Twins Simulation... Tigers vs Twins

by Xeifrank on Oct 5, 2009 1:15 AM EDT Comment 9 comments

I used my simulator to simulate 50000 baseball games between the Tigers and Twins using what to the best of my knowledge will be close to today's starting lineups. The simulator outputs a win probability for each team, along with the average total runs scored, a distribution matrix of all the final scores and how often they occurred, and an averaged box score tally of all games combined. You may find it strange that there are so many one run games listed as the most likely score, and that the home team is always favored to win in the top couple of scores. This is due to basic math and the way that baseball rules play out of having the home team bat last. Here is a great article explaining this phenomenon. I will be simulating all playoff games and posting results on their respective SB Nation blogs.

Today's Results... (Last simulation ran at 1000 PST)

Visitors	Home	Pitching Matchup	Favorite	Vegas Win Prob	Simulator Win Prob	AccuScore
DET	MIN	R.Porcello vs S.Baker	MIN	61.24%	60.32%	TBD

Skinny: Simulator Fun Facts... Nearly perfect agreement between the Vegas lines and my simulator on tuesday's winner take all AL Central finale. The tables are now turned on the Twins, who have gone from longshots to favorites. It's nervous time for Twins fans.

Top 20 Most Likely Scores
1	MIN 4-3
2	MIN 3-2
3	MIN 5-4
4	MIN 6-5
5	DET 4-3
6	MIN 4-2
7	MIN 2-1
8	DET 5-4
9	MIN 5-3
10	DET 3-2
11	MIN 5-2
12	MIN 3-1
13	MIN 6-3
14	DET 5-3
15	MIN 7-6
16	DET 4-2
17	MIN 6-4
18	MIN 4-1
19	DET 6-4
20	DET 6-5
Top 5 Most Likely Shutout Scores
1	MIN 3-0
2	MIN 2-0
3	MIN 1-0
4	MIN 4-0
5	MIN 5-0

Game Pitching Results
Pitcher	WHIP	FIP
R.Porcello	1.482	4.320
S.Baker	1.201	3.639

Note: Keep in mind I did this simulation before knowing the actual starting lineups, so some of the players I used may not be starting. Many of the "Players Most Likely To" stats depend on having the correct lineup. The lineups I used are listed below. Chances are you won't see a big difference in win probability from having the lineup order slightly off. Picking the correct starters is more important, especially if one of the better players is taking the day off. But you still won't see a big change in win probability if the starter and backup are interchangeable.

Simulation Lineups
1	C.Granderson	D.Span
2	P.Polanco	O.Cabrera
3	M.Ordonez	J.Mauer
4	M.Cabrera	J.Kubel
5	A.Huff	M.Cuddyer
6	C.Guillen	D.Young
7	B.Inge	B.Harris
8	G.Laird	M.Tolbert
9	A.Everett	N.Punto

0 recs | Comment 9 comments

Story-email Email Printer Print

Comments

Niiiice

Cool little toy you made there.

Fuck tOSU

by ajs1122 on Oct 5, 2009 1:28 AM EDT reply actions 0 recs

Twins' lineup change

Jose Morales has been starting at DH against righties, so I’d probably pencil him in instead of Harris, still batting seventh.

Excellent work.

"There are only two things that are infinite, the universe and human stupidity, and I'm not sure about the former." - Albert Einstein

by BeefMaster on Oct 5, 2009 10:44 AM EDT reply actions 0 recs

Gomez

I’d like to see Gomez get the nod. Neither he nor Morales have been hitting this month. But having him and Span in the outfield with Kubel as DH gives us our best defensive alignment.

by Jon Kammerer on Oct 5, 2009 11:38 AM EDT reply actions 0 recs

+1

I guarantee that the win probability would go up with Gomez in the lineup.

I just don’t get it, Gardy understands how important that defense is at the end of the game, as he always puts Gomez in. Doesn’t he understand that it would be important at the beginning also? Morales is a better hitter than Gomez, but nowhere near to the degree that Gomez is a better fielder than Kubel.

by lookatthosetwins on Oct 5, 2009 2:51 PM EDT up reply actions 0 recs

And then...

…if Delmon gets on late in the game you have no pinch-runner / defensive replacement for him.

Amused observer of the old ladies backyard laundry gab.

by Johnny Safron on Oct 6, 2009 7:24 PM EDT up reply actions 0 recs

Thanks again Xei,

What do you think about Accuscore? Do you think your simulator is more accurate?

by lookatthosetwins on Oct 5, 2009 2:53 PM EDT reply actions 0 recs

It is hard to tell without comparing over a very long period of time (perhaps a couple years worth of data). I did some comparisons last year when I first started my simulator. AccuScore did better the first half of the year, but once I improved my model and improved my projections I did better than AccuScore. The problem with AccuScore is that they don’t take a lot of stuff into consideration and leave the fine tuning to its users. My simulator beat Vegas odds pretty handily in backtesting, but the lines have gotten tougher. So much tougher that now I pretty much break even against the juice. I am still working on things. I think I have the model working well (not perfect), with room for improvement with the starting pitcher and hitters projections. For example, what is the best way to weight pitcher/hitter projections. How many games, months, years back in the data do you go? Then how much weight do you give to the most recent games. Backtesting with empirical data is the best way to determine this and that’s what I am going to be working on in the offseason. And on the other hand you have to be careful of “over-fitting”. But to answer your question, yes I think my simulator is more accurate than AccuScore, but both have room for improvement.
vr, Xei

by Xeifrank on Oct 5, 2009 4:29 PM EDT up reply actions 0 recs