Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

[Priceless]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

1.UNH
2. BC
3. BU
4. ME

Good to see the days of the Big Four are done

[Stories]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

I think it's safe to say there are no gimmes in that bunch; even though all of the top teams have winning records against the bottom four.

All these games could go 3. Should be an exciting weekend!

[WDR357]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

Nice choke job by Maine to still end up with home ice

[jjmc85]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

I am sure Jack Parker is behind the switch... he is the center of all HE conspiracy theories

It ended up not mattering (the only three way tie we had would have been seeded the same no matter how you do it), but the thought definitely crossed my mind. Seemed convenient that with the new rules, BU clinched a playoff spot regardless of what happened today.

[TonyTheTiger20]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

All these games could go 3. Should be an exciting weekend!

Given how rare weekend sweeps have been for *any* team this year, I have to agree with you.

[Priceless]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

Nice choke job by Maine to still end up with home ice

How'd your team do?

[Todd]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

Doesn't top-down make more sense? That way you're rewarding the team that has done the best against the other teams they are tied with?

I think "more sense" in this case, as it so often is, is arbitrary. The case can be made for any approach.

For example:

They could just take the initial results of the first tie-break calculation and seed everyone there, all at once. Why throw anyone out and recalculate at all? Aren't they all tied with each other?

If teams within the group are tied (say, the middle two of a four-way tie), do an internal tiebreaker to split those and decide those seeds (which one is 2nd of 4 and 3rd of 4), but not have it effect the others (first would still be first and fourth would still be fourth).

Or...

Cutting from the bottom up means that in the end you're comparing the two strongest teams head-to-head. Why should it matter how the two best teams did against the worst team in a five-way tie? Isn't it more important how they did against each other? (Of course, you risk skewing who is "best" every time you recalculate with a different set of members.)

Or...

Placing from the top down means that the best of the group is first and then the best of the re-compared remainders is next, and so on. (See above re: skewing.)

To demonstrate:

........ A ........ B ......... C ......... D ..... Total
A ..... x ...... 0-3-0 ... 3-0-0 ... 2-0-1 = 5-3-1
B .. 3-0-0 ...... x ...... 1-2-0 ... 1-2-0 = 5-4-0
C .. 0-3-0 ... 2-1-0 ...... x ...... 2-1-0 = 4-5-0
D .. 0-2-1 ... 2-1-0 ... 1-2-0 ...... x .. = 3-5-1

With the all-at-once method, you're done. The seedings are A, B, C, D. That seems fair to me. They're all tied. Their mutual records are taken into consideration. We have seedings and can move on.

----

Using bottom-up, you'd toss out D. Then you'd re-compare the others with A at 3-3-0, B at 4-2-0, and C at 2-4-0. C gets tossed. Now you'd compare A and B and the sweep makes B the top seed and A second. I can see the viscerally-pleasing logic in that, where A and B were the top two - separated only by a half-game in the standings - and B swept A, so they get the nod. Final seedings: B A C D.

----

But what about the fact that the only series that B won was against A, yet now they're the top seed of the group? Whether you win a game by 3-0 or 2-1, it's still two points - so why does a sweep against one team take more precendence than losing two other series more narrowly? Couldn't the test be that A won two series and C won two series? Couldn't the top H2H be A v C (with A getting the nod) and then one-series winners B and D match up, making the seeds A C D B? In that case, A, C, and D keep their relative seedings, but B moves from best to worst.

(What if C and D split 1-1-1 - making their overall records 3-5-1 and 3-4-2, respectively? If we go by series wins, we'd then have A at 2-1-0, C and D at 1-1-1 and B at 1-2-0, breaking the C/D tie with league wins, I guess.)

----

Working top-down, you'd take A first and also take the three wins for B, and three losses for C, out of the mix. B is now 2-4-0, C is 4-2-0 and D is 3-3-0. So C is 2nd and D takes 3rd by virtue of their H2H win over B. Final seedings: A C D B. (See above re: best to worst.)

Does it make "more sense" that the second overall team, who swept the top overall team, is the bottom seed of the four? To Hockey East, this year - and apparently you - it does.

----

What seems clear in this scenario is that A, C, and D should be seeded in that relative order. What is unclear is how to evaluate B. Are they best because they beat, even swept, the otherwise best team (bottom-up)? Are they worst because they couldn't beat either of the two otherwise worst teams (top-down & series)? Or should some weight be given to all sets of results and they're somewhere in the middle (all-at-once)?

----

One of the things that seems counter-intuitive to me is the throwing out of teams and re-breaking the tie - which can lead to different criteria being used when the current criteria are sufficient. In the potential case of UMA/NU/UVM heading into last night, they would go to league wins to split the round-robin 3-3-0 tie, and then compare the remaining two of the teams H2H once a winner (or loser) was pulled from the three. I understand the mechanics of how that works, but once we've gone to league wins, we've determined that H2H is insufficient - so why go back up to H2H record when the teams are not tied at the criterion you just used?

FWIW, in the case of UMA/NU/UVM, which tie-breaker is chosen decides whether NU would have been 1st (bottom-up), 2nd (all at once) or 3rd (top-down) of that group.

----

Personally, I think all-at-once is the fairest method - and the most transparent to the casual observer. Since all of the teams are contributing to the tie, shouldn't they all be considered simultaneously to determine the outcome for all of the teams? Given the choice between only bottom-up and top-down, I'd pick top down because at least the top choice seems correct to me, but I think that's a false choice.

Ultimately, for the purpose of this thread, the concern of some (myself included) is more a matter of taking one method, whatever it is, and sticking with it or announcing the change.

[bottomdweller]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

I think "more sense" in this case, as it so often is, is arbitrary. The case can be made for any approach.

For example:

They could just take the initial results of the first tie-break calculation and seed everyone there, all at once. Why throw anyone out and recalculate at all? Aren't they all tied with each other?

If teams within the group are tied (say, the middle two of a four-way tie), do an internal tiebreaker to split those and decide those seeds (which one is 2nd of 4 and 3rd of 4), but not have it effect the others (first would still be first and fourth would still be fourth).

Or...

Cutting from the bottom up means that in the end you're comparing the two strongest teams head-to-head. Why should it matter how the two best teams did against the worst team in a five-way tie? Isn't it more important how they did against each other? (Of course, you risk skewing who is "best" every time you recalculate with a different set of members.)

Or...

Placing from the top down means that the best of the group is first and then the best of the re-compared remainders is next, and so on. (See above re: skewing.)

To demonstrate:

........ A ........ B ......... C ......... D ..... Total
A ..... x ...... 0-3-0 ... 3-0-0 ... 2-0-1 = 5-3-1
B .. 3-0-0 ...... x ...... 1-2-0 ... 1-2-0 = 5-4-0
C .. 0-3-0 ... 2-1-0 ...... x ...... 2-1-0 = 4-5-0
D .. 0-2-1 ... 2-1-0 ... 1-2-0 ...... x .. = 3-5-1

With the all-at-once method, you're done. The seedings are A, B, C, D. That seems fair to me. They're all tied. Their mutual records are taken into consideration. We have seedings and can move on.

----

Using bottom-up, you'd toss out D. Then you'd re-compare the others with A at 3-3-0, B at 4-2-0, and C at 2-4-0. C gets tossed. Now you'd compare A and B and the sweep makes B the top seed and A second. I can see the viscerally-pleasing logic in that, where A and B were the top two - separated only by a half-game in the standings - and B swept A, so they get the nod. Final seedings: B A C D.

----

But what about the fact that the only series that B won was against A, yet now they're the top seed of the group? Whether you win a game by 3-0 or 2-1, it's still two points - so why does a sweep against one team take more precendence than losing two other series more narrowly? Couldn't the test be that A won two series and C won two series? Couldn't the top H2H be A v C (with A getting the nod) and then one-series winners B and D match up, making the seeds A C D B? In that case, A, C, and D keep their relative seedings, but B moves from best to worst.

(What if C and D split 1-1-1 - making their overall records 3-5-1 and 3-4-2, respectively? If we go by series wins, we'd then have A at 2-1-0, C and D at 1-1-1 and B at 1-2-0, breaking the C/D tie with league wins, I guess.)

----

Working top-down, you'd take A first and also take the three wins for B, and three losses for C, out of the mix. B is now 2-4-0, C is 4-2-0 and D is 3-3-0. So C is 2nd and D takes 3rd by virtue of their H2H win over B. Final seedings: A C D B. (See above re: best to worst.)

Does it make "more sense" that the second overall team, who swept the top overall team, is the bottom seed of the four? To Hockey East, this year - and apparently you - it does.

----

What seems clear in this scenario is that A, C, and D should be seeded in that relative order. What is unclear is how to evaluate B. Are they best because they beat, even swept, the otherwise best team (bottom-up)? Are they worst because they couldn't beat either of the two otherwise worst teams (top-down & series)? Or should some weight be given to all sets of results and they're somewhere in the middle (all-at-once)?

----

One of the things that seems counter-intuitive to me is the throwing out of teams and re-breaking the tie - which can lead to different criteria being used when the current criteria are sufficient. In the potential case of UMA/NU/UVM heading into last night, they would go to league wins to split the round-robin 3-3-0 tie, and then compare the remaining two of the teams H2H once a winner (or loser) was pulled from the three. I understand the mechanics of how that works, but once we've gone to league wins, we've determined that H2H is insufficient - so why go back up to H2H record when the teams are not tied at the criterion you just used?

FWIW, in the case of UMA/NU/UVM, which tie-breaker is chosen decides whether NU would have been 1st (bottom-up), 2nd (all at once) or 3rd (top-down) of that group.

----

Personally, I think all-at-once is the fairest method - and the most transparent to the casual observer. Since all of the teams are contributing to the tie, shouldn't they all be considered simultaneously to determine the outcome for all of the teams? Given the choice between only bottom-up and top-down, I'd pick top down because at least the top choice seems correct to me, but I think that's a false choice.

Ultimately, for the purpose of this thread, the concern of some (myself included) is more a matter of taking one method, whatever it is, and sticking with it or announcing the change.

I stopped reading about halfway through. However adding them all up and seeding that way was not something I thought of and would actually probably be the best metric since, like you said, all the teams are tied.

[Todd]

Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

For the record:

UNH 36 - 36 [1]
BC 35 - 35 [2]
BU 28 - 28 [3]
ME 28 - 28 [4]
--- Home Ice- 28+ (BU/ME/UML) ---
UML 28 - 28 [5]
MC 26 - 26 [6]
UMA 26 - 26 [7]
UVM 25 - 25 [8]
--- Out - 25+ (UVM) ---
NU 24 - 24 [9]
PC 14 - 14 [10]

I can't remember another year where, going into the final night, ... :

6 teams had a shot at the 5th seed.

5 teams had a shot at the 6th seed.

A single team could end up anywhere from 3rd to 8th (or 9th...).

4 of the 5 games had major playoff implications, with wins, ties, and losses in each game all shaping the field differently.

The head-to-head game between the 1 and 2 seeds was arguably the least interesting of the night...

...but the results of the game between historically-woeful Merrimack and long-buried, last-place Providence had people repeatedly checking for updates.

Three teams tied at 24 (and one at 25) would be fighting for their season to continue at all, instead of for the 5th through 7th seeds.

There was the potential for a three-way tie for the final two home ice seeds, involving teams not playing H2H, also involving one game ending in a tie - and have it come to pass.

There was also the potential for a five-way tie at 26 - one point under .500 - and one of those teams would be done for the year.

There was also the potential for the gap between 3rd and 9th to be TWO POINTS (same results we got except NU beats BU and UVM beats UML instead of tying would have resulted in Maine 28, UML 27, and the five-way tie at 26).

Tie-breaker methodology caused multiple USCHO posters to contact the league office for clarification (and be better informed about history of the subject than the league office).

After all was said and done, a team only three points under .500 - and ten points ahead of the team behind them - is done for the year, but TWO teams a single point over .500 got home ice. A four-point spread between 3 and 9.

Remarkable...

[Todd]

Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

Also for the record:

Mr. Souris,

I know that several people have written to you so far re: tie-breakers, but a group of us still have questions remaining.

As recently as two years ago (http://www.hockeyeastonline.com/men/.../wp030608.pdf), your Hockey East official tie-breaker layout had the lowest teams being preferentially eliminated from group ties until eventually the top teams went head-to-head. The four-way tie shown in your PDF (linked above) is a good example of this bottom-up method.

According to recent comments, you now have changed that to a top-down approach.

It appears that the text of the rule hasn't changed, so much as the interpretation. This rule can be interpreted as either bottom-up or top-down and still be followed: "If more than two teams finish in a tie, the same criteria will be applied to reduce the number of teams tied, and then the process will commence again." In fact, "reduce the number of teams" could be alphabetical or using a dartboard and still follow that rule.

So, since you are the person behind both the two-year-old press release and the current comments, you seem like the right person to ask.

At what point did HE change their interpretation to a top-down approach? Was this something that the coaches were made aware of? Were they the ones that asked for it? Has there been any thought given to clarifying the text of the rule quoted above?

Personally, I like top-down better, but there are several folks having discussions that want to know whether their teams are in or out and whether/when the rule is applied differently is a key issue.

Thanks for your time.

- Todd

Todd, we have not changed any of the tiebreaker procedures since I have been working for the league, so I am not sure what you are talking about. I don’t see any language in their about the lower team being separated.

-Pete

Some examples from the 08 PDF:

Three-way tie of BC-BU-PC (Seeding would be PC-BC-BU)
BU has worst record in 3-way
PC beats BC based on head-to-head

Three-way tie of BC-PC-UVM (Seeding would be PC-BC-UVM)
UVM has worst record in 3-way
PC beats BC based on head-to-head

Three-way tie of BU-PC-UVM (Seeding would be BU-PC-UVM)
UVM has worst record in 3-way
BU beats PC based on number of wins

Three-way tie of UML-NU-PC (Seeding would be UML-PC-NU)
NU has worst record in 3-way
UML beats PC based either on number of wins or record vs. UNH (head-to-head tied)

Four-way tie of BC-BU-PC-UVM (Seeding would be PC-BC-BU-UVM)
Vermont has the worst record in 4-way
BU has worst record in remaining 3-way
PC beats BC based on head-to-head

In all of these cases, you tossed out the bottom team - in one case, twice - before doing the H2H at the top.

The reason that this makes a difference is exemplified this year by the potential UMA/NU/UVM three-way.

Since they split the mutual round-robin 3-3-0, you'd move on to league wins, which would have the threesome as UMA, NU, UVM. If you just took the first pass, that would be their order. Since you eliminate from one end or the other, we have an interesting situation where NU has the H2H tb against the team above (UMA), but loses to the team below (UVM).

So if you work from the bottom up, throwing out UVM and then comparing UMA/NU, NU actually ends up atop this tie. If you start from the top down, taking UMA as the top means that NU loses the tb and is third of three.

All three are valid ways of breaking the tie, but NU, for example, can have three different positions depending on if you go bottom-up (1st), all at once (2nd), or top-down (3rd).

Is that clearer?

[crickets, so far]

[Patman]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

Todd... Between this and the overall complexity for a non-computer scientist in itself this is why I haven't had the motivation to generate within season NCAA/playoff odds. The rules in the leagues can be opaque and leads to conflicting results. The 3-way tie is usually the best example as which team you eliminate or choose first will affect the order.

The nature of hockey usually means such scenarios are unlikely... But as we see it's not inconcievable.

[JB]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

Personally, I think all-at-once is the fairest method - and the most transparent to the casual observer. Since all of the teams are contributing to the tie, shouldn't they all be considered simultaneously to determine the outcome for all of the teams? Given the choice between only bottom-up and top-down, I'd pick top down because at least the top choice seems correct to me, but I think that's a false choice.

I think this simultanous tie breaker makes the most sense. You only then fall to a sub level if two team remain tied within the simultanous tie breaker.

100% agree with you.

With H.E. experiencing more parity over the last 5 years I think 3+ way tie breakers are going to be come more common than the true days of the big 4.

[TonyTheTiger20]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

Todd, please keep us posted on Mr. Souris' response. I wonder if the Hockey East office is frantically wondering what the eff to do right now.

[TonyTheTiger20]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

Todd, we have not changed any of the tiebreaker procedures since I have been working for the league, so I am not sure what you are talking about. I don’t see any language in their about the lower team being separated.

-Pete

I didn't know jcarter worked for the league.

[Patman]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2009-10 edition

I think this simultanous tie breaker makes the most sense. You only then fall to a sub level if two team remain tied within the simultanous tie breaker.

100% agree with you.

With H.E. experiencing more parity over the last 5 years I think 3+ way tie breakers are going to be come more common than the true days of the big 4.

I think that's awfully premature... for all the talk of parity the standings, by and large, the "parity" of the league hasn't really borne itself out... what you're seeing in general is parity of all of college hockey.

Anyhow, that's besides the point... seeing 3+ way ties aren't all that common... obviously not impossible but they are often not the norm.

Nevertheless, the critical matter is that the league leaves itself open to lawsuits relating to ambiguity. That's why there hasn't been a response yet. I'll bet you we'll quietly find an updated tie-breaking procedure set in the off-season.