Handicapping & Baseball Math
How are ERA and other statistics computed? How often should a team playing say .600 ball beat a team playing .400 ball? Can a game be "modeled" with available stats? With this season not old enough for trends articles, we take a short look at baseball numbers this week.
Earned run average is figured by dividing earned runs (runs not scored on errors) by innings pitched and multiplying by nine. The result is the average earned runs given up by the pitcher in a nine-inning game. ERA developed during the long period when there were many complete games pitched. There are many arguments, qualms and questions about ERA, but its use persists, and will into the forseeable future. It is a simple measure of pitching effectiveness that is widely understood and easy to compute.
Starters ERA are higher than relief pitchers, mainly because runners on base when a starter leaves the game are charged to him if they score off a relief pitcher. A fairer way might be to charge the pitcher who puts a runner on base with 1/2 earned run and the one who lets him score with 1/2 run.
Another difficulty with ERA is how the short awful outing balloons the number. Consider this sequence of starts:
IP ER "game ERA"
6 1 1.50
5 0 0.00
7 2 2.57
1 7 63.00
The ERA for this pitcher is 4.74 after four starts. Yet he is clearly a better pitcher than that, with three good games out of four. His team should be 3-1 in his starts. If a limit of 9.00 is placed on the ‘game ERA" (which is the ERA for only that game), and the games are averaged, His ERA become 3.26, which seems to more accurately describe this pitcher’s record. We could publish this way, but it would suffer from lack of understanding and acceptance and be a bit more difficult to compute.
Some readers do a modeling with our numbers. If a pitcher is throwing 6.33 innings per game, giving 3.5 runs, they
take that and the bullpen ERA and produce a model of his start, sometimes adjusting for home or road or other factors, then adding in a factor for the bullpen and a factor for unearned runs, to produce and expected yield of runs, then compare it to the opponent’s run scoring average. This is just a skeleton for a model, which could be modified any way the handicapper wants.Other handicappers do no such math procedures. they look at ERA and determine that 3.21, 3.36 and 3.67 are fairly good numbers, and that 4.95, 5.32 and 6.09 and not very good and draw conclusions. Some divide hurlers into two main categories, above-average and below-average, with extremely good or bad pitchers in smaller categories, respectively not to be bet against, or bet on, until their performance changes.
Log5 is method developed to quantify the winning chances of one team v. another. Wins, runs or win percentages can be used, just so a comparable loss, runs yielded or losing percentage is used. Using wins:
win% Team A v. Team B = winsA x lossesB
divided by
(winsA x lossesB)+(lossesA x winsB)
So if New York were 88-59 and Montreal were 70-72
if would be 88*72/(88*72)+(59*70) = 6336/10466 = .605
New York would win about 60% of the time. Translating that percentage into odds, New York should be about a -1.55 favorite in the game.
To get odds as percentage, divide the favorite odds by the odds + 1.00, so for a -1.55 favorite use 1.55 divided by 2.55 = .607 A -3.50 favorite becomes .778, a -1.25 favorite becomes .556, etc.
Log5 works using runs scored and runs given up, and is considered more accurate as a predictor that way. Another possibility that works like wins and losses is using winning percentage and 1.000 minus winning percentage ("losing percentage"). The Log5 method was designed by Bill James and first published in one of the old Baseball Abstracts. There are many way to use it in handicapping, although it was not designed for that purpose.
The run scoring averages we publish are good for more than totals playing. If a team is 9-4 at home v. lefts, getting 6.2 runs per game, that is quite a bit different from the team at 9-4 getting 4.6. In the first case they are probably winning by scoring runs, in the second they are likely just getting good pitching from their own pitchers in those games.