, but only if you’re right!

I’ve been running this for 20 years, for 25-30 people, and I love the balance. The winner is usually the person who picks the most games correctly, the sign of a good system. It means there is no real way to “game” the system. It’s not top-heavy or bottom-heavy. If you had picked all the favorites in the first round last year, you would have scored 179. If you had picked all the upsets (there were only 7), you would have scored 147. A good mix, though the upset strategy would have left you with only 7 teams and precious little opportunity to score more points. If you had picked the favorites all the way, you’d have ended up in the middle of the pack. Again, the sign of a good system.

I’ve not seen its equal. Feel free to comment and tell me what you think.

A friend just suggested this nice little twist on the 1,2,4,8,16,32 system: Simply multiply the points you earned by the percentage of games you picked correctly.

In your example above, Player A would score 93 * 0.635 = 59.1. Player B would score 94 * 0.175 = 16.5. Quite a difference. What do you think? It’s simple, and it balances things by returning the importance of picking a lot of games right.

Both very interesting systems.

You touched on the reason why I dislike multiplier systems (it rewards people for picking upsets). When I say that I mean this, as I sit down to my bracket, the first thing I do is ignore the seed number. At the end of the day the seed is an arbitrary number given by a committee trying to do the best job they can. Are they always right? Absolutely not.

I am also not a “trend” guy with regards to selecting winners. I don’t care if X team hasn’t won in a city that has 3 syllables and starts with the letter I since 1972. I examine the matchups as I see them, do my best to project the winner and roll with it.

There is a lot of consideration taken into these events, for example, how a team is playing over the past month, any injuries, how the styles match up, etc. But at the end of the day, I am writing down the name of the team that I believe will win that game. Seed multipliers and things like that don’t reward someone for critically analyzing a game and trying to find a winner as much as they reward picking all the 12&13 seeds into the sweet 16. In other words, it’s a mathematical game rather than a stand-alone prediction.

As I ramble on while thinking about a couple matchups that are very close in my head, I think we both agree on the point of the exercise, to reward the person who has the best bracket based on # wins, % wins, & overall performance.

I was kicking around the idea of supplementing the overall payouts with payouts for performance in each round. I will probably hold off this year because people are already in my pool, but it’s something else to consider.

Thanks for the thoughtful comments and critical rationale.

If you are interested in joining my bracket pool, hit me with an email: jmarlo@sidepoints.com

-Jason