2
$\begingroup$

Let’s say I’d like to estimate the number of runs Team A will score against Team B

Team A (away) vs Team B (home). Starting pitcher for Team B is a lefty named “Ben James”

Team A scores 5.2 runs per game on average

Team A scores 4.3 runs per game on average when AWAY

Team A scores 2.3 runs per game on average against Team B

Team A batters score 6.5 runs per game against “Ben James” on average

Team A batters score 2 runs per game in the specific stadium on average

Etc.

Team B concedes 4.3 runs per game on average

Team B concedes 2.6 runs per game on average againt Team A

Ben James concedes 1.2 runs per game on average

Ben James concedes 3.3 runs per game on average against team A

Ben James concedes 0.5 runs per game on average when playing home.

Etc.

I have a hard time figuring out how to take into account all these different factors in a formula (stats against a lefty, ballpark stats, home/away stats, specific pitcher, etc)

$\endgroup$
5
  • 1
    $\begingroup$ You're not an American, so I'll cut you some slack. But in baseball, teams score "runs", not "points". I am guessing your subject matter expertise in baseball is not very good, which may contribute to your difficulty. $\endgroup$ Mar 17, 2016 at 14:49
  • $\begingroup$ Hi! You are right, I’m Canadian and my main language is French. All Blue Jays games are broadcasted in French so I am truly sorry If I offended you by talking about “points” instead of “runs”. In French we say “point produits” for RBI and “points” for RUNS (not that you care!). When switching language I sometimes use the wrong term. Once again I am very sorry if I misuse a word. $\endgroup$
    – Max
    Mar 17, 2016 at 15:08
  • 1
    $\begingroup$ @MarkL.Stone But don’t get my wrong, I love baseball. It’s not about your nationality or the language you speak. I watched or listened to all Expos games when I was a kid. Same now with the Jays. I know the game pretty well. I am interested in learning how fantasy owners / sport books / sabermetricians can so accurately predict the number of points that will be scored in a game. Cheers. $\endgroup$
    – Max
    Mar 17, 2016 at 15:08
  • $\begingroup$ The guy who created Strat-O-Matic must have a formula that he used to create his player cards. For example, in Strat-O-Matic, many factors are being taken into account: ballpark, opposing pitcher, player offense, etc. Away vs Home is not taken into account. $\endgroup$
    – Max
    Mar 17, 2016 at 15:27
  • $\begingroup$ I recommend you look into multiple regression (as @Marsenau does) and get thoroughly acquainted with overfitting. Baseball player and team performance is a lot less streaky than one might imagine, so I would avoid the ARIMA route. $\endgroup$
    – rolando2
    Jul 29, 2018 at 13:19

1 Answer 1

2
$\begingroup$

You likely will need more data than just averages. Predicting some future outcome relies on past data and trying to estimate how it behaves and then extending that to future values.

I have no idea how Baseball scores behave but I would think that they change a little over time. Teams might go on hot or cold streaks, some key players might get injured, the team might make a trade for a new player. With that in mind, how they played over the last week or so might be more indicative of their upcoming performance than how they have played for the whole season.

With that being said I would assume (again these assumptions are based off little baseball knowledge) that you should look into getting scores for a game by game basis and try treating it as a time series with possible external factors such as 'opponent', 'stadium', etc. To analyze this time series, something like a regression might be applicable or maybe fitting some ARIMA model if there is that time factor.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.