What would the optimal strategy for an NCAA basketball pool be, assuming one knows nothing about basketball?
In such a pool, you get points for correctly picking the winners of the 63 games. Usually 2 points for first round games, and doubling for each successive round.
Now let’s assume that the market (the betting public) knows more about basketball than I do. Since point spreads don’t matter in the pool, I would think one would maximize his expected value by choosing the favorite.
But…this is like those stock picking games (which I consider rather silly) you have to play in high school. You want to come in first, doing well gets you nothing. So to win the stock game with a sufficient number of contestants, you have to lay your balls on the block. Blue chips won’t win the game, as they won’t separate you from the pack.
So, in the pool, I would think you need some underdogs in there. First, is this a reasonable analogy/assumption? Second, is there a way to determine the optimal number of dogs, and which ones based solely on seeds and or point spreads?
I doubt there’s an exact answer, without a complicated linear program, but any insight would be appreciated. This question is fueled by intellectual curiosity, not my desire to win 50 bucks.
Pick the favorites in every game. Yes, correctly identifying an underdog who wins carries a bonus, because it’s a pick that others where likely to have missed. However, picking an underdog who doesn’t win robs you of points that most everybody else will receive.
Also, it’s worth noting that upsets are more common in the early rounds, when the point totals are lower, and less likely in the lower rounds, where they are worse.
If you can correctly pick three of the final four teams, that’s worth much more than correctly picking a half-dozen upsets correctly on the opening weekend.
Only three times since 1985 have 3 #1 seeds made the final four and all 4 #1 seeds making the final four has never happened. OTOH, in 1980 and 2006 no #1 seed made the final four. So they chance of you hitting 3 is almost as likely as you hitting none.
You will NOT win by picking the favourites in each game.
Re: randomness (Posted this in another thread)
"Last year, I played 5 brackets on ESPN and one of them ended up in the 99th percentile where I correctly picked three of the final four and Florida as the eventual champ.
Of course, that was the one where I just picked teams randomly. Sometimes I’d go by the longer school name or the more fiecesome mascot, it’s a crapshoot."
Also, if I remember correctly I picked against schools with religious affiliations.
In the past 3 years, I’ve finished in the 88th, 52nd and 64th percentile with the brackets that I researched thoroughly.
Not sure if this is prohibited by the OP but I’d look at the RPI Ratings , and fill my bracket out accordingly.
Upsets are fun to pick correctly. The problem comes when you pick an upset and it doesn’t happen. Many are touting **Texas A&M Corpus Christi ** as the Cinderella this year. However if Wisconsin shows up to play you could very well have lost 12+ potential points just because you were trying to get cute and you wanted that extra 2 points.
If you feel the need to pick an underdog in the first round, I’d make sure their glory is sure to be short lived. Basically pick a team whose likely to get rolled in the next round anyway. That way you can be +2 on everyone who picked the favorite and even if they CONTINUE their dramatic run, you won’t lose any ground because nobody else picked them either.
FTR…for my upsets I’ve got Winthrop over Notre Dame, GTech over UNLV, Illinois over VTech, Zags over Indiana, Texas Tech over BC, and Xavier over BYU with all of them losing in the second round.* Most of these aren’t biggies though.
I define an upset as a spread of four or more between seeding numbers. So the #9 seed beating the #8 seed, for instance, or even the #10 seed beating a #7 seed don’t really qualify as upsets for me. An upset would be closer to a #12 taking out a #4. It’s okay to have a few upsets in the first two rounds (a #12 seed seems to always win every year) but few, if any, should make it to the Sweet Sixteen or later. George Masons are rare, after all. The bracket I’m using one place has two #3 seeds (Oregon and Washington), a #2 seed (UCLA), and a #1 seed (Ohio St.) in the Final Four.
It depends on what you mean by random. If you mean that the winner flipped a coin for EVERY SINGLE game. . .no way on earth did that person win the pool. No way.
Anyway, yes, picking “favorites” in every single would be an optimal strategy. In every single game, you have chose the team most likely to win. Almost by definition, that’s optimal. I guarantee if you were able to do that in every single March Madness pool from here to eternity, you would win more than your fair share.
Keep in mind, we might be talking about changing your probability of winning from 1/50 (in a 50 person pool) to 1/48. . .so you still might go from here to the end of your career and never win one. But if you were playing from here to eternity, that would be optimal.
HOWEVER, a “favorite” in a game is not necessarily what you think it means. Believe me when I say that the people who set the lines in Vegas have a better idea about the strength of the teams than the selection committee.
So, to pick your favorites in every single game, you would need to be able to determine whether Texas A&M is going to be favored over Memphis (probably) in the Sweet 16, and whether Maryland is going to be favored over Butler in the 32.
For instance, Marquette (8) is an underdog in the first round. BYU (8) is an underdog in the first round. UNLV is an underdog, as a 7 seed playing a 10 seed. Kentucky (8) is a slight dog.
How about “never have 4 #2 seeds made the final four”
“never have 4 #3 seeds made the final four”
#1 seeds are still MOST LIKELY to make the final four. They’ve made more final fours than any other seed. If you are truly some awesome expert that has broken down the Ohio State-UCLA match up, and find that UCLA would come out on top in that game, then more power to you, but just because 4 #1 seeds have never made the FF doesn’t mean that shouldn’t be your choice.
The “opposite” of “4 #1 seeds have never made the final four” is EVERY THING ELSE POSSIBLE. . .
three 1’s and a 2.
three 1’s and a 3.
three 1’s and a 4.
two 1’s and two s’.
two 1’s a 3, and a 2.
etc. etc. etc.
When you finish your pool this year, Kid_A, and you have your two 3’s, one 1, and one 2 in there (or whatever), go back and look at how often THAT has happened.
The fact that four 1’s have never made the final four doesn’t mean it’s not the most likely scenario. It might have a 5% chance of happening, but it’s still the most likely scenario.
If you come from complete ignorance you can’t do better than pick the favorites, because that is the best information you have to go on. However, you still will probably not win the pool if there are a reasonable number of people.
Picking randomly gives you a chance of winning. Picking the favorites gives you no chance. The problem is, there is probably going to be someone else, or several someone elses, who pick all the favorites. So if you pick all the favorites, too, your best possible outcome is to be a small fraction of a multi-way tie.
How many people are in on this pool? That’ll have a significant effect on the optimal strategy. If there are just two or three people, then picking the favorites is the best choice, but if there are a hundred, then it’s almost certainly not. It would also help if you knew anything about the motivations and betting strategies of the other players: This information wouldn’t be needed in a true game theory problem, but game theory assumes that all players are perfectly rational, which probably isn’t a good assumption, here.
The only concrete advice I can give you, off the top of my head, is if there’s an underdog who’s disproportionately favored among the other bettors (this can happen, for instance, if it’s perceived as a “home team” for a lot of the folks). If this is the case, then you definitely want to bet on the favorite in those games, not on the underdog. This way, you get both the advantage of picking the most likely outcome, and of being different from other folks.
Betting against the pack can also be a good idea if the pack disproportionately favors the more likely team (if, for instance, everyone thinks a team is the best thing since sliced bread, but they only have an estimated 55% chance to win their game), but here, it’s harder to tell where the break-even point is, and it’s one of the places where the number of players will be important.
Or, if you can find one, participate in a pool that rewards correct upset picks with more points than correct favorite picks. That really makes it fun!
The best way to score an office NCAA Tournament Pool is to award the point value of the seed. If you had the balls to pick a 14 seed over a 3, and the 14 wins, you get 14 points, not 1.
And, my stat of the day, the odds of picking a perfect bracket are 1 in 9,000,000,000,000,000,000,000,000 or nine trillion times one trillion. If you had a million people filling out a different bracket every second, it would take them over 300 years to fill out that many.
Doesn’t that presume each game is basically a coin flip though? But I have some knowledge, like Niagra will not win make it to the Final Four. So I can eliminate all those brackets in which Niagra is going to the Final Four.
This is definitely true. Since they went to a 64-team format, #9 seeds have actualy beaten #8 seeds more often than the reverse, and #7 only have won the first game 60% of the time (see numbers below).
Nitpick: Washington is not in the tourney. You mean Washington State. Don’t get the two confused; their fans don’t appreciate that.
If you’re going to randomly pick winners, it would probably be best to use historical data on the probability of upsets for a given pair of seedings. From the wikipedia page on the tournament, we find that
The #1 seed has beaten the #16 seed all 88 times (100%).
The #2 seed has beaten the #15 seed 84 times (95%).
The #3 seed has beaten the #14 seed 73 times (83%).
The #4 team has beaten the #13 seed 70 times (80%).
The #5 seed has beaten the #12 seed 59 times (67%).
The #6 seed has beaten the #11 seed 61 times (69%).
The #7 team has beaten the #10 seed 53 times (60%).
The #8 team has beaten the #9 seed 41 times (47%).
That’ll get you the first round. After that, I don’t have statistics, but they’re probably out there somewhere.
This is OT, but this thread on office pool strategy reminds me of a hand from a duplicate Bridge tournament I read about some years ago. In knockout duplicate bridge, two teams of four players split up in two tables; team A places players at N-S on table 1, E-W on table 2, and team B vice versa. Both tables are given identical deals, and the play and scoring between the two tables is compared after the hand is played. If team A’s N-S combo on table 1, for example, takes more tricks than team B’s N-S combo on table 2, team A has “won” that hand (it’s a little more complicated than that, but you get the idea).
Anyway, in one hand, each N-S team bid to an obvious grand slam (bid they could take all 13 tricks). The slam depended on whether the East or West opponent held the queen of spades. The basic strategy here is to “count out the hand”, meaning that each N-S combo played their winners in clubs, diamonds, and hearts and observed in which round each opponent was forced to discard out of suit. Using this info, you can determine how many spades each opponent held, and it turned out the missing spades were split 4-3 between East and West.
The south player at table 1 finished play by assuming the queen was in the four-spade hand, a better than 50% gamble. The south player at table 2 made a different decision: He knew the strategy was a common technique, and that the “book” play was to assume the queen was in the longer-spade hand. Therefore he guessed that’s how his opponent would play. and so if he followed this strategy, the best he could hope for was a tie; both tables would log the same score. He therefore assumed the opposite, that the queen was in the three-spade hand; he would either win on a lucky guess, or lose outright. Since the tournament was in the last few hands and his team was trailing, he thought going for the upset was justified. Of course, the queen was in the three-spade hand, a play the south at table 2 would never have attempted if he weren’t deliberately looking for a game strategy to win.
In applying this lesson to the NCAA tourney pool, I believe that with a large enough pool of contestants (since there are usually 5 or 6 early-round upsets, “large enough” is probably >12), all-favorites sheets are doomed to lose because there is always somebody lucky enough to score on a pick that would otherwise be considered foolish. But then again, that’s the nature of gambling…
One of the points that I think is being missed here is that correctly picking upsets in the first two rounds is largely irrelevant. The objective is not to pick the most number of games right, it’s to amass the most amount of points, and the majority of the points come from picking the teams who advance the furthest.
Most pools are based on a system where the total number of points per round is the same. So for example, the first round has 32 games and each correct pick is worth 1 point. The 16 games in the second round are each worth 2 points, and so on.
If you miss half of the picks in the first two rounds but get the rest right, you’ll have a huge points advantage over somebody who gets all of the picks right for the first sixty games and misses the final four (160-128).
If you want to optimize your strategy by picking underdogs, don’t worry about which number 11 seeds can beat a number 6 – figure out which 2, 3, or 4 seed is likely to reach the final four.