Tuesday, October 14, 2014

What historical odds say about Orioles', Giants' chances

The Baltimore Orioles are in a tough spot. They're down 2-0 and now the series heads to Kansas City, against a team with all the momentum and seemingly destiny on their side too.

What are the chances they come back in this series?

This exact scenario, overcoming the 2-0 deficit hitting the road happened three times beginning 1985: 1985 (Royals over Cardinals), 1986 (Mets over Red Sox), 1996 (Yankees over Braves). Another spin: it's happened once since 1986. So, not likely.

I examined MLB postseason results from all seven game series', 1985-present, when the NLCS/ALCS round was changed from five games to seven, to help find out some more clues.

Here's  the chance of winning a series when leading ...

1-0: 66.7%
2-0: 85.7%; 42 series, 14 sweeps, 10 times it ends in Game 5, seven times 2-0 team closes in Game 6, five in Game 7. Twice the team down 0-2 came back to win in six, four times in seven. 
2-1: 76.7%
3-0: 94.1%; 14/17 sweeps, two went to six, Red Sox famously came back in 2004.
3-1: 82.2%: these have been hard to close out. The 3-1 leader wins Game 5 just 42.2% of the time. Typically the 3-1 leader has to close it in 6 (61.5% winning percentage in game) or else they'll lose the series in the seventh (70%).
3-2: 66.7%: The 3-2 leader finishes it off 54.9% of the time in Game 6, but rarely comes back from losing that game (26.1%).

The San Francisco Giants are in a different boat than the Orioles. The series is tied 1-1 and the Giants are coming home. In a seven game series, a tie after two is just as likely as one team winning both.

You'd think stealing home field would be an advantage, but historically it hasn't at this stage in the series. The team that lost home field advantage (1-1 series') actually won 25 of 42 times.

Anyway, the winner of Game X wins the series X percent of the time ...

1: 66.7% (see 1-0 lead)
2: 69%
3: 61.9%: In a tie series, the Game 3 winner actually moves on 73.8% of the time. This has typically been the game a 2-0 series leader drops (59.5% of the time).
4: 70.2%
5: 61.4%: In tie series, Game 5 winner moves on 64% of the time.
6: 88.2%
7: 100% (obviously) ... team at home is 17-6 in this game.

Final point: 5 or 7-game series ... The team that gets to three wins in a 7-game series first hangs on 79.8% of the time.

Monday, October 13, 2014

Latest BCS rankings: One Mississippi (State), Two Mississippi

The latest BCS ranking are just like the old counting method: One Mississippi (State), Two Mississippi. First time that's ever happened. 

I'll try to update this each week of the season. You can imagine there will be controversy if the BCS No. 4 doesn't match the selection committee's No. 4 team (the committee's first ranking comes out in two weeks). 

For Sagarin, I used Elo Score, which was the closest to his most recent ranking that he used in the BCS. He has North Dakota State ranked in this; I'm not sure if the others include FCS teams in their rankings. The Harris Poll has been replaced by the AP, which used to be used in the rankings. I got Wolfe's ratings from Massey's comparison web site. 

Wednesday, October 8, 2014

If AP poll members were on the College Football Playoff committee...

The College Football Playoff is new, and there's a lot of mystery behind its process. Instead of voters, we have committee members. The committee will have its own ranking, but it's a multistage process to get there. 

Here's how it works:

A. Each committee member selects its best 25 teams (unordered). Schools have to appear on at least three ballots to be eligible. 
B. Each committee member provides the best six teams (again unordered). The six teams on the most ballots proceed.
C. Each voter then ranks those six teams 1-6. The three teams with the lowest combined totals are ranked in order and the other three are subject to the next round.
D. Repeat B and C until 25 teams are ranked

I decided to test how this process works and used ballots from 13 AP voters. 

My sample: Adam Jude (Seattle Times), Seth Emerson (Macon Telegraph), Jon Wilner (San Jose Mercury News), Brett McMurphy (ESPN), Sam Werner (Pittsburgh Post-Gazette), Chris Murray (Reno Gazette-Journal), Daniel Berk (Arizona Daily Star), Harold Gutmann (Durham Herald-Sun), Kellis Robinett (Wichita Eagle), Steven Sipple (Lincoln Journal Star), Joey Knight (Tampa Bay Times), Josh Kendall (The State (South Carolina)), Grant Ramey (Volquest)

Here's the result from just adding up their ballots, the way polls do it.   

1. Auburn 319
2. Florida State 311
3. Mississippi State 290
4. Ole Miss 284
5. Baylor 272
6. Notre Dame 252
7. Alabama 231
8. TCU 228
9. Arizona 223
10. Oregon 208
11. Oklahoma 203
12. Michigan State 195

Note: Auburn is not first in the actual AP poll (trailing by two points). My sample just favored the Tigers (and the Wildcats) relatively more than the other AP voters. 

All 13 voters included the poll’s top four teams – Florida State, Auburn, Ole Miss and Mississippi State in the top-six, while 10 had Baylor and nine had Notre Dame. Those six schools move on because they were on the most top-six lists,

Using their Top 25s, I can tell the the preference each team has with each voter. So assigning 1-6 values to those six teams, this is how they would have voted, sticking to their ballot:

1. Auburn (19)
2. Florida State (25)
3. Mississippi State (46)
NR Ole Miss (52)
NR Baylor (58)
NR Notre Dame (73) 

Auburn, Florida State and Mississippi State are the 1-3 teams in the playoff if the season were to end this week and these AP voters were committee members. The process then repeats itself. 

In the next round, Ole Miss, Baylor, Notre Dame, Alabama, Arizona and TCU were on the most top-sixes of remaining teams. 

4. Ole Miss (21)
5. Baylor (30)
6. Notre Dame (45)
NR TCU (57)
NR Arizona (59)
NR Alabama (61)

In the third round, Alabama, TCU, Arizona, Oregon, Oklahoma and Michigan State appear on the most ballots.

7. Alabama (31)
8. TCU (36)
9. Arizona (38)
NR Oregon (51)
NR Oklahoma (58)
NR Michigan State (59)

In the last round (for this blog post), Oregon, Oklahoma, Michigan State, Georgia, Texas A&M, UCLA were on the most top-six ballots of remaining schools.  

10. Oregon (24)
11. Oklahoma (28)
12. Michigan State (36)
NR Texas A&M (52)
NR Georgia (61)
NR UCLA (72) 

In short, the Top 12 is the exact same as their AP ballots. 

1. Auburn
2. Florida State
3. Mississippi State
4. Ole Miss
5. Baylor
6. Notre Dame
7. Alabama
8. TCU
9. Arizona
10. Oregon
11. Oklahoma
12. Michigan State

It's interesting that this complex methodology reveals the exact same ranking as just a straight 1-25 poll (though I only checked the Top 12). 

There's two ways this method produces a different result. The first is when there is a close race between two teams and one team has more variability than the other (as in a ranking lower than sixth). 

Hypothetically say Auburn and Florida State are tied for third in a three-person poll. Auburn has 2 third place votes and 1 fourth place vote. Florida State has two second place votes and one seventh place vote. Adding it up, they each have 67 points (or 11 if starting the other way). In this system, Florida State's seventh place vote gets replaced with a sixth place vote and the Seminoles would prevail.

The second way is a voter having a team ranked in the top six that is not one of the six most frequent teams.  Let's say one voter's top six is Auburn, Arizona, Florida State, Ole Miss, Mississippi State and Notre Dame. Arizona is not one of the the six most frequent teams listed in voters' top six, which means every school this voter had 3-6 gains a point on Auburn. 

Tuesday, October 7, 2014

Inaugural Mock BCS Standings: Auburn top team

Just because the BCS doesn't exist anymore, that doesn't mean we can't compute the BCS standings. Thankfully, the much-maligned computer rankings are still around. Jeff Sagarin is probably the most well-known one. Anderson & Hester was released today. Here's Billingsley, the Colley MatrixMassey, and Wolfe websites (although Wolfe's aren't updated, USA Today has an updated version).

One note about Sagarin: USA Today and others are using his overall rating for the BCS (which is a weighted synthesis of three ratings). I'm using Elo Score, one of those three, which was the closest to his Pure Elo used in the BCS. Prior to Pure Elo, Elo Score and may or may not have been included in his BCS standings under the name Elo Chess. I haven't seen Elo Chess or Pure Elo posted this year.

The BCS used the Harris and Coaches Poll and computers to determine the BCS standings. Each was worth one-third. A team's best and worst computer rankings were dropped. The Harris Poll isn't around, so I'm subbing in the AP poll as its replacement (which used to be a part of the BCS).

Wednesday, October 1, 2014

Is the Wild Card round fair? Game 1 LDS winner least likely to move on

MLB's do-or-die Wild Card round is exciting television drama. But is it fair for the participants? After all, teams play 162 games a season, plus spring training, and then the season comes down to one game.

I decided to look at the history behind the Division Series round to see if teams actually come back from 1-0 deficits.  If teams frequently mount comebacks, then there's evidence in favor of lengthening the Wild Card round. Here are the stats for the LDS round, which was reinstated following the 1994 player strike:

Up 1-0: 71.1% of teams win series
Up 2-0: 88.8%
Up 2-1: 75.5%

Winner of Game X moves on Y% of the time

1: (see 1-0 lead)
2: 75% ... Teams that won Game 1 win 59.2% of Game 2s.
3: 73.7% ... In 1-1 series, the Game 3 winner moves on 77.4% of the time.
4: 77.6% ... Team with 2-1 lead finishes the series 53.1% of the time in Game 4.
5: 100% (obviously) ... Teams that trailed 2-1 in series and forced a deciding game Game 5 win 52.2% of the time... Teams have forced 8 Game 5s after trailing 2-0 in the series, winning 5 of them.

What's interesting is the Game 1 winner is the least likely of any game in the series to decide the overall winner, though it's still a heavy favorite.

Just for kicks and giggles, I included each team's overall series record in the LDS round (post-strike).

Angels: 3-3
A's: 1-6
Astros: 2-4
Blue Jays: 0-0
Braves: 6-7
Brewers: 1-1
Cardinals: 9-2
Cubs: 1-3
D-backs: 2-3
Dodgers: 3-4
Giants: 3-3
Indians: 4-2
Mariners: 3-1
Marlins: 2-0
Mets: 3-0
Nationals: 0-1
Orioles: 2-1
Padres: 1-3
Phillies: 3-2
Pirates: 0-1
Rangers: 2-4
Rays: 1-3
Red Sox: 6-4
Reds: 1-2
Rockies: 1-2
Royals: 0-0
Tigers: 4-0
Twins: 1-5
White Sox: 1-2
Yankees: 10-7