I’m pleasantly surprised by the amount of clamoring for “non-traditional” Bracket content. Perhaps the tide is turning and the masses are finally growing sick of the garbage they are being fed by the TV people…
Anyway, this intro will be brief, as there’s very little mystery this year as to how approach “optimal” strategy. There’s nothing especially fun about filling out brackets in this manner, and in an ideal world, I’d fill just one bracket, which would be strictly based on my own opinions and biases. But alas, I want to win money, so I pick what is likeliest to yield high finishes in pools. After four months of grinding this sport every single day, digging through endless spreadsheets, and forming and un-forming overall opinions on teams, I take all of the information I’ve acquired throughout the season and…dump it in the trash. The bracket is a math problem, not a basketball problem.
As always, everything is dependent upon pool size, entry fee, and skill level of your pool. I’m in a few $250 and $500 contests where I already know that I can throw in the garbage the publicly-available numbers from Yahoo/CBS, as the player pool in pricier contests is, naturally, exponentially sharper than a free ESPN contest or your wife’s $10 family pool. In those small contests, you can reliably pick almost straight chalk because you know that Uncle Larry is going to pick a 5 seed to win it all and have a 13 seed in his Elite 8.
The model referenced in the below screenshots is from TeamRankings/PoolGenius, but you can plug in any projection model (KP, Torvik, EvanMiya, etc). and get very similar results.
Last season, in standard 50-100-person contests, I entered a bracket with Houston defeating Purdue in the title game, and Alabama joining them in the Final Four. Yeah…done, right? Not quite. This bracket still finished in the top 10 percent everywhere that I entered it, simply by playing an extremely +EV spot with (4) UConn over (1) Kansas in the Sweet 16 and then advancing UConn to the Final 4 (losing to Houston). I knew ahead of time simply by looking at futures odds and metrics that KU would be a very small favorite in a Round 2 game against Arkansas (closed -3) and would be a small dog to UConn in the Sweet 16. Despite this arduous path, KU was still the third-most popular championship pick with roughly 10% of entrants playing the Jayhawks to win the whole damn thing. I also picked up massive ground by advancing (3) Kansas State to the Elite 8. KSU was (easily) the least popular 3 seed pick but projected as a coin flip against (2) Marquette in a hypothetical Sweet 16 game (which never came to fruition). The “field” was heavily overweight on Marquette there, and a sizeable percentage played (6) Kentucky to take out KSU a round earlier, even though the Tang Gang had superior pre-Tourney odds to reach the Sweet 16 (44% vs. 32%).
Point is, with just a few well-calculated shots, you can withstand being wrong about an absolute shitload of stuff elsewhere. Your focus should not be on trying to get everything right, but to differentiate in the highest-leverage spots in the bracket where projections and public sentiment diverge.
This year, current data from PoolGenius shows the following title odds and public pick percentages:
As a preliminary matter, it’s not that you “can’t” take UConn — you certainly can in smaller pools. After all, they’re approximately twice as likely as any other team to cut down the nets. And yet, their chances are still “only” in the 20-25% range depending on your preferred model. In a more mid-sized pool (i.e. 50-100 people), a UConn title pick almost certainly must be paired with a slightly more contrarian title game opponent. The most popular title game matchups are UConn/Purdue (12.8%) and UConn/Houston (10.6%), both of which are too chalky for all but the smallest pools. North Carolina holds the key to unlocking major leverage in this year’s Tournament. Arizona is the clear favorite to win the West, but UNC will be the more popular pick in all but the sharpest, high-stakes contests. Rather than jamming a chalky and grossly over-leveraged UNC into a hypothetical title matchup with UConn, you can pivot to Arizona with less ownership but better odds. This isn’t a trick. You really did just read that sentence correctly. Less ownership + better odds. OK then.
With the big boys out of the way, I’ll throw out a few specific spots in each region worth considering. The size of your pool and the ability of your competition determines how much added risk you need to take on. PoolGenius does a phenomenal job showing the added risk for any hypothetical play you make within the bracket. Before advancing that 9 seed over the 1 seed in the second round, ask yourself if it’s reaaaaally worth taking on the massive risk to do so when you can accomplish nearly the same thing a round or two later with substantially less risk and almost the same reward. Balance. Balance. Balance.
East Region
Outside of very deep contests, I see no reason to even consider eliminating UConn prior to the Elite 8 (at the earliest). You can get approximately the same leverage eliminating them in the Final 4 game.
Auburn is popping like crazy in most predictive models and the Tigers bear some similarities to last year’s UConn as a top 5 overall team who had no business being saddled with a 4 seed. In sharper contests, Auburn over UConn will be a frequent play, but I’m likely staying away from it because I think I can find enough ways to differentiate elsewhere while keeping UConn in until the E8/F4. Alternatively, Auburn is also the 5th-likeliest team to win the whole damn thing, so an Auburn-as-champion bracket would allow you to be extremely chalky elsewhere. Not a bad look for a massive contest (i.e. 1000+ entries).
In the bottom half, I really like BYU over Illinois in what would be lined as a near-PK. I project Illinois -2, but the field will be exceedingly overweight on Illinois. In larger pools where you’re in need of additional differentiation —- but also don’t want to risk eliminating UConn too early — advancing BYU to the Elite 8 pairs well with UConn in the regional final to offset the overwhelming Husky chalk.
Midwest Region
Not nearly as much intrigue here as there is in the East as the top four seeds all have negative leverage in advancing to the Sweet 16. The “bust it wide-open” play in very large pools is probably Gonzaga over Purdue, but the Boilers are essentially tied with Houston as the second-most likely team to win the whole thing. Do you really want to assume a mountain of added risk for a fairly milquetoast reward? In my mid-sized pools, I’m advancing Purdue to at least the championship game, where, when matched up with someone other than UConn or UNC (i.e. Arizona, Iowa State, Auburn, etc.), actually creates a realistic-yet-contrarian title game matchup.
There’s really nothing to be gained from putting Creighton into the E8 over Tennessee. Subjectively — and I’m trying to avoid this infusing this type of analysis into a strictly math-based process — Creighton’s has zero floor due to its absurdly high variance. All things equal — and with nothing to be gained from advancing Creighton over Tennessee — I’ll swallow the chalky 1 vs 2 final here with Purdue and Tennessee in mid-sized pools.
Every year, the pre-play in game data shows “the public” picking the 6 seed (in this case 7 seed) at a laughably high clip. Once a team advances (Colorado State) and it’s no longer “either/or,” the numbers start to move closer to reality, though will likely still fall well short of reflecting the true odds of the Colorado State vs. Texas matchup (project Texas -3).
South Region
This is a fun one. Kentucky is an overseeded 3 whose national profile almost always leads to an ownership popularity “tax,” of sorts, but I’m not inclined to eliminate them until the hypothetical S16 game with Marquette, who is likewise way underwater in Round 2 but might close as a small fave against the ‘Cats a round later. This is similar to the BYU/Illinois question in the East, as Texas Tech is a natural, +EV pivot off of Kentucky in the R32 and again has significant positive leverage a round later in the S16, potentially setting up a contrarian-but-feasible 1 v 6 regional final against Houston for consideration in the deepest of large-field pools.
If you stay chalky in the bottom half here, there’s a little bit of a leverage op with Duke or Wisconsin against Houston in the S16. I’d only consider this route in large-field pools, but it’s a viable contrarian path if you’re keeping the other 1 seeds alive deep into the Tournament.
As I mentioned in the Midwest, the play-in game numbers are way off for now due to the unclear outcome of Boise/Colorado. I’m reasonably sure that Colorado would close as a small fave against the Gators after docking a point or so for the Handlogten injury. It won’t move the needle much, but taking the CU/Boise winner (particularly CU) to beat Florida is a cheap, low-risk way of grabbing an extra point in traditional scoring formats.
West Region
The one you’ve been awaiting. The below table speaks for itself. Every metrics site hates UNC as a one seed and sees them more appropriately seeded on the 3-line. I won’t get into the merits of that discussion, but we do know from recent history that the field will gravitate towards any 1 seed — no matter how dubious — and in particular when the 2 seed is widely-considered “untrustworthy.” So, when to eliminate UNC? The model (any projection model will show the same) starts lighting up as early as R2, but as with a few of the other over-leveraged top seeds, why take on such massive risk so early when you can wait a round or two to achieve the same effect? In most of my entries, I’ll remove them in the E8. Worth mentioning here that in sharper/high-stakes contests, these UNC numbers won’t be in the same stratosphere, as most of those players are also engaged on the betting side and will be very much attuned to current prices in the futures market, where UNC is trading around +350 to make the Final Four.
Should you opt to bounce UNC in the S16, there’s very little difference in whether you advance Saint Mary’s or Alabama. Dealer’s choice.
I’ve seen some support for Baylor as a Final Four team. No issue there, as the key is to get away from UNC in some fashion. In a hypothetical Baylor/Zona S16 game, Zona likely opens in the -4/4.5 range.
Someone out there will go full YOLO with the hope that Michigan State finally lives up to its analytical profile and wins the region. I get it…but it’s something that only makes sense in one of those 2000+ entry contests on DraftKings or something, because if you whiff here, you’re immediately drawing dead. As I mentioned on Twitter, the hypothetical UNC -3/3.5 line against Sparty would knock the socks off of the casuals. The KU/Arkansas comps from last year are very real, no matter which of 8/9 MSU ends up advancing to play UNC.