Re: The New WCHA IV (2013-14): Unlocked and Cocked
Well, let's look at it. Bemidji's final three games are a home game against UAH (almost assuredly a win, especially after losing tonight) and two road games against Bowling Green. BG is really hung up right now, and a split seems pretty likely. Going 2-1 takes you to 11-13-4, 26 points. Now look at LSSU: they are now 12-13-0, but they are unlikely to win a game for the rest of the season. But for what it's worth, say that they get the split tomorrow night or pick off Ferris State next weekend. That would make them 13-15-0, also 26 points. But the Beavers have the 3-1-0 season series win, so they win any tiebreaker with the Lakers. Bemidji may slump to end the season, but the fact that they get UAH another time gives them a pretty decent shot at the playoffs, because the other team on the 8/9 line has its final games against the two best teams in the league.
It will be hard for the Lakers to get ahead of the Beavers in points. Their max is 15-13-0, but that's pretty unlikely. In fact, a probabilistic model has the Lakers sweeping the Bulldogs around 0.02% of the time, which is as close to zero as you can get. It has about a 50% chance of a split against Ferris, and it has an 83% shot of a split with Mankato. If we're optimistic, the Lakers get to 14-14-0. (Getting to 14-13-1 is 0.03% getting a tie against Ferris, 6.42% against the Mavs.) But if you saw the Lakers tonight, you'd really doubt their ability to sweep the rest of their games. The most likely scenario is 13-15-0, and the Beavers can easily get to 26 points.
I have not built a model that says how often a certain result set happens, that is 10,000 runs of what happens in the five games in Week 24. I'm going to try to get there by Tuesday, but I have school stuff to do. I really wish that I could spend all day tomorrow and Sunday on it because I really want to do it.
As for the Siouxsports thing, I could probably do something like that. My goal for next season is something that self-iterates throughout the season, using some sort of regression model from the 2013-14 KRACH to an estimate for the 2014-15 numbers, and then letting the model iterate based on the probabilities from each run. It's ambitious, but I've been meaning to learn Python anyway.
GFM
Originally posted by gmann
View Post
It will be hard for the Lakers to get ahead of the Beavers in points. Their max is 15-13-0, but that's pretty unlikely. In fact, a probabilistic model has the Lakers sweeping the Bulldogs around 0.02% of the time, which is as close to zero as you can get. It has about a 50% chance of a split against Ferris, and it has an 83% shot of a split with Mankato. If we're optimistic, the Lakers get to 14-14-0. (Getting to 14-13-1 is 0.03% getting a tie against Ferris, 6.42% against the Mavs.) But if you saw the Lakers tonight, you'd really doubt their ability to sweep the rest of their games. The most likely scenario is 13-15-0, and the Beavers can easily get to 26 points.
I have not built a model that says how often a certain result set happens, that is 10,000 runs of what happens in the five games in Week 24. I'm going to try to get there by Tuesday, but I have school stuff to do. I really wish that I could spend all day tomorrow and Sunday on it because I really want to do it.
As for the Siouxsports thing, I could probably do something like that. My goal for next season is something that self-iterates throughout the season, using some sort of regression model from the 2013-14 KRACH to an estimate for the 2014-15 numbers, and then letting the model iterate based on the probabilities from each run. It's ambitious, but I've been meaning to learn Python anyway.
GFM
Comment