ALCS Simulated

by Xeifrank on Oct 12, 2007 1:44 PM EDT

Dodger Sims blog will be running 2000 simulations of each ALCS game to come up with a win expectancy for each game. These results along with the LV Hilton Sports Books win expectancies will be posted in this diary before each game. The simulator was written in Visual Basic and uses advanced sabermetric formulas to determine likely at-bat outcomes. The simulations is a Monte Carlo method for determining win expectancies. Regular season 2007 statistics (park adjusted) are used as data input, and where sample sizes are too small, 2007 ZIPS projections are used. This simulator has been simulating NL West games for the entire 2007 season.

vr, Xeifrank

Tweet Comment 15 comments | 0 recs |

Do you like this story?

Comments

Game #1 Results

LV Hilton Sports Book gives a win probability of 59.51% to the Boston Red Sox.

DodgerSims simulator gives a win probability of 52.45% to the Boston Red Sox.

Good luck.
vr, Xeifrank

by Xeifrank on Oct 12, 2007 1:47 PM EDT reply actions

Re: ALCS Simulated

Funny how every game was favored to the Yankee's and we still beat the crap out of them.

Playoffs! LGT resident kineisologist

by E5 on Oct 12, 2007 1:53 PM EDT reply actions

Re: ALCS Simulated

Yes. Good job! Your team is to be commended on that accomplishment. One more dragon to slay! :)
vr, Xei

by Xeifrank on Oct 12, 2007 2:11 PM EDT up reply actions

Re: ALCS Simulated

Well, why bother even playing then?!

Now the Lord can make you tumble, and the Lord can make you turn, and the Lord can make you overflow... but the Lord can't make you burn

by Turkmenbashi on Oct 12, 2007 2:38 PM EDT reply actions

Re: ALCS Simulated

It predicted 4 winners out of 14 games in the postseason. Let's hope it continues to be wrong. (Then again, Vegas was wrong on most of the games, too)

by Spidey on Oct 12, 2007 4:46 PM EDT reply actions

Benchmark

Vegas is it's benchmark. Good eye to point that out. vr, Xei

by Xeifrank on Oct 12, 2007 6:49 PM EDT up reply actions

Re: Benchmark

Not sure what you mean by that, but the Vegas odds are tilted toward balancing the bets on both sides, and that will work against the Indians whenever they face New York, Chicago, LA, Boston, Detroit or Texas, or any starting pitcher from Japan. It's just like the All-Star Game, basically.

by Jay on Oct 12, 2007 6:58 PM EDT up reply actions

Re: Benchmark

The benchmark to how well the simulator is doing predicting games is to compare it to other simulators doing the same thing (don't know of any) or Vegas Oddsmakers. Both systems not only predict winners, but also give a win probability. If Vegas gives the BoSox a 59% chance of winning, and I give them a 52% chance of winning, and the BoSox win, then Vegas did better than the simulator for that particuliar game, even though we both picked the BoSox to win. If the Tribe win, then the simulator outperformed the Oddsmakers on that one particuliar game. I have been doing this for all NL West games this season and comparing the results of the simulator with the Vegas Oddsmakers. Beating Vegas is an extremely difficult task and the simulator was within 1.12% of doing so this year. The simulator's margin of error is approximately 1%. So it did pretty good for it's first year. I will try to get that number a little lower for the 2008 season. Alot of it depends on the input data used, not so much the code for the simulator. I can also use the simulator for various sabermetrical experiments. For example, I can see if a certain managerial move actually helped or hurt a team's chances of winning. I can determine a team's win probability at any point during a game, similiar to what Fangraphs does, but there numbers don't take into account who is pitching, hitting, what part of the lineup is up etc...

vr, Xei

by Xeifrank on Oct 12, 2007 7:11 PM EDT up reply actions

Re: Benchmark

You are a genius, the Yankees should hire you.

Playoffs! LGT resident kineisologist

by E5 on Oct 12, 2007 10:55 PM EDT up reply actions

Re: Benchmark

Good point about the benchmarking, (it is not necessarily a measure against accuracy). Performance is measured against other predictors. I would expect Vegas to be the best since it reflects collective expectations (if even the larger markets are overweighted).

My point is that once the playoffs begin, most methods of predicting outcomes go out the window. Maybe you have some data about performance of Vegas predictions, but I look at the past several years and think that predictions are poor in the playoffs.

It's a new season and "intangibles" play a big role in outcomes.

by Spidey on Oct 13, 2007 12:58 PM EDT up reply actions

Game #2

Simulator: 51.16% for the Indians

Vegas: 43.48% for the Indians

Quite a gap in predictions today. I'm a little worried, because of the way the Red Sox are playing. But the numbers favor the Indians slightly to win this game. They better! :)

vr, Xei

by Xeifrank on Oct 13, 2007 12:13 PM EDT reply actions

Re: Game #2

Yeah, well, I just sacrificed a goat this morning and - according to my Delphic reading of the entrails - the Tribe is a 57% to 43% favorite to win.

"the most vehement Yankee-hating guy I know" - Jay

by mauichuck on Oct 13, 2007 12:19 PM EDT up reply actions

Re: Game #2

This gives me some hope for tonight, and might help wake some of us from our slumber. LGT has been rather quiet and the Tribe appears to have a bit of a hangover since beating the Yankees.

I suspect many bettors are reading too much into last night's game. On the other hand, I have no idea how we can stop Ortiz and Ramirez.

by Spidey on Oct 13, 2007 1:04 PM EDT up reply actions