One Run games and the run line

[DougOzzzz]

I performed an analysis of data from 1960-2003 to see if there is any correlation between a team's record in 1 run games the first half of the season and the second half. The results were fairly straight forward. There is none. A team's 2nd half record in 1 run games can be predicted solely by their current won-loss record and their pythagorean record (calculated by their RS-RA), with the later counting twice as much as the first. Adding in a team's record in 1 run games helps very little. In fact, the coefficient for a team's 1st half record in 1 run games is actually negative. This is probably because a team's actual record, when compared to their pythagorean record, does a pretty good job of predicting a team's record in 1-run games. For instance, if a team is winning games at a rate much higher than their pythagorean record would predict, they are probably winning most of their 1 run games.

So I tried ignoring a team's actual W-L record and ran a regression analysis based on their pythagorean records and their record in 1-run games. A team's 2nd half winning pct. in 1 run games can be expressed with this formula: 0.33+.327*Pythagorean record + .0127*record in 1 run games. With a coefficient of .0127, a team's record in 1 run games halfway through the season adds very little to their pythagorean record when projecting their record in 1 run games the 2nd half of the season.

So, to wrap that up - there's virtually no value in looking at a team's season record + or - 1.5 runs when evaluating their chances against the run line. Their current record (or better yet, pythagorean record) is a much better predictor of their future success vs. the run line.

I also checked to see if I could estimate the percent of games a team wins and loses by 1 run based on which team is home and the expected winning percentage (with RPG included). Additionally, I estimated the percentage of games that could be expected to be 1 run games based on the expected mean total runs, and the expected run difference.

The 1st formula for the home team is: 0.1725+.6975*WPct+.0068*RPG
The 1st formula for the road team is: 0.1307+.7339*WPct-.0089*RPG
The 2nd formula is the following: 0.521445-.02407*RPG-.0103*RDIF

Using these formulas you can estimate the relative value of the run line compared to the money line.

For example, the line on NYN at WAS tomorrow for WAS is -149 and -1.5 +143. If the ML is considered fair, then WAS should win 59.84% of the time. The over-under is set at 8.5 runs with considerably more juice on the under (-132). We'll estimate that the mean runs scored for the game will be 9 even (the median is less than that, of course). The expected difference in runs should be about 0.96 runs (actually less, but my formulas do not adjust for home/road advantage).

Thus, the % of 1 run games should be: .521445-9*.02407-.96*.0103= .295 or 29.5%
Washington should win: .1725+.5984*.6975+9*.0068 = .651 or 65.1% of the 1 run games.
Washington should win by 1 run: .295*.651= .192 or 19.2% of the time.
Their expected winning percentage against the run line then is 59.84-19.2 = 40.64%. Unfortunately, they need to win by 2+ at least 41.15% of the time to make the run line bet a better value than the money line bet. So if you're going to bet Washington, it's probably best to take the money line.

[mrbaseball]

[ QUOTE ]
there's virtually no value in looking at a team's season record + or - 1.5 runs when evaluating their chances against the run line. Their current record (or better yet, pythagorean record) is a much better predictor of their future success vs. the run line.


[/ QUOTE ]

It's really pretty simple. Good teams win by one more often and bad teams lose by one more often. And I will continue to use the records (both actual and +-1.5 runs) because they are pertinant. If a team covers the runline 40% of the time up until today I'm not gonna make a wager tommorrow where the break even is 42% since historically they have covered it only 40% of the time.

How often it has happened is important. How often it will happen is unknown. But making decisions of off how often it has happened makes sense. We know one run games happen. Historically we know they happen about 32% of the time. Historically we know winning teams win them more often and losing teams lose them more often. We can rest assured they will continue to happen in this fashion into the future. Maybe even more often as juice is less of a factor and small ball gains respectability in the post steriod era.

I think it's impossible to "predict" a one run game. But knowing how often they happen and in particular with the teams involved (manager style and bullpen are extremely important I think) will keep you on the right side of the arithmetic when making run line wagers.

[calling all cars]

Here's some data I've been using:

HOME FAV ROAD FAV

Win by 2 Runs or More
39.1 % 43.6 %

Win by Exactly 1 Run
18.4 % 11.6 %

Lose by Exactly 1 Run
11.0 % 16.4 %

Lose by 2 Runs or More
31.5 % 28.5

The pertinent data being, of course, the second line. Additionally, I've put together a rough guide comparing the O/U line (assuming it's fair) to the chances of a one run victory by the favorite:

home favorites
o/u 7.5=20%
o/u 8=19%
o/u 8.5=18.5%
o/u 9=18%
o/u 9.5=17.5%
o/u 10=17%
o/u 10.5=16%

road favorites
o/u 7.5=12.5%
o/u 8=12%
o/u 8.5=11.5%
o/u 9=11%
o/u 9.5=10.5%
o/u 10=10%
o/u 10.5=9.5%

Yes, these are numbers I extrapolated out of my imagination--but head over to Tradesports, and you'll find they correlate very nicely with the going prices:

MIL v FLA--Fish are favored, O/U is 7.5, spread between their money line and win line is about 18 points

CWS v TB--Sox are favored, O/U is 9.5, spread is about 16

OAK v TOR--A's favored on the road, O/U is 8.5, spread has sold for as little as 8, but could be had for around 12 probably.