Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2011-12 edition
Another thing...
Some of you may remember a couple of years ago when the league was very tight 3-9 and there was a lot of discussion here about multi-way ties. (By definition, all ties are "multi". One entity can't be in a tie by itself. What I mean here is "more than two".) One key point of the discussion was an apparent change in how the league was breaking such ties. Precedent was Bottom Up (see below), but that year they were projecting based on Top Down (also see below).
Part of the problem was that the league rule regarding multi-way ties was vague when it came to how to do it. In fact, based on our discussion here, and those I had around the rinks, I contacted the league office and had a lengthy back-and-forth via email, with lots of examples on my end (no... really?), which stopped very suddenly from their end without any further explanation. It turned out that the league had changed their interpretation without changing the text - and without realizing they had done so until it was broken out for them. Right when they realized it, they stopped corresponding.
Since then, Bertagna has hinted acknowledgement at the prior confusion during TV-game interviews and happily noted that they have internally resolved some issues. Note that the league had never actually had a more-than-two-way tie prior to that season. In sum, while all precedent had historically been Bottom Up, they have now apparently switched either to Top Down or All At Once.
In case it comes up, I can use examples from this week's outcomes to illustrate all the scenarios - and why they matter.
With two teams tied, you simply go down the hierarchy of pre-determined priorities.
In the case of Hockey East, they have decided to use the following:
Hockey East said:
For playoff seeding purposes, the following tiebreakers will be used at the conclusion of the regular season
1. Head-to-head results between the tied teams
2. Number of wins in conference play
3. Best record against the first-place team(s), then the second-place team(s), then the third-place team(s), and so on
4. Coin flip
The trouble comes in with the language around more-than-two ties, as seen here:
Hockey East said:
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.
"...reduce the number of teams tied..."? Couldn't be any clearer than that?
If you have three teams tied, as you move through the hierarchy of tie-breakers, you will eventually get to a situation where one team is better than the other two (BC/BU/UML w/ ME 4th), one team is worse than the other two (PC/NU/UMA), or all three teams have different values (BC/BU/UML w/ MC 4th). (The same general process applies for four or more, but three is sufficient, and easier, to demonstrate.)
In a three-way tie at 35 between BC, BU and UML, I see it breaking down as follows:
- Round robin of 2-1-0 results gives all three a 3-3-0 record.
- With one tie each and the same number of points, they'd have the same number of wins (unless BU had two ties which would mean BC 1 over UML 2 on H2H tb and BU 3 with fewer wins).
- Since these would be the top three teams, tie breaking would next compare their record against 4th place team.
- - if ME: UML 2-1-0, BC/BU 1-2-0 UML 1, BU 2, BC 3 (based on BU 2-1-0 tb once UML is placed #1)
- - if MC: BC 2-0-1, BU 2-1-0, UML 1-1-1 and the seeds would fall that way.
In a tie w/ PC, whether H2H or in three-way combo, it works for NU (1-0-2), but doesn't fare well for UMA (0-2-1). If they all tied, both PC and NU would be 2-1-3, UMA 1-3-2. Seeding NU 7, PC 8, UMA 9.
For the first two, the general consensus across all leagues everywhere is to either promote the solo top team or demote the solo bottom team and then start over with a two-way tie between the remining pair. That fits with HE's description.
The tricky situation is what to do when three tied teams are suddenly all different - Scenario 3. Basically, there are three ways to break that tie.
The simplest - and fairest to me - is to simply seed them as they are, All At Once (AAO). That is how I described that split:
- - if MC: BC 2-0-1, BU 2-1-0, UML 1-1-1 and the seeds would fall that way.
However, historically, on
all prior projection occasions (remember, they never had to do it for real until the controversial year), HE had done Bottom Up (BUp). In our demo case, HE would have dropped UML as the lowest team and seeded them 3. Then, starting over with BC/BU, BU would be 1 based on H2H and BC would be 2nd.
In the controversial year, they suddenly switched to Top Down (TDn) in their projections. In our demo case, HE would have promoted BC as the highest team and 1 seed. Starting over with BU/UML, UML would be 2 based on H2H and BU would be 3.
So there you have a stark example of how and why the unspecific "...reduce the number of teams tied..." is insufficient.
AAO seeding would be BC, BU, UML.
BUp seeding would be BU, BC, UML.
TDn seeding would be BC, UML, BU.
Three different seedings. Three different opponents for the 7 seed (each of the three teams could end up #2). No clear description of what's going to happen.
As it turned out, in the one year with a three-way tie, the league kind of lucked out in that one team was 2-1-0 aganist both of the others and the other pair was also a definitive 2-1-0. With a 4-2-0, 3-3-0, 2-4-0 split, all three methods give the same result.
AAO seeds them in the above order: A, B, C.
BUp drops C (2-4-0) and then settles the other two beased on A's 2-1-0 tb to keep them at A, B, C.
TDn promotes A (4-2-0) and then seeds B over C with the 2-1-0 tb. Result, again, A, B, C.
A, B, C was the right result, no matter how you slice it. The problem is that since all outcomes were the same, we still don't know
for sure how the league will split a situation that doesn't break as cleanly - like our demo.
I'm not positive at this point, but I think they described the real tb, when it happened, by listing the AAO records, so that's why I describe the splits as I do.
In the end, it may not matter. We've only had one actual more-than-two in nearly 30 years of league play, so they're rare. That said, we've had at least one two-way almost every year. In fact, we had three pairs of two-way ties in the standings in one season before we had our first three-way tie. However, it could happen again this year, and will probably happen again at some point - even if it's another 25 years.
Here's why it's important
during play, as opposed to after the RS is over:
In our demo, BU could end up either 1, 2, or 3 depending on how the tb happens. It is conceivable that everyone else could be in the locker room and BU could be in OT with a tie putting them into our demo scenario. If they know that BUp prevails, they could clamp down on D and try to ride out the tie for 1st. On the other hand, if they know that TDn would be used, it would make sense to pull the goalie late in OT to go all out for the win - surpassing the tie for 1st instead of ending up 3rd with either a tie
or a loss.
This is reminiscent of a few years ago when BU and NU were fighting it out in the standings for the 8th and final seed. NU had the tb and was a point behind coming into the final night. NU had finished with a win and BU was in OT at UNH. BU knew that a loss would have NU pass them, but a tie would also leave them out of the playoffs on the NU tb. As it turned out, BU scored about half-way through OT for the win, but had the game gone on much longer, it is likely that Parker would have pulled the goalie in OT. He would only have the opportunity to make that informed decision because everyone knew where the tb procedure stood.
Ironically, NU winning that tb was due to their better record against 1-seed BC, so it would seem that, of the two options that year, BC would have preferred to not face NU. However, as a result of BU squeaking ahead into 8th, that year was also the second time in league history that 8 beat 1 in a best-2-of-3 when BU went to Conte and beat BC in three games.
What is also unclear, based on this:
Hockey East said:
3. Best record against the first-place team(s), then the second-place team(s), then the third-place team(s), and so on
... is how the league would resolve our demo if ME and MC were themselves
tied for 4th. Since we've seen that who is 4th changes who is 1-3, that's a real issue.
Do they compare records against the combined pair at 4th? Do they resolve who wins the tb at 4th first?
What if
both the Trio at 1st
and the Pair at 4th had tbs that came down to the 3rd tb (i.e., if ME and MC were 1-1-1 H2H with the same number of wins - not the case, but for the sake of example)? You can't settle 1st until you settle 4th, but you can't settle 4th until you settle 1st. It's recursive. They don't state which gets decided first. Do they then go to records against 6th place?
If so, which tie do you break first, the Trio or the Pair? Because once you break either the Trio or the Pair, you can use that broken tie to break the other one.
Further, in theory, UNH and PC could be tied at 6 and UMA and NU could be tied at 8th. Would that then mean that everyone's record against UVM - who is the one team definitely not making the playoffs - would decide the entire playoff seedings?
My point is that everyone knowing what they're playing for is important.
Guessing shouldn't be involved.
It's about time.