RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

[jmhusker]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

I do include a flat 13.45% chance of tie factor (I counted how many ties happened in the last five years or something like that). I'm not rigorously checking every single possibility, because as I and others have noted, that would require >10[sup]17[/sup] simulations. My tiebreaking can't handle anything greater than a three way tie, and if two teams are still tied after the ECAC win tiebreak, I flip a coin and break the ties that way.

EDIT: Outcomes with an "x" rather than a number never occurred in my simulation, but may be possible. I defer to FD08 on tiebreaking and whether something is possible or not.

MORE EDIT: I just don't understand how you're getting us in ninth (alone or as a tie) 69% of the time, which is nearly double what I have. Here is the relevant line from my post in the other thread:

   |  KRACH |     1     2     3     4     5     6     7     8     9    10    11    12 |Avg Rk |AvgPts
-----------------------------------------------------------------------------------------------------
RP |   53.0 |     x   0.0   0.0   0.2   0.5   1.3   3.2   8.5  37.2  48.9   0.1     x |  9.27 | 18.43

Would you mind sharing your code? I can post mine too and we can compare.

While I don't doubt the mathematics here, the question is the model. Using Krach as an estimate assumes a team is as good (or bad) as it's overall record and that we are going to play at that level the rest of the way. Absent any trends or other information that makes sense. In our case, there are a number of factors that might lead you to question the wisdom of declaring this as the most likely set of outcomes. First off, we really stunk it up earlier in the season when playing OOC (especially against Bentley). So is the OOC Krach a good indicator of our team or should you consider ECAC only? Secondly, what are we doing lately. One might argue we are even worse however other than the Union game, it seems like we are still playing hard and staying in games. Third, our team is like a box of chocolates so...

Here's another way to look at this. We are favored to beat Brown (I should hope). If we do, and Cornell or Colgate loses then we are one point out of 8th with 5 games left. Say what you want about that scenario but I'd put money down against 10-1 odds that we can squeak into 8th from there.

 

[lugnut92]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

While I don't doubt the mathematics here, the question is the model. Using Krach as an estimate assumes a team is as good (or bad) as it's overall record and that we are going to play at that level the rest of the way. Absent any trends or other information that makes sense. In our case, there are a number of factors that might lead you to question the wisdom of declaring this as the most likely set of outcomes. First off, we really stunk it up earlier in the season when playing OOC (especially against Bentley). So is the OOC Krach a good indicator of our team or should you consider ECAC only? Secondly, what are we doing lately. One might argue we are even worse however other than the Union game, it seems like we are still playing hard and staying in games. Third, our team is like a box of chocolates so...

Here's another way to look at this. We are favored to beat Brown (I should hope). If we do, and Cornell or Colgate loses then we are one point out of 8th with 5 games left. Say what you want about that scenario but I'd put money down against 10-1 odds that we can squeak into 8th from there.

I don't deny that KRACH has it's problems, but I include a disclaimer listing these issues in every single post in my predictions thread. KRACH is the only (popularly known) model that can be easily used for simulation of outcomes, so it's what I (and most everyone doing something like this) use. I know burgie12 used a model that better accounted for ties, but I don't know how that model worked.

 

[jmhusker]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

SA is an extremely glib and savvy man. he makes a wonderful personal appearance, he speaks incredibly well, and he is a great salesman for our program and our school. He is an A+ recruiter. I suspect he is also honest with the recruiting kids and their parents. I would never want that to change. I think what is needed is a truly offensively oriented and fundamental tactical assistant who can make use of all the talent that SA brings to the program. I hate to point out the obvious, but if we just won faceoffs yesterday, we probably do not lose the game. It was one of the most glaring examples of how poor fundamentals have cost us. There are so many examples of the past few years (being on a 5 on 3 against QU late in the game and giving up a short handed goal). It seems like every game we make a few to many fundamental errors or lapses that other teams just either do not make or make less often.

I've always seen SA as a bit of a snake oil salesman myself. The whole Alfred E Newman grin and the slick hair and the over-hyped confidence from day 1 gave me some reservations. I've watched long enough to know that he is not a good bench coach. And beyond that he is either the most stubborn or stupid or masochistic coach there is when it comes to end of game situations. I've stated my reservations previously about the transfer of Koudys and early departure of Zaliewski (who should be in the ECHL right now since he is getting almost no playing time with the Comets). There's a little more smoke there. In the end, I would not be sad to see him go, even immediately to be frank.

All of that said, I thought this year would be rough and in my mind I'd already targeted next season as his last chance to put up or be shown the door. I'm still fine with that as a guideline if whoever at RPI has the decision making authority here is thinking the same way. If there is one further year of suck and he stays on, then the entire admin of RPI needs to have their heads examined. I sense from things written about the Empress of late, that's probably the situation on the ground already.

 

[jmhusker]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

I don't deny that KRACH has it's problems, but I include a disclaimer listing these issues in every single post in my predictions thread. KRACH is the only (popularly known) model that can be easily used for simulation of outcomes, so it's what I (and most everyone doing something like this) use. I know burgie12 used a model that better accounted for ties, but I don't know how that model worked.

OK let me clarify... I don't think the model is wrong. I think stating conclusions in posts based on what the model says needs to have that disclaimer. Beyond that, I suggest we could all agree that projecting Dartmouth and Harvard based on Krach is also a little tricky right now. Conditions on the ground have changed a lot this year. That doesn't always happen but this year just seems wackier than normal, which is wacky enough!

 

[FlagDUDE08]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

I've always seen SA as a bit of a snake oil salesman myself. The whole Alfred E Newman grin and the slick hair and the over-hyped confidence from day 1 gave me some reservations. I've watched long enough to know that he is not a good bench coach. And beyond that he is either the most stubborn or stupid or masochistic coach there is when it comes to end of game situations. I've stated my reservations previously about the transfer of Koudys and early departure of Zaliewski (who should be in the ECHL right now since he is getting almost no playing time with the Comets). There's a little more smoke there. In the end, I would not be sad to see him go, even immediately to be frank.

All of that said, I thought this year would be rough and in my mind I'd already targeted next season as his last chance to put up or be shown the door. I'm still fine with that as a guideline if whoever at RPI has the decision making authority here is thinking the same way. If there is one further year of suck and he stays on, then the entire admin of RPI needs to have their heads examined. I sense from things written about the Empress of late, that's probably the situation on the ground already.

Remember to add into account that the second Appert has the makings of a good season, the administration offers him a huge extension. Obviously the administration isn't quite right in the head, and unfortunately, an oust was attempted with the no confidence vote around 2007, which resulted in the disbanding of the Faculty Senate.

It's worth it to push the coach to resign, or buy out the contract if necessary, and get someone better it. Falling short of that, no more contract extensions unless either a national championship or three consecutive tournament appearances are had. He's had his grace period, and now it's time to produce results.

 

[sk8tronthepond]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

Good question since Fridgen's recruiting went downhill his last couple of years.

I think the continuity of Assistant Coaching changes as well as the pool of highly skilled, academic student athletes shrinking were factors. However, all recruits did fit the academic profile of the Institute.

 

[Waite21]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

My concern is that if he stays he or whoever the new AD is will try to change us into a football/basketball school.

The good thing is that it would save me money.

Ralph, even if a new AD wants to try to build up RPI's football and/or basketball programs, it would make sense to do everything possible to maintain a top-notch hockey program.

At Penn State, which is unquestionably first and foremost a football school, photographers from local newspapers are sent to take pictures of coach James Franklin escorting football recruits to events on campus. And the photographs I've seen have been from the Pegula Ice Arena, where SRO crowds of over 6,000 routinely show up to watch their top-20 ranked team.

I haven't seen any photos of Franklin taking football recruits to the Bryce Jordan Center, where Penn State routinely curtains off the upper deck to conceal the large number of seats that go unsold for basketball games.

Of course, we will know that Penn State hockey has truly arrived if and when the photographers at Pegula Arena ever start taking pictures of action on the ice, rather than in the stands.

 

[Ralph Baer]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

Ralph, even if a new AD wants to try to build up RPI's football and/or basketball programs, it would make sense to do everything possible to maintain a top-notch hockey program.

Sadly, RPI ≠ sense.

 

[Ralph Baer]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

I think the continuity of Assistant Coaching changes as well as the pool of highly skilled, academic student athletes shrinking were factors. However, all recruits did fit the academic profile of the Institute.

There are always a lot of factors, but we supposedly lose out on our top six or seven targets one year.

 

[sk8tronthepond]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

There are always a lot of factors, but we supposedly lose out on our top six or seven targets one year.

Obviously there are a lot of factors, but "supposedly" is not an accurate fact!

 

[DeepRed72]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

I do include a flat 13.45% chance of tie factor (I counted how many ties happened in the last five years or something like that). I'm not rigorously checking every single possibility, because as I and others have noted, that would require >10[sup]17[/sup] simulations. My tiebreaking can't handle anything greater than a three way tie, and if two teams are still tied after the ECAC win tiebreak, I flip a coin and break the ties that way.

EDIT: Outcomes with an "x" rather than a number never occurred in my simulation, but may be possible. I defer to FD08 on tiebreaking and whether something is possible or not.

MORE EDIT: I just don't understand how you're getting us in ninth (alone or as a tie) 69% of the time, which is nearly double what I have. Here is the relevant line from my post in the other thread:

   |  KRACH |     1     2     3     4     5     6     7     8     9    10    11    12 |Avg Rk |AvgPts
-----------------------------------------------------------------------------------------------------
RP |   53.0 |     x   0.0   0.0   0.2   0.5   1.3   3.2   8.5  37.2  48.9   0.1     x |  9.27 | 18.43

Would you mind sharing your code? I can post mine too and we can compare.

pastebin with my python code

Your numbers are different enough from mine to cause me some concern about my model. (by the way which thread did you post those results originally). Mine was, as I first stated, a back of the envelope Excel model with macros. I haven't written robust code in years. And I also stated up front that I did not take tie-breaks into consideration. I will revisit my model in the next few days. Hopefully I can either agree with your numbers or reconfirm mine. I do like yours better however as you show almost a 14% chance of home ice. As for some of the arguments against KRACH given by some in the few preceding posts, what they are overlooking is that it includes our putrid performance out of conference which earns us a rating way below our fellow ECAC teams, including Union. When evaluating a team's strength one can't summarily ignore some of the data because it doesn't fit one's preconceived notion.

 

[FlagDUDE08]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

Your numbers are different enough from mine to cause me some concern about my model. (by the way which thread did you post those results originally). Mine was, as I first stated, a back of the envelope Excel model with macros. I haven't written robust code in years. And I also stated up front that I did not take tie-breaks into consideration. I will revisit my model in the next few days. Hopefully I can either agree with your numbers or reconfirm mine. I do like yours better however as you show almost a 14% chance of home ice. As for some of the arguments against KRACH given by some in the few preceding posts, what they are overlooking is that it includes our putrid performance out of conference which earns us a rating way below our fellow ECAC teams, including Union. When evaluating a team's strength one can't summarily ignore some of the data because it doesn't fit one's preconceived notion.

Take a look in the Projected Standings thread; the same one where I post the possibilities.

 

[Ralph Baer]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

Obviously there are a lot of factors, but "supposedly" is not an accurate fact!

A member of this forum, who will remain nameless, told me about it. What I don't recall is the number. (And I meant "lost" not "lose".) I wrote "supposedly" because the English language doesn't have a good subjunctive case.

 

[DeepRed72]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

Take a look in the Projected Standings thread; the same one where I post the possibilities.

Thanks FD. Yes it occurred to me all I had to do was click on lugnut and I found the Projected Standings thread right away. Anyway I found several more logic and math (mostly copying) errors in my model which I believe are now debugged. I also adjusted my KRACH values to reflect the ones that LN used as well as increasing the tie rate to his assumed 13.45%. Here is a comparison of the outputs (Lugnut 1 million iterations vs. my model 500 iterations) when I split the tiebreaks evenly (Lugnut's probability given first, then mine last):

4th - .002, .001
5th - .005, .001
6th - .013, .019
7th - .032, .028
8th - .085, .060
9th - .372, .357
10th - .489, .533
11th - .001, .001

Since a plurality, if not majority, of the 10th place ties are between us and Union and we win the tiebreak on head-to-head sweep, my 10th place is probably overstated and 9th is understated (remember I split the 9th place ties evenly as I do not have a tiebreak subroutine). That given, I believe my model is now in sync with LN and I bow to the superior intellect. It was not my intention to reinvent the wheel, and not a very round one at that. I was unaware of the other thread. I will now cease and desist in the knowledge that this matter is in good hands. Thanks Lugnut.

 

[fishcore12]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

We finish bottom 4, go on the road and pull a Brown/Colgate on a few teams and end up in Lake Placid. I will be at the bar waiting for you all to buy me drinks.

 

[RPI67]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

Thanks Doc great post. SA salesman recruits great young men and obtained a contract extension with a poor losing record. Let the assistants do everything else. I have a spreadsheet that has the lineups of every game for the last nine years. Continuous change. I did see the UConn and Harvard games in person. I don't question effort just tactics. LGR. Can still get to win first home playoff series under SA.

 

[FlagDUDE08]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

We finish bottom 4, go on the road and pull a Brown/Colgate on a few teams and end up in Lake Placid. I will be at the bar waiting for you all to buy me drinks.

If we can actually win two playoff series, I would actually take you up on that bet. Only because you already owe me one from a few years ago.

 

[sk8tronthepond]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

A member of this forum, who will remain nameless, told me about it. What I don't recall is the number. (And I meant "lost" not "lose".) I wrote "supposedly" because the English language doesn't have a good subjunctive case.

With regards to this type of information, any source other than the coaching staff is not credible in my opinion, even if it comes from a parent!

 

[burgie12]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

I know burgie12 used a model that better accounted for ties, but I don't know how that model worked.

I used the Rutter Method. Slide show from Rutter here. It assumes that each team's ranking is not a single value, but instead a normal distribution centered around a specific value. When comparing two teams, you create a new standard deviation centered at the difference of their means. A tie only occurs within "phi" of 0.

It supposedly does a better job of handling ties because it automatically eliminates the "straight tie" problem. But, it does have its faults. It assumes each team has an equal standard deviation, which is obviously false. It has the ability to handle home-ice advantage, but most simple representations don't.

And, like KRACH, it still suffers from "you are as good as your record says you are," hot streaks, injuries, etc. that are extremely difficult to model.

KRACH is the only (popularly known) model that can be easily used for simulation of outcomes, so it's what I (and most everyone doing something like this) use.

I like CHODR, too. But, haven't taken to actually coding and using it. VBA is my only language currently (especially since I already have the tiebreakers already coded in and I don't want to have to redo it in a different language). Since FD took over the mathematical possibilities and lugnut has been doing an excellent job on the modeling, there's not much demand for a third numbers guesser, anyway.

 

[FreshFish]

Re: RPI Hockey 2014-2015 Part II: Dedicated to Rich Curadi

I thought this year would be rough and in my mind I'd already targeted next season as his last chance to put up or be shown the door. I'm still fine with that as a guideline if whoever at RPI has the decision making authority here is thinking the same way.

Good summary.