An interesting read from ESPN Insider, more for filling out your brackets but info can be helpful when wagering on these games.
Upset Prediction Value -- Balancing forecasting accuracy with frequency
So how can you tell which long shots are most likely to sneak up on front-runners? Before this question can be addressed, you need to ask yourself one basic question: What kind of a Cinderella spotter do you want to be -- an accurate one, or a prolific prognosticator? You could create an elaborate rule that has never failed to predict an upset, but the rule would almost certainly apply to just a handful of games. By the same token, you could pick every single upset -- if you're willing to be wrong 80 percent of the time -- and end up dead last in your pool.
In determining which factors have the biggest influence on upset prediction, you have to look at how much the rule increases the odds of picking an upset (accuracy) and how many upsets it describes (frequency). Here's an example: Long shots have won 17.3 percent of the 504 games in which an upset could happen in Round 1. These giant killers are rarely 15- or 16-seeds. By eliminating these teams from consideration, you improve your chances of picking an upset by 42 percent -- a 24.6 percent winning rate (83-255) versus 17.3 percent. Just as important, 83 of 87 first-round Cinderellas satisfy this rule. By multiplying the increase (42 percent) above the typical upset rate by the percentage of upsets described (95 percent) we arrive at the upset prediction value (UPV), for the "15- and 16-seed exclusion" rule: an even 40.0. When you evaluate rules by their UPV, you're better able to compare their relative worth in predicting upsets.
General guidelines for spotting Cinderellas
If the deadline for finishing your bracket is approaching, then you may not have time to dig into the nitty gritty of upset probabilities. Still, it's worth knowing a few general guidelines to boost your odds of picking an upset. Remember these three rules and you'll dramatically improve your odds of spotting an upset.
1. Don't pick any upset springers lower than a 12 seed. Yes, 13-, 14- and 15-seeds do pull the occasional upset, but it's at a much lower rate than higher-seeded teams. These long shots are 39-245 against opponents with a seed position of at least four rungs higher than them. That's an upset rate of 13.7 percent. The average underdog wins 20.4 percent of the time -- nearly a 50 percent improvement. And 16 seeds? Forget about them; they're 0-84 in the tourney.
2. Never pick a top seed to be an upset victim. You don't have to look any further back than Kentucky's loss to UAB in the 2004 tourney to know that top seeds can be toppled. But it's not worth predicting. In the 223 games they've played against prospective Cinderellas, one seeds have shattered the slipper 201 times. That's more than a 90 percent success rate -- much too strong to bet against.
If you only applied these two rules to your search for Cinderellas, you'd improve your odds by nearly 57 percent. Underdogs seeded lower than 13 facing all favorites at least four seeds higher, except top seeds, are 116-246. That's a 32.0 percent upset rate and the rules describe nearly two-thirds (116 of 180) of all upsets. Taken together, these simple seed restriction rules yield a UPV of 36.6. You can do a little bit better if you remember one more guideline:
3. Make sure your upset choice beats opponents by more than five points a game. Potential Cinderellas (after the two rules above) with a regular-season victory margin of more than five points a game are 100-183 against favorites for a 35.3 percent upset rate. That's 73.0 percent better than the typical long shot, and the guideline accounts for about 56 percent of all upsets, for a UPV of 40.5.
Round-by-round rules for predicting upsets
If you're willing to examine potential upset squads more carefully, you can boost your Cinderella-spotting success rate -- to the point where you're right more often than you're wrong. Here what you need to do in each round:
Round 1: Take high-scoring 11-seeds and frontcourt-dominant 12-seeds. Given the 13- to 16-seed exclusion rule, you only need to concern yourself with 11- and 12-seeds in the first round.
• Take 11-seeds that score at least 73 points a game and beat their opponents by seven or more points. These teams are 18-17.
• Take 12-seeds with coaches who've been to the tourney at least once before, have a winning percentage higher than .700 and get more than 55 percent of their points from the frontcourt. These 12-seeds are 10-3.
By applying these two rules, you'd be 28-20, or 58.3 percent accurate, in predicting first-round upsets. That's 237.9 percent better than the typical first-round upset rate of 17.3 percent, and it describes more than 32 percent of the upsets.
Upset percentage increase: 237.9 percent | Upsets described: 32.2 percent | UPV: 76.6
Qualifying 2006 long shots: No. 11 seeds San Diego State and Wisconsin-Milwaukee; No. 12 seeds Utah State and Montana.
Round 2: Pick battled-scarred 10- to 12-seeds. While ninth-seeded UAB's 2004 bracket-busting upset of top-seeded Kentucky may still be fresh in the minds of tourney pool players, there's still no reason to pick a ninth seed to topple a top seed in Round 2. They're just 3-43 against the big dogs. That leaves seventh, eight, 10th, 11th and 12th seeds. Here are the rules for these seeds in their various matchups:
• In 1 vs. 8 games, take battle-tested eighth seeds (less than a .675 winning record) who were in the previous year's dance, with coaches making at least their third tourney appearance. They don't come around very often, but these teams are 7-5.
• In 2 vs. 7 games, take seventh seeds from the Big Six conferences (ACC, Big East, Big Ten, Big 12, SEC and Pac-10) who beat opponents by at least six points a game. These teams are 7-6.
• In 2 vs. 10 games, take 10-seeds with at least four junior or senior starters. They're an astounding 8-2 against their favored opponents.
• In 3 vs. 11 games, take 11-seeds that have won fewer than nine of their last 10 pretourney games and have a winning record either below .700 or above .800. Teams satisfying these conditions are 6-3.
• In 4 vs. 12 games, take 12-seeds coming into the tourney on either a losing streak or a single-game winning streak that beat opponents by less than 10 points a game. They're a perfect 6-0 against fourth seeds.
If you applied all these rules, then you'd accurately predict 34 upsets in 50 games for a 68 percent success rate. That's 166 percent better than the typical 25.5 percent Round 2 upset rate, and it describes about 56 percent of the upsets. Upset percentage increase: 166.3 percent | Upsets described: 56.7 percent | UPV: 94.3
Qualifying 2006 long shots: No. 7 seeds Marquette and Georgetown; No. 8 seeds Kentucky and Arizona; No. 10 seeds North Carolina State and Northern Iowa; No. 11 seed Southern Illinois; No. 12 seeds Utah State and Texas A&M.
Sweet 16: Avoid Nos. 1 and 2 seed victims; pick tested No. 7 to 12 seeds with solid guards. Seeding is the key indicator of Sweet 16 Cinderellas. Avoid underdogs that face No. 1 or 2 seeds and any long shot seeded higher than 12th. All other potential upset pullers are 14-18. If you further restricted your upset choices to teams that went to the tourney the previous year and have won at least six of their last 10 pretourney games, then you'd improve your prediction success to 10-3 (0.769). That's more than three times the typical Sweet 16 upset rate -- and it describes 40 percent of the upsets.
Upset percentage increase: 226.2 percent | Upsets described: 40.0 percent | UPV: 90.5
Qualifying 2006 long shots only if they don't face No. 1 or No. 2 seeds: No. 8 seeds Kentucky, George Washington and Arizona; No. 9 seeds UAB and Bucknell; No. 10 seed Alabama; No. 11 seeds Southern Illinois and Milwaukee-Wisconsin; No. 12 seeds Utah State and Montana.
Elite Eight: Pick Nos. 6 to 8 seed rookie squads. Picking upsets in the Elite Eight is risky business. Only 4 of 26 long shots have snuck up on favorites. No. 8 seed Villanova beat No. 2 seed North Carolina in 1985; 11-seed LSU knocked off top-seeded Kentucky the following year; Rick Pitino's sixth-seeded Providence Friars beat No. 1 Georgetown the next year; and Michigan's sixth-seeded Fab Five was the last team to spring a Round 4 upset when it beat Ohio State in 1992. Other than LSU, no team seeded lower than eighth has won in the quarterfinals (the rest are 0-10). So restrict your upset choices to sixth, seventh and eighth seeds. Then focus on teams that didn't make the tourney the previous year. It's a small sample of underdogs, but they're 2-2 in the Elite Eight, while the rest are 2-20.
Upset percentage increase: 224.7 percent | Upsets described: 50.0 percent | UPV: 112.3
Qualifying 2006 long shots: No. 6 seed Indiana; all the No. 7 seeds; No. 8 seed Arkansas.
Final Four and championship: Pick Nos. 5 to 8 seed long shots led by coaching legends. Two teams account for the four upsets that have occurred in the Final Four and championship rounds. Eighth-seeded Villanova beat second-seeded Memphis State, then shocked Georgetown on its way to the championship in 1985. And sixth-seeded Kansas toppled second-seeded Duke and top-seeded Oklahoma to win the tourney in 1988. The other six teams seeded fifth or lower have all lost in the last two rounds of the tournament. What did Villanova and Kansas have that they didn't? Simply put, experienced coaches with strong programs. Both schools had legendary coaches with at least seven tourney appearances (Rollie Massimino and Larry Brown) who had taken their teams to the Big Dance at least four straight years, longer than the impact of any single class. Upset percentage increase: 250.0 percent | Upsets described: 100.0 percent | UPV: 250.0
Qualifying 2006 long shots: No. 5 Syracuse; No. 6 Michigan State; No. 8 Kentucky and Arizona.
Putting all the upset rules together
If you followed all of the rules listed above for each round of the tourney, then you would pick upsets correctly in 78 of 119 games, for a 65.5 percent success rate. That's a 221 percent improvement over the typical upset rate of 20.4 percent. It describes 78 (or 43.3 percent) of the 180 upsets that have occurred in the 64- and 65-team era. That works out to an impressive 95.7 UPV. Let's face it: If you could identify 78 Cinderellas (an average of nearly four per tourney), then you'd be the envy of your fellow bracketeers, whether you won your pool or not.
So how can you tell which long shots are most likely to sneak up on front-runners? Before this question can be addressed, you need to ask yourself one basic question: What kind of a Cinderella spotter do you want to be -- an accurate one, or a prolific prognosticator? You could create an elaborate rule that has never failed to predict an upset, but the rule would almost certainly apply to just a handful of games. By the same token, you could pick every single upset -- if you're willing to be wrong 80 percent of the time -- and end up dead last in your pool.
In determining which factors have the biggest influence on upset prediction, you have to look at how much the rule increases the odds of picking an upset (accuracy) and how many upsets it describes (frequency). Here's an example: Long shots have won 17.3 percent of the 504 games in which an upset could happen in Round 1. These giant killers are rarely 15- or 16-seeds. By eliminating these teams from consideration, you improve your chances of picking an upset by 42 percent -- a 24.6 percent winning rate (83-255) versus 17.3 percent. Just as important, 83 of 87 first-round Cinderellas satisfy this rule. By multiplying the increase (42 percent) above the typical upset rate by the percentage of upsets described (95 percent) we arrive at the upset prediction value (UPV), for the "15- and 16-seed exclusion" rule: an even 40.0. When you evaluate rules by their UPV, you're better able to compare their relative worth in predicting upsets.
General guidelines for spotting Cinderellas
If the deadline for finishing your bracket is approaching, then you may not have time to dig into the nitty gritty of upset probabilities. Still, it's worth knowing a few general guidelines to boost your odds of picking an upset. Remember these three rules and you'll dramatically improve your odds of spotting an upset.
1. Don't pick any upset springers lower than a 12 seed. Yes, 13-, 14- and 15-seeds do pull the occasional upset, but it's at a much lower rate than higher-seeded teams. These long shots are 39-245 against opponents with a seed position of at least four rungs higher than them. That's an upset rate of 13.7 percent. The average underdog wins 20.4 percent of the time -- nearly a 50 percent improvement. And 16 seeds? Forget about them; they're 0-84 in the tourney.
2. Never pick a top seed to be an upset victim. You don't have to look any further back than Kentucky's loss to UAB in the 2004 tourney to know that top seeds can be toppled. But it's not worth predicting. In the 223 games they've played against prospective Cinderellas, one seeds have shattered the slipper 201 times. That's more than a 90 percent success rate -- much too strong to bet against.
If you only applied these two rules to your search for Cinderellas, you'd improve your odds by nearly 57 percent. Underdogs seeded lower than 13 facing all favorites at least four seeds higher, except top seeds, are 116-246. That's a 32.0 percent upset rate and the rules describe nearly two-thirds (116 of 180) of all upsets. Taken together, these simple seed restriction rules yield a UPV of 36.6. You can do a little bit better if you remember one more guideline:
3. Make sure your upset choice beats opponents by more than five points a game. Potential Cinderellas (after the two rules above) with a regular-season victory margin of more than five points a game are 100-183 against favorites for a 35.3 percent upset rate. That's 73.0 percent better than the typical long shot, and the guideline accounts for about 56 percent of all upsets, for a UPV of 40.5.
Round-by-round rules for predicting upsets
If you're willing to examine potential upset squads more carefully, you can boost your Cinderella-spotting success rate -- to the point where you're right more often than you're wrong. Here what you need to do in each round:
Round 1: Take high-scoring 11-seeds and frontcourt-dominant 12-seeds. Given the 13- to 16-seed exclusion rule, you only need to concern yourself with 11- and 12-seeds in the first round.
• Take 11-seeds that score at least 73 points a game and beat their opponents by seven or more points. These teams are 18-17.
• Take 12-seeds with coaches who've been to the tourney at least once before, have a winning percentage higher than .700 and get more than 55 percent of their points from the frontcourt. These 12-seeds are 10-3.
By applying these two rules, you'd be 28-20, or 58.3 percent accurate, in predicting first-round upsets. That's 237.9 percent better than the typical first-round upset rate of 17.3 percent, and it describes more than 32 percent of the upsets.
Upset percentage increase: 237.9 percent | Upsets described: 32.2 percent | UPV: 76.6
Qualifying 2006 long shots: No. 11 seeds San Diego State and Wisconsin-Milwaukee; No. 12 seeds Utah State and Montana.
Round 2: Pick battled-scarred 10- to 12-seeds. While ninth-seeded UAB's 2004 bracket-busting upset of top-seeded Kentucky may still be fresh in the minds of tourney pool players, there's still no reason to pick a ninth seed to topple a top seed in Round 2. They're just 3-43 against the big dogs. That leaves seventh, eight, 10th, 11th and 12th seeds. Here are the rules for these seeds in their various matchups:
• In 1 vs. 8 games, take battle-tested eighth seeds (less than a .675 winning record) who were in the previous year's dance, with coaches making at least their third tourney appearance. They don't come around very often, but these teams are 7-5.
• In 2 vs. 7 games, take seventh seeds from the Big Six conferences (ACC, Big East, Big Ten, Big 12, SEC and Pac-10) who beat opponents by at least six points a game. These teams are 7-6.
• In 2 vs. 10 games, take 10-seeds with at least four junior or senior starters. They're an astounding 8-2 against their favored opponents.
• In 3 vs. 11 games, take 11-seeds that have won fewer than nine of their last 10 pretourney games and have a winning record either below .700 or above .800. Teams satisfying these conditions are 6-3.
• In 4 vs. 12 games, take 12-seeds coming into the tourney on either a losing streak or a single-game winning streak that beat opponents by less than 10 points a game. They're a perfect 6-0 against fourth seeds.
If you applied all these rules, then you'd accurately predict 34 upsets in 50 games for a 68 percent success rate. That's 166 percent better than the typical 25.5 percent Round 2 upset rate, and it describes about 56 percent of the upsets. Upset percentage increase: 166.3 percent | Upsets described: 56.7 percent | UPV: 94.3
Qualifying 2006 long shots: No. 7 seeds Marquette and Georgetown; No. 8 seeds Kentucky and Arizona; No. 10 seeds North Carolina State and Northern Iowa; No. 11 seed Southern Illinois; No. 12 seeds Utah State and Texas A&M.
Sweet 16: Avoid Nos. 1 and 2 seed victims; pick tested No. 7 to 12 seeds with solid guards. Seeding is the key indicator of Sweet 16 Cinderellas. Avoid underdogs that face No. 1 or 2 seeds and any long shot seeded higher than 12th. All other potential upset pullers are 14-18. If you further restricted your upset choices to teams that went to the tourney the previous year and have won at least six of their last 10 pretourney games, then you'd improve your prediction success to 10-3 (0.769). That's more than three times the typical Sweet 16 upset rate -- and it describes 40 percent of the upsets.
Upset percentage increase: 226.2 percent | Upsets described: 40.0 percent | UPV: 90.5
Qualifying 2006 long shots only if they don't face No. 1 or No. 2 seeds: No. 8 seeds Kentucky, George Washington and Arizona; No. 9 seeds UAB and Bucknell; No. 10 seed Alabama; No. 11 seeds Southern Illinois and Milwaukee-Wisconsin; No. 12 seeds Utah State and Montana.
Elite Eight: Pick Nos. 6 to 8 seed rookie squads. Picking upsets in the Elite Eight is risky business. Only 4 of 26 long shots have snuck up on favorites. No. 8 seed Villanova beat No. 2 seed North Carolina in 1985; 11-seed LSU knocked off top-seeded Kentucky the following year; Rick Pitino's sixth-seeded Providence Friars beat No. 1 Georgetown the next year; and Michigan's sixth-seeded Fab Five was the last team to spring a Round 4 upset when it beat Ohio State in 1992. Other than LSU, no team seeded lower than eighth has won in the quarterfinals (the rest are 0-10). So restrict your upset choices to sixth, seventh and eighth seeds. Then focus on teams that didn't make the tourney the previous year. It's a small sample of underdogs, but they're 2-2 in the Elite Eight, while the rest are 2-20.
Upset percentage increase: 224.7 percent | Upsets described: 50.0 percent | UPV: 112.3
Qualifying 2006 long shots: No. 6 seed Indiana; all the No. 7 seeds; No. 8 seed Arkansas.
Final Four and championship: Pick Nos. 5 to 8 seed long shots led by coaching legends. Two teams account for the four upsets that have occurred in the Final Four and championship rounds. Eighth-seeded Villanova beat second-seeded Memphis State, then shocked Georgetown on its way to the championship in 1985. And sixth-seeded Kansas toppled second-seeded Duke and top-seeded Oklahoma to win the tourney in 1988. The other six teams seeded fifth or lower have all lost in the last two rounds of the tournament. What did Villanova and Kansas have that they didn't? Simply put, experienced coaches with strong programs. Both schools had legendary coaches with at least seven tourney appearances (Rollie Massimino and Larry Brown) who had taken their teams to the Big Dance at least four straight years, longer than the impact of any single class. Upset percentage increase: 250.0 percent | Upsets described: 100.0 percent | UPV: 250.0
Qualifying 2006 long shots: No. 5 Syracuse; No. 6 Michigan State; No. 8 Kentucky and Arizona.
Putting all the upset rules together
If you followed all of the rules listed above for each round of the tourney, then you would pick upsets correctly in 78 of 119 games, for a 65.5 percent success rate. That's a 221 percent improvement over the typical upset rate of 20.4 percent. It describes 78 (or 43.3 percent) of the 180 upsets that have occurred in the 64- and 65-team era. That works out to an impressive 95.7 UPV. Let's face it: If you could identify 78 Cinderellas (an average of nearly four per tourney), then you'd be the envy of your fellow bracketeers, whether you won your pool or not.
