Monthly Archives: March 2015

KRACH predicts the NCAA tournament

One of the neat things about everyone’s favorite college hockey ranking scheme, KRACH, is that it can be easily applied to predict the likelihood of the winner of a game. I took the current KRACH and used to predict each team’s chances of winning it’s opening round game, advancing out of the regional, winning the semifinal at theFrozen Four, and winning the national championship.

The full data is below, but here are a few interesting observations:

  • Though Mankato’s KRACH is only a bit higher than North Dakota’s, the Mavericks are given a much better chance of both emerging from the regional (62%) and winning the national championship (23%) due to significantly weaker competition in the Midwest than in the West.
  • In fact, Mankato is the only 1-seed given a better than 50% chance of emerging from its region.
  • The regions all have strong teams and weaker teams. This same analysis last year featured a region where each team had between a 20-30% of emerging; no region is even close to that parity this year.
  • All four 2 vs. 3 opening round games are very tightly contested with KRACH prediction ranges of only 58-62% vs. 42-38%.


KRACH East Game 1 Game 2 (Region Champ) Game 3 (Frozen four semifinal) Game 4 (National Champ)
386.382 1. Miami (OH) 65.19% 36.49% 18.05% 9.39%
206.325 4. Providence 34.81% 14.15% 4.94% 1.82%
364.441 2. Denver 61.96% 33.38% 16.04% 8.12%
223.772 3. Boston College 38.04% 15.99% 5.86% 2.28%
523.143 1. Mankato 92.36% 62.16% 38.54% 22.89%
43.2923 4. RIT 7.64% 1.13% 0.14% 0.02%
289.538 2. Omaha 57.94% 22.92% 10.93% 4.92%
210.181 3. Harvard 42.06% 13.80% 5.51% 2.06%
487.207 1. North Dakota 69.05% 40.59% 24.86% 13.96%
218.395 4. Quinnipiac 30.95% 12.13% 5.06% 1.87%
399.556 2. Michigan Tech 59.87% 30.61% 17.30% 8.88%
267.776 3. St Cloud St 40.13% 16.67% 7.78% 3.24%
375.473 1. Boston University 66.36% 37.22% 18.57% 9.25%
190.328 4. Yale 33.64% 13.26% 4.49% 1.52%
339.557 2. Minnesota Duluth 58.76% 31.18% 14.79% 7.01%
238.363 3. Minnesota 41.24% 18.34% 7.15% 2.78%

What teams need to do to advance

Looking back at Saturday – what’s still possible, 11 teams are already a lock for the 16 team tournament.

The winner of Atlantic Hockey will definitely take a spot, leaving 4 spots for other teams

The winners of NCHC and WCHA are guaranteed to come from among the 11, so have no additional effect on number of slots.

Only Hockey East can go to a team that is already guaranteed a spot, making the Boston University-Mass.-Lowell game the most important game for most teams hoping to make it at large. If Mass.-Lowell wins, they take another spot, leaving only 3. If BU wins, there are still 4.

Finally, the Big Ten and ECAC tournaments each contain one team that can only advance by winning the tournament and one team that can make it at-large, so they will effectively each be taking a spot, leaving just 1-2.

What each at-large team needs

The bubble team scenarios are all pretty complex given the questions about how many slots will be available for at-large teams

Minnesota’s slim risk comes from losing and having their PWR fall as low as #15. Because Minnesota losing would involve Michigan winning the Big Ten and thus taking another slot, #15 in the PWR would not make it. To avoid that, if they lose, Minnesota would need any of Harvard, Boston University, or St Cloud State to win. If all of them lose, then Minnesota makes it only about 58% of the remaining scenarios, dependent on weird combinations of the outcomes of UND/Denver, Mich Tech/Mankato, and RIT/Mercyhurst.

Bowling Green’s slim opportunity comes from climbing to #15 and having that spot get an at-large bid. So, no low ranked teams can win their conference tournaments and steal a spot, meaning BG needs wins from Minnesota, Harvard, and Boston University. They also needs Michigan Tech to beat Mankato, and some specific combinations of RIT/Mercyhurst, UND/Denver, and Miami/SCSU outcomes.

Yale’s opportunity has gotten a bit larger. They can climb a little higher than BG, as high as #13-14 in PWR, so have a more opportunities even if lower ranked teams win conference tournaments. They seem to need Harvard and Boston University to win. If those two came through, Yale would advance in about 90% of remaining scenarios. If Mankato won, they would advance; if Mich Tech won, they would need more help. A Mich Tech and Denver win seems to get them in, while a Mich Tech and North Dakota win would leave them still needing help. A Mich Tech, North Dakota, and RIT win seems to get them in, while a Mich Tech, North Dakota, and Mercyhurst win leaves them needing a little more help. A Mich Tech, North Dakota, Mercyhurst, St Cloud St, and Michigan win would get them in.

Providence is a quagmire of all the leftovers from the above. Sorry for mailing it in on this one, but the single biggest factor is Minnesota winning which jumps them from 48% to 93% of remaining scenarios. Beyond that, it’s all weird combinations.

Saturday – what’s still possible

Locks for tournament:

  • UND
  • Mankato
  • Denver
  • Boston University
  • Mich Tech
  • Minnesota Duluth
  • Miami
  • Nebraska Omaha
  • Boston College
  • Quinnipiac
  • St Cloud St

Can still make it at-large

  • Bowling Green (about 4%)
  • Minnesota (about 95% if they lose today)
  • Providence (about 48%)
  • Harvard (about 55% if they lose today)
  • Yale (about 23%)

Can only make it by winning conference tournament

  • Colgate
  • Mass.-Lowell
  • Michigan
  • Mercyhurst
  • RIT


  • Vermont
  • St. Lawrence
  • Robert Morris
  • Northeastern
  • Dartmouth
  • Western Michigan
  • New Hampshire
  • Penn State
  • Michigan State
  • Ferris State
  • Ohio State
  • Canisius
  • Wisconsin


Tonight’s headlines – PWR effects of Hockey East games

The bulk of games are tonight, so there will plenty of PWR tables, simulations, blogs, and tweets about PWR implications of the outcomes. I’ll try to give you a preview of some of those in advance.

These claims are a little less certain than my usual articles because they involve a fair amount of human tweaking, tallying, and writing. So, I may have introduced some error, so I’ll use the words “appear to be” a lot.

Like the ECAC, Hockey East has quite a few teams near the bubble. However, the outcomes of their games tonight have far less impact on other teams’ scenarios than do the ECAC teams’.

  • Quinnipiac appears to be able to secure an at-large bid with just one additional win if Boston University and Mass.-Lowell win tonight.
  • Harvard appears to move into the position of needing at least one win to still have a chance at an at-large bid if New Hampshire and Vermont win tonight.

Plus the effects on the playing Hockey East teams themselves, previously discussed:

  • New Hampshire appears to need to win the conference tournament to advance
  • Mass.-Lowell appears to need at least one win to advance, so losing tonight would eliminate them
  • Vermont appears to need at least one win to advance, so losing tonight would eliminate them


Tonight’s headlines – PWR effects of ECAC outcomes

The bulk of games are tonight, so there will plenty of PWR tables, simulations, blogs, and tweets about PWR implications of the outcomes. I’ll try to give you a preview of some of those in advance.

These claims are a little less certain than my usual articles because they involve a fair amount of human tweaking, tallying, and writing. So, I may have introduced some error, so I’ll use the words “appear to be” a lot.

I’m starting with the effects of ECAC outcomes because the large number of bubble teams in the ECAC makes those among the most interesting.

  • Minnesota appears not to be able to make it at-large with 0 wins if Harvard wins today.
  • Boston College appears to be guaranteed an at-large bid if Harvard and St. Lawrence win today.
  • Mass.-Lowell appears not to be able to make it at-large (though could still be alive in their own tournament) if Harvard wins today.
  • Yale appears not to be able to make the tournament if Quinnipiac and Colgate win today.
  • Vermont appears not to be able to make it at-large (though could still be alive in their own tournament) if Harvard and Colgate win today.

Plus the effects on the playing ECAC teams themselves, previously discussed:

  • St. Lawrence appears to need to win its conference tournament to get a bid
  • Colgate appears to need at least one win to make the tournament, so would be eliminated with a loss tonight


Friday morning’s PWR headlines today

There are two Big Ten games today, so while people will be poring over PWR tables and running simulations tonight, I’ll predict their headlines now:

[OSU or Penn St] and [Michigan or UW] advance in first round of Big Ten tournament, knocking [Penn St or OSU] and [UW or Michigan] out of NCAA contention. Share of remaining scenarios in which Minnesota advances with one win [cut in half/double].

That is pretty much it. The primary tournament effects appear to be that the teams that lose are eliminated, which you probably already knew. The possibilities of other teams advancing in certain scenarios will shift a few percentage points here and there, with the only significant shift being Minnesota’s chance of advancing with 1 win going from 35% as of now to 50-55% if Michigan wins or 16-18% if Wisconsin wins.

No teams other than those playing today seem to have any potential scenarios closed off as a result of today’s action.

Unlikely scenarios and how they might happen

How Yale could make the NCAA tournament

Here’s an example:

The keys are:

  • A large number of conference tournaments being won by top teams. In this case 4 conference tournaments go to top teams so the top 14 in PWR will make the NCAA tournament.
  • Enough teams lose for Yale to climb to #14. In this case Quinnipiac, Bowling Green, Colgate, Minnesota, and Mass.-Lowell all fall below Yale. Passing five instead of four is needed to allow Vermont to pass Yale so it can absorb the autobid.

That set of conditions come together in only about 4% of scenarios to get Yale in.

How Boston College could miss the NCAA tournament

Here’s an example:

The keys are:

  • A large number of conference tournaments won by low ranked teams to take up more slots. In this example, all but 2 conference tournaments go to lower ranked teams so only the top 12 in PWR will make the NCAA tournament.
  • Move idle BC’s PWR down as much as possible. In this case, BC falls from #9 to #13 by losing the comparison to Bowling Green, St Cloud, and Colgate, to fall from 49 comparisons won to 46 (see BC’s PWR). It does so on a combination of their RPIs rising just a hair and a BC’s falling just a hair.
  • The intersection between the above two is what makes this scenario very unlikely — teams that pass BC must do so without winning the conference tournament so lower ranked teams can still take up spots. Looking back at all the possibilities, only a small number of teams have the potential to climb to #12 or higher without winning their conference tournaments. In this case, St Cloud St and Bowling Green are able to flip their comparisons with BC without too much success because they’re already so close in RPI to BC.

That set of conditions come together in only about .8% of scenarios to keep Boston College out.

How St Cloud St could make the NCAA tournament without any more wins

Here’s an example:

The keys are:

  • As discussed in Odds and Ends, SCSU is perilously close to a .500 record so must take advantage of the possibility of a tie in the NCHC consolation game to end up with exactly a .500 record without winning another game.
  • Have only 3 conference tournaments go to lower ranked teams so #13 makes the NCAA tournament.

That set of conditions come together in about 40% of the scenarios in which SCSU loses the semifinal but goes on to tie in the consolation game.

How Hockey East could send five teams to the NCAA tournament

Here’s an example:

The keys are:

  • In addition to BU (which makes it for sure), we need four of Mass.-Lowell, Vermont, Boston College, Providence, and New Hampshire.
  • It’s easy to get BC and idle Providence highly ranked (that’s two).
  • Because Vermont and Mass.-Lowell are long shots who play each other, we can really only get one highly ranked (one more).
  • Give New Hampshire the conference championship (outbid for one more), but also give 3 conference championships to highly ranked teams so the HE teams we positioned in 11-13 all make it.

That set of conditions come together in about .3% of scenarios to get 5 Hockey East teams in.

How Hockey East could be limited to only two teams in the NCAA tournament

Here’s an example:

The keys are:

  • Because Boston University is going to make the tournament anyway, the conference championship either needs to go to BU or another team that’s going to make it at-large.
  • BC is exceptionally hard to push out (see above), so instead preventing any of Mass.-Lowell, Providence, or Vermont from climbing is easier. Vermont is coming from far enough behind that despite a first round win, it’s easy to keep their PWR low with wins by neighbors Harvard, Bowling Green, and St Cloud St.
  • Finally, enough conference tournaments (in this scenario, 3) go to low ranked teams to keep out any remaining Hockey East teams on the bubble.

Those factors come together to limit Hockey East to two NCAA bids in only about 5% of scenarios.

How the ECAC could send three teams to the NCAA tournament

Here’s an example:

The keys are:

  • Colgate, Harvard, and Quinnipiac are all on the bubble. But, because Harvard plays Quinnipiac in the first round, the winner of which plays the winner of St. Lawrence vs. Colgate, it’s really hard to get all three in. Both Harvard and Colgate are long shots, so in this scenario we sacrifice one to get the other in, sending Harvard all the way to the championship.
  • Advancing Harvard damaged Quinnipiac, so we then need to make sure as many conference tournaments as possible go to top teams. In this scenario, that’s 4, allowing #14 Quinnipiac in at-large. That requirement also allows Yale to slip in (see “How to get Yale in” above).

Those factors come together to get 3 ECAC teams in the NCAAs in about 6% of scenarios, so it can also happen without Yale being one of the teams

Conference participation possibilities for NCAA tournament

Even though Boston University is Hockey East’s only current lock for the NCAA tournament, they’re sure to get at least two entrants. That suggests that the only scenarios that knock Boston College out involve at least one Hockey East team other than BU making it (Mass.-Lowell, Vermont, Providence, or New Hampshire).

Hockey East is actually most likely to send 3-4, with 5 still possible.



The Big Ten is only guaranteed to place its conference tournament champion in the NCAA tournament. But, if Minnesota does well but fails to win the championship, the Big Ten could end up sending both the Gophers and the champ.



The NCHC is guaranteed to send five teams (Denver, Miami, Minnesota-Duluth, North Dakota, and Omaha), but is most likely to send a sixth (adding St. Cloud St).



The ECAC is only guaranteed to send its conference tournament champion, though its likely to send a second from Colgate, Harvard, Quinnipiac, or Yale. Harvard, Colgate, and Yale are all reasonably long shots, so not all can make it resulting in a maximum of three participants from the ECAC.



Atlantic Hockey will only send its tournament champion to the NCAA tournament.



The WCHA is guaranteed to send two teams, Michigan Tech and Minnesota State. They’re quite likely to send three and possibly even four from those two plus Bowling Green and Ferris State.


Odds and ends – .500, ties, consolation game and SCSU

There are a few oddities in this year’s tournament outlook and structure that have been the subject of numerous comments—

  • The .500 rule (teams must have a record of at least .500 to get an at-large bid)
  • The tie possibility (the NCHC consolation game, unlike all the other tournament games, can end in a tie)
  • The consolation game (the NCHC consolation game is now the only one of its kind)

I examined the effects of each of these on the possible scenarios and found that none of them are a very big deal. While each can, of course, result in a different team making the tournament in some situations, none distort the outcomes in particularly unusual or undesirable ways.

The .500 rule will keep one otherwise eligible team out of the tournament in a handful of scenarios, the tie possibility will give its participants the opportunity for an outcome a bit better than a loss (but a bit worse than a win), and the consolation game will give an additional opportunity for some movement to two teams that would be done playing without it.

On the .500 rule

For the first time in recent memory, it’s possible for a team with an under .500 record to be in PWR position for an at-large bid. St Cloud St could lose two games to earn an 18-19-1 record, but still be ranked high enough for an at-large bid.

I ran the scenarios a second time, tweaking the rules to let SCSU into the tournament from that position despite the <.500 record to see how much that rule changed things. While it will clearly be a big deal to the teams affected if it happens, the overall impact is pretty small. Without the .500 requirement, SCSU would make the tournament in about 21% of its win 0 scenarios (which include the possibility of a consolation game tie), compared to 19% with the .500 requirement in place. So, absent that rule, SCSU would make the tournament at-large in about 4% of its two loss scenarios and bump another team. The potential victims of that bump would be other bubble teams, primarily Minnesota, Harvard, Colgate, Mass.-Lowell, and Yale.

On the possibility of a tie in the NCHC consolation game

There is only one game in the remaining conference tournaments that can end in a tie — the NCHC consolation game. Though like any game it could definitely change who makes the tournament, it doesn’t throw a huge wrench into the process. Other than St Cloud St’s .500 situation described above, no team can achieve a different ranking this year due to the existence of the tie than they could achieve without it.

The possibility of a tie has the biggest obvious impact on SCSU, giving them an opportunity to go winless on the weekend on still make the NCAA tournament (with a loss and a tie).

Not surprisingly, each of the other three NCHC teams with a potential for a loss and a tie also fare slightly better in that scenario than they would with two losses—UND can finish #1 more frequently, Miami can finish #7 more frequently, and Denver has significantly more potential to finish #3-6 then the #7 they’d likely finish with two losses.

On the existence of the NCHC consolation game

Similarly, the existence of the NCHC consolation game doesn’t have any unusual effects on the field this year.

By giving an extra game to two teams that have lost, each will have an additional opportunity to either make up some of the lost ground or lose even more ground than if the consolation game weren’t played.

By virtue of being on the bubble, SCSU is again most affected. Without the consolation games, they would advance in a decent share (about 30%) of scenarios in which they lost in the semifinals. Forcing them to play another game after such a loss puts them back in control of their own destiny, facing certain elimination if they lose or a much improved tournament outlook if they win.