Sweet 16 Power Rankings

A nice bonus from the way I do this analysis is that it also gives me a power rating for each team. Here are the Sweet 16 participants in order of team strength. Louisville is only the 4th best team remaining, but due to match-ups, has the 2nd best odds of winning.

Sweet 16 Update

In case you missed the explanation from last week, I've taken the Vegas odds for various events (odds for each team to win the title, odds for each team to advance to the Final 4, and odds for the Sweet 16 match-ups) and used them to determine the strength of each team. This allows me to then calculate the probabilities for each team in each round.

The Vegas lines are generally incredibly accurate because even the slightest mistake can cost the sportsbooks millions of dollars from savvy bettors. Here are the updated numbers for the second weekend:



Not much to analyze here as the numbers really speak for themselves and there aren't really any surprises in the numbers like we saw in some of the earlier rounds. Kansas has about a 47% chance of advancing to their 6th Elite 8 in the past 8 seasons. All but one of the 2 vs 3 match-ups look like near toss-ups. The Tarheels and Cardinals should advance fairly easily into the next round, and UNC is still the clear favorite.

Bracket update - Right side

I've taken the projections from earlier in the week that were based on Vegas odds and updated them for the results thus far. Here's the right side...





Gonzaga benefits from the Illini upset, though they're still looking at a probable UNC match-up in the Sweet 16. No upsets of any kind thus far in the East.

Bracket update - Left side

I've taken the projections from earlier in the week that were based on Vegas odds and updated them for the results thus far. Here's the left side...





The only significant upset on this whole half of the bracket so far is Dayton over West Virginia. It's good to be Kansas.

Calculating what to watch

Thanks to CBS's March Madness on Demand we now have the power to watch any game we want and I intend to make the most of it. I've made a simple formula using the strength of each team, closeness of the match-up, strength of the offenses, and upset chances to calculate the "games to watch" today. It's more an art than a science, but it's a start. 100% represents the "ideal" game.

Morning Games
12:20 EDT (9) Butler vs (8) LSU (56.8%) This game will at least be close and is a bit better than the other 8 vs 9.
12:25 EDT (15) CSU Northridge vs (2) Memphis (36.7%)
12:30 EDT (9) Texas A&M vs (8) BYU (59.3)

Afternoon Games
2:30 EDT (12) Northern Iowa vs (5) Purdue (52.6%)
2:50 EDT (16) Radford vs (1) UNC (26.6%)
2:55 EDT (10) Maryland vs (7) California (54.9%) I'll pass on the 1 vs 16 blowouts, solid evenly-matched game here.
3:00 EDT (16) Chattanooga vs (1) Conecticut (23.1%)

Evening Games
4:55 EDT (13) Mississippi St. vs (4) Washington (53.9%)
7:10 EDT (10) Minnesota vs (7) Texas (57.3%) This is a game I was already looking forward to, should go down to the wire.
7:10 EDT (10) Michigan vs (7) Clemson (54.3%)
7:20 EDT (14) American vs (3) Villanova (33.8%)
7:25 EDT (13) Akron vs (4) Gonzaga (40.0%)

Night Games
9:40 EDT (15) Binghamton vs (2) Duke (25.4%)
9:40 EDT (15) Morgan St. vs (2) Oklahoma (26.9%)
9:50 EDT (11) VCU vs (6) UCLA (66.6%) Possible upset, both teams are underrated. WKU vs UI shouldn't be too close.
9:55 EDT (12) Western Kentucky vs (5) Illinois (37.4%).

Of course, the beauty of MMOD is you can alway switch over for an upset in the making. I'll do this again early tomorrow morning for day 2 of the tournament. I'll also post some fresh NCAA tournament bracket calculations after the first two days of games

Bracket secret #1

March Madness Bracket Secret #1 - Be contrarian
In the previous post I established that you generally want to pick the best team in any given match-up. Now we can talk about the other side of that argument. There is an advantage to be gained by going against the grain with a pick.

Here's a super simplified example. Let's say 10 people are tied coming into the championship game. UNC is playing Memphis and is 60% to win. Which team do you hope you've picked in your bracket? The answer is it depends on the other 9 people. For example, let's say 7 of them picked UNC and 2 picked Memphis:

If you picked UNC
60% of the time they win, you are now in an 8 way tie and have a 1 in 8 shot of winning the tiebreaker (60% * 1/8 = 7.5% chance of winning)
40% of the time UNC loses, and you lose

But, if you picked Memphis
60% of the time they lose, and you lose
40% of the time they win, you are in a 3 way tie and have a 1 in 3 shot (40% * 1/3 = 13.3%)

As you can see picking the less likely result has actually nearly doubled your chances of winning. Of course, your real bracket situation is not nearly this simple, but the same principles apply. Try to identify underrated teams to select, especially in the Final 4 and beyond. I'm not talking about trying to pick the next George Mason, the odds just don't justify it, but considering looking for an unpopular 2 or 3 seed (or even a 1 seed) with a good shot.

This year UNC is a huge favorite, but they're also an extremely popular pick. A team like UConn on the other hand has a really difficult path (thanks mostly to Memphis), but still has a great shot. If you're in a pool that is likely to lean a certain way (a pool full of Pitt fans, or UNC-haters, or whatever else) use that information to your advantage because it is a big advantage. Another great resource is the National Bracket on ESPN that shows the popular picks through out the NCAA basketball tournament.

That's all for now, I'd love to hear your feedback on this, and suggestions for teams that are flying under the radar this year that can help you differentiate from the field.

Bracket fallacy #1

I wanted to take a quick few minutes to point out 1 common bracket fallacy and also give you a positive piece of advice to go with it. I'll have a couple more posts in the vein this week.

Bracket Fallacy #1 - Upsets happen, so you have to pick upsets to win
There are valid reasons to pick upsets, but this isn't one of them. Inevitably you'll hear someone tell you something like "typically at least one 12-seed beats a 5-seed during each tournament" (12s have won 32% of the time against 4s), so you'd better pick a 12 in there somewhere!" If you look at those numbers though, you're twice as likely to be right when picking a 5 as you are when picking a 12. 5s are usually just flat out better basketball teams.

We can actually calculate some odds on this:

Your odds of being right picking all #5s: 68% * 68% * 68% * 68% = 21.4%
Your odds of picking 3 #5s right AND a #12 right: 68% * 68% * 68% * 32% = 10.1%

The reality it's more likely the 5s will sweep this year than it is that you'll pick just the right 12 seed. In fact, what often happens is your 12-seed pick loses while another one wins meaning you've now lost 2 of the 5/12 match-ups. Now there are plenty of cases where a 12 is a valid pick on it's own (see Wisconsin and Arizona), but never pick an upset just for the sake of picking an upset. I see it all the time and you're really shooting yourself in the foot. Predicting that upsets will happen during March Madness and predicting which upsets will happen are completely different things.

New high-quality March Madness on Demand (P2P technology?)

If you're a true college hoops fan I don't need to tell you how great March Madness on Demand has been. MMOD, similar to MLB.TV, is one of those things that while you experience it you can't help but say to yourself "THIS is why we invented all this technology in the first place."

The big news this year is that they are rolling out "high quality" broadcasts that are supposed to be tv-quality or better, and they're doing away the waiting rooms. As far as I know they haven't said this anywhere, but those two changes point to a move to peer-to-peer technology which saves them a ton of bandwidth and gives us a higher quality feed. Just make sure your ISP doesn't mind all that uploading.

Anyway, they have a preview up of the new high-quality version. I recommend heading over there and checking it out. Great stuff.

Using the odds - South

These new odds use Vegas lines for various phases of the tournament to create one unified projection.

No surprise here as the Tarheels more than double their odds of a title. The oddsmakers are obviously well-aware that this team has a lot in common with a great team from last season. While they haven't performed that far ahead of anyone this season, you have to think their ceiling is higher than the other competitors.

OU gets a big bump which, in my opinion, they don't deserve. Oklahoma is a great example of a team that gets overrated due to winning a lot of close games. The few games they did lose are being unfairly excused due to Blake Griffin's absence. They are just not that good this year.

Finally, Gonzaga receives a much-needed downgrade, but still shows a fighting chance of breaking into the Final 4.

Using the odds - East

These new odds use Vegas lines for various phases of the tournament to create one unified projection.

Pitt and 'Nova are the big winners in the update. Vegas has been high on Pitt all year long, so it's no surprise they get a bump up here (plus all that share Memphis lost had to go somewhere). 'Nova is surely getting a boost due to opening up the tournament at the Wachovia Center which the old rankings were, of course, not "aware" of.

I actually think UCLA is being underrated here. Getting through Villanova is no small task but this is an experienced and talented team that could really cause some trouble.

Using the odds - West

These new odds use Vegas lines for various phases of the tournament to create one unified projection.

Memphis gets a monstrous downgrade. The old rankings were clearly overrating Memphis and that error compounded round-by-round. These new numbers look much more reasonable. The West is still a near coin-flip between the Tigers and Huskies. Everyone's least favorite team, the Missouri Tigers, get a modest downgrade in this version. They are now given an 86.9% chance of continuing their 0-for-forever streak on advancing to the Final 4.

Using the odds - Midwest

I've got another version of bracket odds assembled, and these should be the most accurate numbers you'll find anywhere. These are built using the odds from various sportsbooks to determine the strength of each team. Sportsbooks aren't in the habit of losing money, so you can generally trust that their numbers are very accurate. I used odds for the NCAA champion, odds for the champion from each region, and odds for first rounds games together to create one unified projection. A few of the opening round games don't have moneylines out yet, so those may need some tweaking laster this week.

Here's the Midwest region:


MSU gets a bump, and WVU is downgraded as compared to the old version. My Jayhawks get a modest bump, most of it due to the downgrade of the Mountaineers.

Update coming soon

I'm working on another version of bracket projections using the current Vegas lines for various things (1st round spreads, odds for each team to win their region, odds for each team to win the championship).

I'm actually tweaking the KenPom rankings to get the percentages to line up with those 3 sets of data points (roughly, it won't hit all 3 exactly of course). From there any match-ups that aren't among those data points can be calculated.

It should be pretty cool, and it should definitely be the most accurate solution thus far.

Breaking down the bracket - South

This is the last region. I will be doing more in-depth analysis of the brackets throughout the week, as well as updates as the tournament goes on, so take a second to bookmark this page or subscribe via RSS.

UNC is the obvious favorite out of the region, but KenPom's numbers have Gonzaga posing an extremely serious threat. Gonzaga and Memphis have a lot in common. Both have a resume full of blowouts against less-competitive teams. It seems undeniable that the system is overrating these teams to a degree, but I believe both are more dangerous than they're given credit for. The 'Zags boast 2 wins over Tennessee, as well as victories against Oklahoma St, Washington St, and an OT loss to UConn.

Essentially 5 different teams have at least a 10% chance to advance out of this region: UNC, Gonzaga, Syracuse, ASU, and (nearly) OU. I have the West region, with UConn and Memphis as the most likely to produce the eventual champion, but this region has by far the most depth.

Illinois is a great team for a 5 seed, but they're staring at potential match-ups with Gonzaga, then Syraucse, then UNC. You can pretty much write them off at this point. They'd need a near-miracle to escape this region.

The first round games all look very lopsided here. I suspect my bracket will be all chalk for round 1 in the South. That's all for now, but check back soon for more breakdowns. Hope you enjoyed it.

These charts show an analysis of each team's chance of advancing to each round. The numbers are based purely on the rankings provided by Ken Pomeroy and a powerful little formula called log5 which converts those rankings into percents for a given match-up.

Breaking down the bracket - East

Nothing is guaranteed for teams in the East bracket. There are several teams with a fighting chance of coming out victorious.

Pitt, Duke, and UCLA are all top 10 teams in KenPom's rankings. 'Nova is a major factor as well especially playing their first 2 games in Philly at the Wachovia Center which isn't accounted for in these numbers.

Florida State as a 5 is actually the underdog here to the 12th-seeded Wisconsin. This bracket is my pick to see the most upsets. Wisconsin and Minnesota have great chances in round 1, UCLA looks solid in round 2, and either Duke or UCLA presents a major threat against Pitt.

These charts show an analysis of each team's chance of advancing to each round. The numbers are based purely on the rankings provided by Ken Pomeroy and a powerful little formula called log5 which converts those rankings into percents for a given match-up.

Breaking down the bracket - West

This is where it gets interesting. KenPom's formula loves Memphis. In some ways it's not hard to see why. Memphis hasn't lost in almost 3 full months. In that time they've won 25 straight by an average of over 19 pts/gm. On the other hand, Conference USA doesn't offer a whole lot of challenge.

On the whole, I think Memphis is overrated by these numbers, but underrated by almost everyone else. They haven't played many tough games, but they've been downright dominant. I absolutely think the Tigers are the favorite in this region.

Aside from Memphis over UConn, the only projected "upset" in this bracket is 5-seed Purdue over 4-seed Washington by a fairly narrow margin. BYU is a very tough 8 seed. They should move through the first round easily and then cause some serious headaches for top-seeded UConn.

Chime in if you think Memphis doesn't have a 45% chance of winning this region.

These charts show an analysis of each team's chance of advancing to each round. The numbers are based purely on the rankings provided by Ken Pomeroy and a powerful little formula called log5 which converts those rankings into percents for a given match-up.

Breaking down the bracket - Midwest

The selection show is over and the analysis has started. I've never been a big fan of Digger Phelps and his "eye" test, or Dickie V's loud ridiculous opinions. On the other hand, it's hard to find analysis that I consider to be any better. Hopefully I can fill that void a bit.

The following charts show an analysis of each team's chance of advancing to each round. The numbers are based purely on the rankings provided by Ken Pomeroy and a powerful little formula called log5 which converts those rankings into percents for a given match-up. Now, KenPom is not flawless, but he's extremely good at college basketball. Better than any analyst or other computer ranking. I hope to do some projections using other sources (including Vegas lines) later in the week.

Let's get down to it. Here's my chart for the Midwest bracket:


A couple of things stand out here. First, West Virginia shows up as the most likely to escape the Midwest after the 1 seed Louisville. WVU has been knocked around to the tune of 8 losses in the Big East (11 losses overall), but they've had a knack for hanging in there against teams like Louisville and Pitt and notched a win over UConn while dominating many of the other good Big East squads.

Looking at the first round, I don't believe Arizona should've been in, but they appear likely to cause trouble for the 5th-seeded Utah. Similarly, USC looks like a favorite against 7th-seeded BC in the opening round.

As a Kansas fan I think North Dakota St. is actually pretty tough for a 14, and 'm disappointed to see West Virginia as our 6 seed, but beyond that I can't really complain.

Overall, I have no real qualms with the way the numbers came out on this one. Let me know if you disagree though.

Welcome

Hey everyone, welcome to Building a Better Bracket. In future years, this space will be used to post my pre-Selection Sunday bracket predictions. I've made some mathematical innovations in the science of bracket-building that I believe will give me the ability to out-predict the other "bracket experts" out there.

For this year, the field has already been announced, so I'll be posting some probabilities and projections for the tournament.