The same thing is said every year when it comes to the MLB All-Star game. “The sharp money is on the under.”

In some ways, this one game is similar to the Super Bowl – a lot of recreational bettors will bet on this game. As in most sports, recreational bettors are more likely to bet the over than under. In these widely-viewed premier events, the mobs can overpower the smart money, and may offer value on one side to the prudent bettor. But is that the case here in the MLB All-Star Game?

A starting point for this analysis might be to review how betting on the under for this game has done in the last 15 years:

 Year NL AL Result 2010 3 1 UNDER 2009 3 4 UNDER 2008 3 4 UNDER 2007 4 5 PUSH 2006 2 3 UNDER 2005 5 7 OVER 2004 4 9 OVER 2003 6 7 OVER 2002 7 7 OVER 2001 1 4 UNDER 2000 3 6 UNDER 1999 1 4 UNDER 1998 8 13 OVER 1997 1 3 UNDER 1996 6 0 UNDER

If you had bet the under for the last fifteen years, you would have gone 9-5-1. This would have shown a moderate profit, but it is hardly convincing because of the small sample size.

If you handicapped this game, what odds would you set for the moneyline and total? One way to do this is to use a base runs approach, developed by Smyth. This methodology is explained in much greater detail here. You estimate how many singles, doubles, triples, homers and walks each team will achieve. You can make this estimate by reviewing MLB statistics for one team’s batters, and the other teams’ pitchers. Armed with those estimates, you plug them into Smyth’s base runs formula.

Because there are so many reserve players, you have to make some concessions to make the data workable. Go with the starting hitters, and assume that they all play the entire game. This ignores all reserves, which greatly simplifies things. Another assumption you have to make is that whatever statistics these batters have achieved up to this point in the season is an accurate measure of how they will do going forward. There is a problem with this: “lucky” players are more likely to be on the All-Star team than players with average or poor luck. There is a risk that this approach will inflate your All-Star hitters’ likely stats. This is offset by the same factor with pitchers; if a pitcher is over-achieving due to luck, he is more likely to make the All-Star team as well. If you concede that both pitchers and hitters have stats that are higher than they are likely to do going forward, you can hope these offset fairly. If you had more time and energy, a better approach might be to use averages for the last 3 years, or a set of projected stats that incorporates past and present form.

One other difficulty you have is that you do not know how long each pitcher will stay on the mound. No pitcher is allowed to go longer than three innings. With twelve pitchers on each roster, this averages to a little over 2/3ds of an inning per pitcher. You can improve this estimate if you know who is starting, or who is likely to pitch longer.

First, consider the NL lineup:

 AB H 2b 3b HR BB CS SB PA Brian McCann 279 87 14 0 14 42 1 2 321 Prince Fielder 305 92 20 1 22 65 0 0 370 Rickie Weeks 350 96 20 2 15 37 2 7 387 Placido Palanco 321 88 11 0 4 29 0 3 350 Jose Reyes 350 124 22 15 3 34 6 30 384 Ryan Braun 306 98 19 3 16 42 4 19 348 Lance Berkman 255 75 12 0 23 58 0 3 313 Matt Kemp 311 101 18 2 22 56 3 24 367 Hunter Pence 333 109 24 2 10 23 0 4 356

Both teams now use a designated hitter. No NL player was voted to that position, so you can arbitrarily select any player you want for that. Hunter Pence had the highest batting average of any player with at least 300 at bats, so he was chosen for this example. If you assume each player had about 4.3 plate appearances, you can estimate how much offense they are likely to contribute against an average pitcher. However, the AL pitchers are anything but average.

Consider the AL pitchers:

 IP K BB H HR Josh Beckett 106 91 31 65 6 Aaron Crow 41.3 41 21 29 4 Gio Gonzalez 109 85 50 85 6 Felix Hernandez 137 134 42 116 8 Brandon League 36.7 25 8 33 1 Chris Perez 31.3 20 15 22 1 David Price 124 122 24 107 10 Mariano Rivera 33 28 6 29 1 James Shields 134.7 132 34 102 13 Jose Valverde 36 35 20 29 3 Justin Verlander 143.3 138 31 95 12 Jered Weaver 131.3 114 30 91 5 CJ Wilson 125 109 39 112 8

If you assume each of twelve pitchers will pitch an equal amount over nine innings, you can reduce them to a single defensive set of statistics:

AL Pitchers vs. Average Hitters for 9 innings

 IP K BB H HR 9 8.56 3.15 7.58 0.60

The runs scored formula requires singles, doubles and triples. Most pitcher stats do not break that down nicely. Another assumption you can make is that non-home run hits made off these pitchers will follow the normal league distribution: 77.4% are singles, 20.5% are doubles, and 2.1% are triples.

With an offensive and defensive rating for each team, we can project the NL’s offensive stats for this game by averaging the projected hitting stats with the projecting pitching stats, and using Smyth’s “base runs” formula, and adjusting for Arizona’s park factor of 1.05. You do not want to use a straight average though. Instead, multiply the base runs of offense by defense, and divide by the conference average to normalize. In this example, the NL has an offensive rating of 5.20 base runs, and the AL has a defensive rating of 3.43 base runs. Using 2010 stats, the average base runs is 4.39. Therefore, the NL is projected to have 5.2 * 3.43 / 4.39 runs, or 4.06 base runs.

You can do a similar exercise with the AL’s projected runs.

Consider their starting lineup:

 Name AB H 2b 3b HR BB CS SB PA Alex Avila 232 68 17 3 10 34 0 3 266 Adrian Gonzalez 348 121 28 3 16 40 0 1 388 Robinson Cano 325 96 21 5 14 21 1 6 346 Alex Rodriguez 297 88 19 0 13 33 1 4 330 Derek Jeter 272 70 10 1 2 23 2 7 295 Jose Bautista 281 93 14 1 28 84 3 5 365 Curtis Granderson 313 87 11 7 25 44 7 15 357 Josh Hamilton 198 58 13 2 10 25 1 4 223 David Ortiz 285 86 22 1 17 41 0 1 326 Juan Miranda 417 107 24 2 15 39 0 1 417

Again using 4.33 plate appearances, this generates an AL offensive base runs of 5.11. This is slightly less than the NL’s rating of 5.20.

And then there is the NL defense: its pitching.

 NL Pitchers IP K BB H HR Heath Bell 37 27 13 31 0 Matt Cain 120.3 101 32 101 7 Tyler Clippard 48.3 61 16 25 7 Roy Halladay 136.3 131 18 123 8 Cole Hamels 124 115 24 96 7 Joel Hanrahan 39.3 33 8 29 1 Jair Jurrjens 104.2 63 25 88 5 Clayton Kershaw 122.7 138 35 97 9 Cliff Lee 129.3 128 27 113 11 Tim Lincecum 117.3 126 44 99 7 Jonny Venters 53 56 20 31 1 Ryan Vogelsong 84.3 68 27 70 7 Brian Wilson 38.7 36 22 33 1

This reduces to:

 IP K BB H HR 9 9.1 2.9 7.7 0.55

This generates a defensive base runs rating of 3.33, also better than the AL’s rating of 3.43.

What does all this work tell you in relation to baseball betting? The NL is projected to score 4.061 runs, versus 3.87 for the AL. Since neither team is truly a home team here, you can use an unadjusted Pythagorean theorem to set the fair moneyline price. The probability of the NL winning should be about (4.06 ^ 2) / (4.06 ^2 + 3.9 ^2), or .524. Despite the NL winning the All-Star game only once in the last ten years, they should be a small favorite in the betting odds, and the fair total is about 8. Although the total is not out for the game yet, it has usually been close to the league scoring average in years past, suggesting an opener of 8.5 to 9.

If you can take the NL at +110, it is worth a bet. Similarly, under 8.5 has value, and under 9.0 is golden.