PDA

View Full Version : 2016-2017 WCHA Season Thread



Pages : 1 2 3 4 5 6 7 8 9 10 [11] 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

bigblue_dl
11-13-2016, 10:02 AM
Bemidji State is running away with this, clearly, but by how much? (http://blog.wchaplayoffs.com/2016/11/13/2016-17-week-7-below-rating/) The number surprised me.

GFM

They're doing well, but running away? They're only 2 games up on Tech in the loss column, with 4 games to play.

gfmorris
11-13-2016, 03:18 PM
They're doing well, but running away? They're only 2 games up on Tech in the loss column, with 4 games to play.

My math says that you'd expect Bemidji to win about 3-games-in-4 the rest of the way — to note, that's a winning percentage below what they've been at to-date. That's 13-14 more wins. Call it 14. That's no less than 71 points, right?

To get to 71 points, Tech needs 18 wins (or equivalent) in their 20 remaining games. Either Bemidji has to come back to Earth or Tech/Mankato has to go on an amazing run.

Is 13-14 more wins realistic for the Beavers? Sure, when you're already 9-1-0-0 in league play. Their remaining slate: v. NMU, @ Tech, v. LSSU, @ UAA, @ BG, v. FSU, v. Tech, @ UAH, v. MSU.

Even if you argue that they fall off strongly — going 9-0-0-9 the rest of the way — who are they losing those nine games to? Say Tech and Mankato — that makes those series almost .500 (Mankato has the edge). Figure they lose once at BG after the trip to Anchorage. You still have to find four more losses! Even if Bemidji finishes something like 18-1-0-9, that's still 56 points. Tech would need to go no worse than 13-0-0-7 just to tie that, which is certainly doable for the Huskies (I'd even argue expected!), but you're already playing the game where Bemidji isn't going to play half as well as they have to date.

I don't have any statistical wood to put behind that arrow. Soon, I hope.

GFM

Jimjamesak
11-13-2016, 03:49 PM
Well I think we had 1 win ooc this weekend and three losses. sigh. Congrats to Bemidji on their record so far. I wonder what has happened to mankato. After their sweep of St. Cloud they don't seem to be the same team.
Hey don't blame UAA, we lose every game conference or non-conference...





:(

Jeff
11-13-2016, 04:11 PM
My math says that you'd expect Bemidji to win about 3-games-in-4 the rest of the way — to note, that's a winning percentage below what they've been at to-date. That's 13-14 more wins. Call it 14. That's no less than 71 points, right?

To get to 71 points, Tech needs 18 wins (or equivalent) in their 20 remaining games. Either Bemidji has to come back to Earth or Tech/Mankato has to go on an amazing run.

Is 13-14 more wins realistic for the Beavers? Sure, when you're already 9-1-0-0 in league play. Their remaining slate: v. NMU, @ Tech, v. LSSU, @ UAA, @ BG, v. FSU, v. Tech, @ UAH, v. MSU.

Even if you argue that they fall off strongly — going 9-0-0-9 the rest of the way — who are they losing those nine games to? Say Tech and Mankato — that makes those series almost .500 (Mankato has the edge). Figure they lose once at BG after the trip to Anchorage. You still have to find four more losses! Even if Bemidji finishes something like 18-1-0-9, that's still 56 points. Tech would need to go no worse than 13-0-0-7 just to tie that, which is certainly doable for the Huskies (I'd even argue expected!), but you're already playing the game where Bemidji isn't going to play half as well as they have to date.

I don't have any statistical wood to put behind that arrow. Soon, I hope.

GFM
For someone that doesn't know how to spell their own name you seem to understand the numbers pretty well. Do you prefer MiniTab, SAS, SPCC or JMP?

gfmorris
11-13-2016, 04:14 PM
For someone that doesn't know how to spell their own name you seem to understand the numbers pretty well. Do you prefer MiniTab, SAS, SPCC or JMP?

I see what you did there. ;) My University thrust Minitab upon me. I'm not really using it for this — it's a pretty simple Monte Carlo simulation.

GGFM

bigblue_dl
11-14-2016, 08:20 AM
My math says that you'd expect Bemidji to win about 3-games-in-4 the rest of the way — to note, that's a winning percentage below what they've been at to-date. That's 13-14 more wins. Call it 14. That's no less than 71 points, right?

To get to 71 points, Tech needs 18 wins (or equivalent) in their 20 remaining games. Either Bemidji has to come back to Earth or Tech/Mankato has to go on an amazing run.

Is 13-14 more wins realistic for the Beavers? Sure, when you're already 9-1-0-0 in league play. Their remaining slate: v. NMU, @ Tech, v. LSSU, @ UAA, @ BG, v. FSU, v. Tech, @ UAH, v. MSU.

Even if you argue that they fall off strongly — going 9-0-0-9 the rest of the way — who are they losing those nine games to? Say Tech and Mankato — that makes those series almost .500 (Mankato has the edge). Figure they lose once at BG after the trip to Anchorage. You still have to find four more losses! Even if Bemidji finishes something like 18-1-0-9, that's still 56 points. Tech would need to go no worse than 13-0-0-7 just to tie that, which is certainly doable for the Huskies (I'd even argue expected!), but you're already playing the game where Bemidji isn't going to play half as well as they have to date.

I don't have any statistical wood to put behind that arrow. Soon, I hope.

GFM

I don't need your fancy math to tell me that Tech is behind BSU by 2 games, and play them 4 times. That is not a runaway.

gfmorris
11-14-2016, 08:48 AM
I don't need your fancy math to tell me that Tech is behind BSU by 2 games, and play them 4 times. That is not a runaway.

Well, it's obvious that Tech has their future in their hands with the four games. If I carry the math to the end, I'm willing to bet that Tech, on average, finishes in the top two and sometimes finishes in the top spot. That's entirely possible. I just don't know how likely it is.

GFM

John J. MacInnes
11-14-2016, 08:53 AM
Bemidji State is running away with this, clearly, but by how much? (http://blog.wchaplayoffs.com/2016/11/13/2016-17-week-7-below-rating/) The number surprised me.

GFM

The Huskies also still have two games in hand on the Beavers.

gfmorris
11-14-2016, 11:07 AM
The Huskies also still have two games in hand on the Beavers.

My argument stands. If you presume that the Beavers will win about 3-in-4 (and I do) — whether they be against Tech or not —*it gets very hard for Tech to catch up. Even if Tech beats Bemidji four times, it's still a lot of ground to catch up.

You can argue with the assumptions all you like. This is not a conclusion that I expected when I started thinking about it, because I was all, "Nobody runs away with the league title 10 games into the season / before Thanksgiving." They haven't, but they've got one hell of a start.

GFM

John J. MacInnes
11-14-2016, 11:30 AM
My argument stands. If you presume that the Beavers will win about 3-in-4 (and I do) — whether they be against Tech or not —*it gets very hard for Tech to catch up. Even if Tech beats Bemidji four times, it's still a lot of ground to catch up.

You can argue with the assumptions all you like. This is not a conclusion that I expected when I started thinking about it, because I was all, "Nobody runs away with the league title 10 games into the season / before Thanksgiving." They haven't, but they've got one hell of a start.

GFM

They are off to a great start. And Ferris State made a pretty good argument in 2013-14 that you can get off to a great start and 'run & hide' from the chasing pack. But if Tech matches Bemidji's points this weekend and then sweeps the following week, the gap is 6 points - or 2 wins. It only looks like a huge lead because of the new point system.

Shirtless Guy
11-14-2016, 11:34 AM
They are off to a great start. And Ferris State made a pretty good argument in 2013-14 that you can get off to a great start and 'run & hide' from the chasing pack. But if Tech matches Bemidji's points this weekend and then sweeps the following week, the gap is 6 points - or 2 wins. It only looks like a huge lead because of the new point system.
I agree to some extent. The real question is can Bemidji keep up the scoring defense?

camilo
11-14-2016, 11:45 AM
Well, it's obvious that Tech has their future in their hands with the four games. If I carry the math to the end, I'm willing to bet that Tech, on average, finishes in the top two and sometimes finishes in the top spot. That's entirely possible. I just don't know how likely it is.


My argument stands. If you presume that the Beavers will win about 3-in-4 (and I do) — whether they be against Tech or not —*it gets very hard for Tech to catch up. Even if Tech beats Bemidji four times, it's still a lot of ground to catch up.

Dicey presumption for sure. How are you at investing based on your presumptions and math?

Shirtless Guy
11-14-2016, 01:13 PM
Not that things are in full swing, the Tech Hockey Guide WCHA Web Media Poll is out:
Check out the results of the first poll (http://techhockeyguide.com/news/2016/11/14/wcha-week-7-thg-web-media-poll?utm_source=message%20board&utm_medium=USCHO&utm_campaign=WCHA%20Thread)
Wild guess who's number 1 :P

gfmorris
11-14-2016, 02:40 PM
Dicey presumption for sure. How are you at investing based on your presumptions and math?

If it's dicey, refute it.

I let professionals manage the stock market. Also, these are probabilities. I wouldn't put BSU's chance of being the #1 seed much higher than 60% right now, but I would put them having home ice at something like 95%.

Hey, I'm going to be fine if you all want to come around in mid-February and call me a dunce or something. That's fine.

GFM

camilo
11-14-2016, 11:39 PM
If it's dicey, refute it.

I let professionals manage the stock market. Also, these are probabilities. I wouldn't put BSU's chance of being the #1 seed much higher than 60% right now, but I would put them having home ice at something like 95%.

Hey, I'm going to be fine if you all want to come around in mid-February and call me a dunce or something. That's fine.

GFM

You probably know that most professional management of stocks can pretty much be explained by randomness.

Now that I'm reading you more carefully, I don't care to argue at all because it all seems reasonable to me. Thanks. But the true test is long run beating simplistic predictions or WAGs - sort of like how investment managers should be measured, but that's a losing game for them.

In doing this sort of prediction, what sort of factors have you used. Just some thoughts that come to my mind are: Have you considered whether the future schedule equal in difficulty to the past? Home vs. away balance, and which teams, where (good ones at home, bad ones away or vice versa)? Are potential injuries or other random challenges considered - viruses that can knock a team down for a weekend or two, potential academic issues, etc. How about the relative youth or experience of the team vs. teams they've played so far vs. in the future , and how does that factor statistically? Do experienced teams tend to improve less or more over the season than inexperienced teams?

I have no idea what impact these factors would have. I don't know if you've accounted for factors you can fairly objectively predict or measure (strength of schedule, home vs. away, etc. past vs. future), or have plugged in some sort of probability of random events, so I ask. I don't claim that any of them refute your presumption, but wondering if they or others are factored in.

In the a little knowledge can be dangerous department: I've taken some statistics and simulation courses in college (business school - operations management) about 25 years ago and it was fascinating, and very complicated and difficult for me, even at the extremely simple level I was able to "master". I had pretty good luck with my capstone project, because it involved a process I knew pretty well, and I was able to identify critical factors and gin up probabilities for enough of them to achieve good predictions and model for adjusting the process. I say "gin up" but really I tried to do sampling and probabilities/distributions within the limits of how much time and access I had at the time. A little better than a WAG.

But I can't argue or refute with any degree of expertise in the field, which I assume you have. At this point, I'm a general manager (called CEO) of a small company working shallowly but broadly in the non-profit, for-profit and governmental worlds simultaneously. My primary expertise is tolerating uncertainty and being comfortable making imperfect decisions. That being the culture I work in, I always am skeptical of people who think they can calculate stuff like this but interested in how they do it.

All that said, it's hard to disagree that BSU is the team to beat, but I don't claim to figure that out with Math, but just because they're a darn good team this year and only have a couple of teams in the league that can be said to be "good". It will be very interesting to see how they do in their 4 games with MTU.

Shirtless Guy
11-14-2016, 11:53 PM
Whether you're right or wrong at any given point in any given season won't pursuade me that your model or whatever it is is actually correct. Let's see your modeling beats other guesses year after year to a significant degree. One season success in predicting a league championship can happen if enough people - or even one person - throws darts.

As far as "refute", I just think it's too simplistic, unless I've missed what objective criteria you've used to make the presumption you stated, what factors, and on what weighting or probabilities you're including them. If so, I'm seriously interested, not being argumentative for its own sake.

What I hear you saying is simply that past performance predicts future performance. Question's I'd ask (maybe you have) about your presumption are: Have you considered whether the future schedule equal in difficulty to the past? Home vs. away balance, and which teams, where (good ones at home, bad ones away or vice versa)? Are potential injuries or other random challenges considered - viruses that can knock a team down for a weekend or two, potential academic issues, etc. How about the relative youth or experience of the team vs. teams they've played so far vs. in the future , and how does that factor statistically? Do experienced teams tend to improve less or more over the season than inexperienced teams?

I have no idea what impact these factors would have. I don't know if you've accounted for factors you can fairly objectively predict or measure (strength of schedule, home vs. away, etc. past vs. future), or have plugged in some sort of probability of random events, so I ask. I don't claim that any of them refute your presumption, but think they should be factored in. Maybe your simulation or model or whatever you call it accounts for these or more factors in a rigorous way. If so the dicey-ness factor I stated earlier is reduced! If so, again I'm truly interested and not just trying to be argumentative.

In the a little knowledge can be dangerous department: I've taken some statistics and simulation courses in college (business school - operations management) about 25 years ago and it was fascinating, and very complicated and difficult for me, even at the extremely simple level I was able to "master". I can't argue or refute with any degree of expertise in the field, which I assume you have.

All that said, it's hard to disagree that BSU is the team to beat, but I don't claim to figure that out with Math, but just because they're a darn good team this year and only have a couple of teams in the league that can be said to be "good". It will be very interesting to see how they do in their 4 games with MTU.He's basing this solely on his ELO rankings (adjusted for WCHA games only). ELO is one of the gold standards for odds and is used all over the place, FiveThirtyEight uses it for predicting odds on pro sports and it was created out of predicting winners in international chess.

gfmorris
11-15-2016, 03:10 AM
To answer camilo's basic question: have I applied an Elo-based prediction model week-by-week in a Monte Carlo simulation to the end of the year? No, I haven't. I wish that I'd had the time to pull it together by now, but that hasn't happened. (My work hours suck, but I'll have some free time while watching the Earth go by over the weekend.) It would indeed be valuable.

To put it cleanly, Bemidji has four potential sweeps: UAH, UAA, NMU, FSU. That gives them eight wins. As I noted above, if the Beavers go .500 down the stretch, it's still a situation where Tech has to win 13 in 20 to catch up. Is that possible? Yeah! But four of those games are indeed against Bemidji, and at this point, BELOW/ABOVE would argue that they teams would split those games — and Tech sweeping Bemidji would be something like flipping heads four times in a row (although the probabilities favor Tech with every win).

It's important to note here that BELOW is simply an estimate of teams' strength. Tech is really high because they were really high last year and have played decently well to-date. Bemidji is high because BOOM! Seriously, they're +160 on the year. BELOW is saying, "Hey, we thought this was a .500 team, but really it's a .740 team." Is Bemidji that good? Well, they're playing well above that right now. I'd say that the truth is somewhere between .500 and .950, so .740 seems as good of a fit as any.

That said, Monte Carlo simulations with probabilistic estimates are only as good as their inputs and assumptions. We have a pretty small data set — 140 games for 10 teams — so I'd probably estimate that the error is anywhere between 0-50 points, and again, we're dealing with probabilities —*a 75% chance that Bemidji wins any average game is still a 25% chance that they'll lose. That's about the odds that FiveThirtyEight had on the last election, and the lesser odds proved true. [Note: no desire to discuss politics.]

GFM

beaverhockeyfan
11-15-2016, 08:27 AM
I can say with 98% confidence that Geof's blood pressure rises a little bit every time he has to type:

B-E-M-I-D-J-I
or B-E-A-V-E-R-S
or B-S-U

:p

moose97
11-15-2016, 10:11 AM
So, BSU's pretty good this year? Fun. Hi to all the old timers. Gonna try to catch one of the BSU @ BGSU games, if anyone will be around in January. Sorry to bother everyone - carry on...

giwan
11-15-2016, 10:29 AM
Not that things are in full swing, the Tech Hockey Guide WCHA Web Media Poll is out:
Check out the results of the first poll (http://techhockeyguide.com/news/2016/11/14/wcha-week-7-thg-web-media-poll?utm_source=message%20board&utm_medium=USCHO&utm_campaign=WCHA%20Thread)
Wild guess who's number 1 :P

Classic reason showing why pre-season polls mean squat.