All posts by gfmorris

2014-15 WCHA tiebreakers and their relevance

Here’s how the tiebreakers work in the WCHA.  You will regularly see me call these “the A tiebreaker” or similar.  All tiebreakers are used to compare two or more teams.  If three or more teams are involved, a tiebreaker pulls the winning team out and resets the comparison process.  More on that later.

  • A tiebreaker: a team wins the tiebreaker if they won a four-game season series.
  • B tiebreaker: a team wins the tiebreaker if they had more conference wins.
  • C tiebreaker: teams are compared based on their winning percentage in head-to-head matchups.
  • D tiebreaker: teams are compared based on their winning percentages head-to-head against teams down the standings table.
  • E tiebreaker: winning margin (goals for minus goals against) in conference contests.
  • F tiebreaker: coin flip.

Now, last year showed that it’s unlikely that we’ll pass the D tiebreaker.  Let’s see if that’s true based on where we are with everyone having played 24 league contests.

The spread

At this point, it’s Mankato-Tech 1-2 for the McNaughton Cup.  BG getting swept by Northern ended their shot at one of the top two seeds, and they’re quite clear of Bemidji and Northern in 4th — but we’ll get to that, too.

My concern here will be in the following comparisons:

  • Who wins the McNaughton, Tech or Mankato.
  • The edge case of Bemidji and BG/Northern interacting.
  • Who comes away with home ice in the Bemidji-Northern-Ferris trio.
  • Which two teams make the playoffs at the bottom of the bracket.

The McNaughton Cup

This is probably the easiest setup: Mankato has the edge by one point, and the two teams play each other two more times.

The A tiebreaker will be in play, as the Mavericks swept the Huskies earlier in the season.  If the Huskies sweep, they make the season series even and push things to the B tiebreaker.

  • If Mankato sweeps, they are 21-3-2 (44 points), and they are the regular season titlist by being five points clear going into the final weekend.
  • If Mankato wins and ties, they are 20-3-3 (43 points) while Tech is 19-5-2 (40).  Tech would have to sweep Northern while Bemidji wins at least one against the Mavs, because Mankato still wins the A tiebreaker.
  • If the teams split, Mankato is 20-4-2 (42) and Tech is 20-5-1 (41).  Mankato still gets the A tiebreaker because they went 3-1-0 on the season.  As such, Tech is going to have to get two more points on the final weekend than Mankato does.
  • If Tech wins and ties, the Huskies are 20-4-2 (42) and the Mavs are 19-4-3 (41).  Mankato still has the A tiebreaker, so they will need just one more point than the Huskies get on the final weekend.
  • If Tech sweeps, the Huskies are 21-4-1 (43) and the Mavs are 19-5-2 (40).  The A tiebreaker is now out of play.
  • Mankato would have to sweep Bemidji to get to 21 conference wins, and the Huskies would be okay as long as they got a win versus Northern. However, if the Wildcats take three points — and they did to start the seasons — both teams would be 21-5-2 on the season.
  • At that point, you go to the C tiebreaker, which is also out, and then the D, which starts with performance against teams below them.  Both teams split with BG.  Both teams would have swept Bemidji.  Mankato would win because they picked up a win against Northern.

BG is very likely to finish in third

Northern Michigan ended Bowling Green’s chances to make the McNaughton chase a three-team affair.  The Falcons can only end up tied for second in points with Michigan Tech at 39 if they win out and the Huskies lose out.  That makes Tech 19-8-1 and BG 18-7-3, and Tech wins the B tiebreaker, conference wins, as Tech and BG played just twice this season.

There is the unlikely event that the Falcons lose out and end up 14-11-3, 31 points.  Bemidji State or Northern Michigan could win out for 32 points and a shot at third.  (More on any BSU-NMU tiebreakers later.)  If either (or both) of those teams ended up with just 31 points, they’d go 3-0-1 in their final four games to finish 13-10-5, while the Falcons would be 17-7-4.  The Falcons’ edge in conference wins (and thus the B tiebreaker) won’t matter in this scenario, as the Falcons were 2-1-1 against Bemidji and 1-2-1 against Northern.  In an insane, three-way tie, BG wins with Northern in third, having lost both comparisons.

In short: BG can’t get to 2nd; they can get to 5th if they lose out and Bemidji and Northern win out; if they end up tied with 31 points, they beat Northern in the A tiebreaker but lose to Bemidji.

The race for home ice and, in consolation, realizing that they avoid the top two teams

Bemidji State and Northern Michigan both sit tied for fourth at 10-10-4 (24 points).  Ferris is two points behind them at 11-13-0.  Ferris has the potential for winning a B tiebreaker, as they have one more conference win at this point.

Let’s ignore the chances that ties get involved here and just say that there are wins and losses here.  Here are the remaining games for each team:

  • Bemidji State: @ Ferris State, v Minnesota State
  • Ferris State: v Bemidji State, @ Lake Superior
  • Northern Michigan : v Lake Superior, home-and-home with Michigan Tech

A tiebreaker: Bemidji and Ferris this upcoming weekend for the first time.  Bemidji won the season series 3-0-1 over Northern.  Ferris and Northern only played twice, splitting the series.

B tiebreaker: conference wins is pretty straightforward.  Any tie between Bemidji and Northern is very likely to result from their winning the same number of games.  Ferris will have to win one more game than either or both of the other two win, so they’re guaranteed to win the tiebreaker in that scenario.

C tiebreaker: head-to-head comes back into play again, and the only unknown is the BSU-FSU ranking.  Again, Bemidji wins over Northern and Ferris and Northern push to the D tiebreaker.

D tiebreaker: It’s pretty unlikely that we’re going to get here.  Bemidji wins the A tiebreaker over Northern, so equal runs there still pushes the Beavers ahead.  A three-way tie means that Ferris wins the B tiebreaker, and cycling back again favors the Beavers.

If Ferris and Bemidji split, the C tiebreaker is irrelevant.  But then Ferris has to win one more game than the Beavers, which means that the Mavericks sweep while the Bulldogs split with the Lakers or the Mavs and Beavers split while the Bulldogs sweep.  But then Ferris State is 13-15-0 or 14-14-0 while Bemidji is either 11-13-4 or 12-12-4, and either of those scenarios results in the Bulldogs winning the B tiebreaker.

You could negate Ferris’s advantage by having the Bulldogs tie four times, finishing them at 11-13-4.  But Bemidji could then be no more than 12-10-6 if they swept the Mavericks, and they’d be 11-11-6 if they split.  That result gets us to the D tiebreaker, as the teams played just twice, have the same conference wins, and were then tied in HTH.  Then we do the D tiebreaker:

  • MSU: Bemidji would be 1-3-0, and Ferris went 0-4-0.
  • MTU: Bemidji and Ferris were winless against the Huskies.

As you can see, the only way for Ferris to jump into 4th is for them to win one more conference game than both Bemidji and Northern.  Ferris only controls its destiny by playing the Beavers this weekend; otherwise, they are fans of Lake Superior and Michigan Tech, a position that’s surely strange in Big Rapids.

The races for last

I want to stop for a second and lament Alaska’s ineligibility.  They’d be tied for sixth with Ferris State, and that pairing of four would be more interesting than the three pack, mainly because the Nanooks are 10-12-2 and ties just make for chaos.

But the UAH-LSSU-UAA race is pretty simple.  I wrote earlier on UAHHockey.com that the Chargers need three points to get to tiebreakers and four to win outright.  They can’t get to 6th, but they can miss entirely if they win just one game (8-19-1, 17) while Anchorage wins out (8-18-2, 18) and Lake State gets at least five points (8-18-2, 18 or 7-17-4, 18).  Should all three teams end up 8-18-2, UAH ends up in 7th.  If UAH and LSSU end up tied in points, UAH wins based on their 2-1-1 season series.

LSSU then really only has UAA to worry about if they play such that they go to tiebreakers.  UAA hasn’t won in the lower 48 this year (one tie against Maine), and they’re on an 0-8-0 run.  They get Bowling Green this week in Anchorage before hosting Alaska the next, so if they’re going to make a run, this is it for them.

Tiebreakers: The teams split two meetings this year, so our concerns lie with the B and D tiebreakers.  Let’s consider the D tiebreakers first before going through the permutations on the other end.

  • MSU: both teams were winless against the Mavs.
  • MTU: both teams were winless against the Huskies.
  • BGSU: the Lakers and the Falcons split two games, while the Seawolves have yet to face BG.
  • BSU: UAA went 0-2-2, while LSSU went 2-0-0.
  • NMU: UAA went 1-3-0, while LSSU is 0-2-0 against NMU going into this weekend.

Whew.  Now let’s look at every way that these teams can end up tied.

18 points:

  • Anchorage sweeps Bowling Green (winning the D tiebreaker at that point) and then Alaska to get to 8-18-2.
  • LSSU would need five points to get to 18, which is either two wins and a tie (8-18-2) or a win and three ties (7-17-4).
  • Two LSSU wins ensures that Anchorage wins the D tiebreaker; one win ensures that Anchorage wins the B tiebreaker.

17 points:

  • Anchorage would get three wins and a tie (7-18-3).  This means that Anchorage wins the D tiebreaker, since they wouldn’t lose to the Falcons.
  • LSSU would need four points, which means they either win two games (8-19-1), a single win and two ties (7-18-3), or the insanity of four ties (6-17-5).
  • Two LSSU wins ensures the Lakers win the B tiebreaker.
  • One LSSU win and two ties sees that Anchorage wins the D tiebreaker
  • Four ties ensure that Anchorage wins the B tiebreaker.

16 points:

  • Anchorage either finishes 3-1-0 (7-19-2) or 2-0-2 (6-18-4).  At best, they sweep the Falcons and win the D tiebreaker; at worst, they get both ties against the Falcons and sweep the Nanooks, which leads them to being Wildcat fans.
  • LSSU would need three points, which is either 1-2-1 (7-19-2) or 0-1-3 (6-18-4).
  • If the Lakers get their points via ties only, they lose the B tiebreaker.
  • If the Lakers win-and-tie, they win the tiebreakers if: 1) the Seawolves go 2-0-2 (win on B); 2) the Seawolves split with the Falcons and the Beavers edge the Wildcats in the standings.
  • The Seawolves can win if the Lakers win-and-tie and they go 3-1-0 and they save that loss for the Governor’s Cup.

15 points:

  • Anchorage finishes 2-1-1 (6-19-3) or 1-0-3 (5-18-5).
  • LSSU gets two points, either 1-3-0 (7-20-1) or 0-2-2 (6-19-3).
  • LSSU wins the B tiebreaker if they win a game or if their two-tie effort is met with a 1-0-3 effort for the Seawolves.
  • If LSSU goes 0-2-2 and UAA goes 2-1-1, the Seawolves win the D tiebreaker if they go at least 1-0-1 against the Falcons.  They could lose the same tiebreaker if their two wins come against the Nanooks or if it’s Bemidji edging Northern.

14 points:

  • LSSU ties just one of its four remaining games (6-20-2).  It won’t matter which.
  • UAA finishes either 2-2-0 (6-20-2), 1-1-2 (5-19-4), or 0-0-4 (4-18-6).
  • LSSU wins the B tiebreaker except when UAA goes 2-2-0.
  • If UAA goes 2-2-0, it wins the D tiebreaker when it sweeps the Falcons.  If it splits with the Falcons, they’re rooting for Bemidji.  If they are swept by the Falcons, LSSU wins the comparison.

13 points:

  • LSSU loses out (6-21-1).
  • UAA either goes 1-2-1 (5-20-3) or 0-1-3 (4-19-5).
  • LSSU wins the B tiebreaker.

Now, wasn’t that fun?

Week 21 Predictions

B9r-n_XCYAAo4if.png-large

 

And I’m back!  I have prettier graphs!  Well, at least this one is pretty.

The WCHA standings are ranked in three tiers: Mankato-Tech-BG, NMU-BSU-FSU, and UAH-LSSU-UAA.  This is true in the actual standings as well as the predicted ones that you’ll see below.

Here’s a legend on how to read the above graphic: the team in a bold color is on the road (with the best approximation of their colors that I can get in Excel), while the team at home is in white.  So if you look at Bemidji-Northern:

  • Bemidji wins about 1.09 times in the 10,000 runs that the model does.
  • Northern wins about 0.88 times .
  • The teams get an average of 0.04 ties.

You get this math from the ones below it, which are HM0% (the chance that the home team is swept), HM1% (the chance that they lose and tie), etc.   I’ll explain a little bit more about the math in a bit.

As you can see in the Bemidji-Northern series, a split is the most common option (40.69% of all runs), with a Bemidji sweep second (32.94%) and Northern sweeping third (22.63%).  The remainder are events where the model thinks that the schools might tie.

The math — expected values

The model I’m presenting here uses KRACH only for a way to generate an expected value of the series.  KRACH is a rigorous mathematical answer to the improper application of the transitive rule, which could be considered from a UAH perspective like so:

“Hey, Omaha is 16-7-3, eighth-best in the country winning percentage-wise with the second-toughest strength-of-schedule.  But you know what?  UAH tied them one night and lost by only a goal another night.  We’re not that bad!”  Except, you know, Tech RUTS’d UAH and has almost an inverse record to the Red Cows (7-18-3).

KRACH compares everyone to everyone with matrix mathematics that accounts for the fact that, well, everyone doesn’t play everyone.  So when teams cross over into non-conference play, it matters.  The three teams with 20-win seasons so far are Minnesota State, Michigan Tech, and Robert Morris, but the relative non-conference records of those two conferences makes a difference, as does the schedules that each of those three schools played:

  • Mankato: @Omaha 2x, H/H with Duluth, v Princeton, North Star College Cup (Minnesota, Bemidji State)
  • Tech: v Michigan 2x, @ Duluth, GLI (Michigan, Ferris State), @ Wisconsin (the split there hurts them)
  • RoMo: Lake Superior, Three Rivers Classic (Penn State, Colgate), H/H with Bowling Green

KRACH gives a numerical ranking  that can be used to do a backwards look.  If Robert Morris (103.5 in KRACH before games on 2015-02-13) played Air Force (23.82), you’d expect the Colonials to win 81.2% of the time [103.5/(103.5+23.82)].  This lines up  fairly well with reality, as the Colonials are 2-1-1 against the Falcons this season (.750), with all four contests going to overtime.

A greater disparity can be seen in the MTU-UAA matchup, where the Huskies get the sweep 100% of the time.

Mind you, I had confused the M*U with the UA* matchups this weekend (I blame the cold medicine).  But Mankato is a virtual lock to sweep UAF, too (93% of the time).

The math — distributions

Now once you get this expected value from KRACH, you can consider the results to be normally distributed, i.e. on a bell curve.  This is to say: all things told, if you know how likely Team A is to defeat Team B, you can set that as the expected value of the distribution and then run simulations on that distribution.

Now wait, you’re saying, how are these things equally distributed?  Didn’t you say that Tech was going to crush Anchorage?  Yes, I did.  Tech’s KRACH of this moment is 411.8; UAA’s is 47.44.  Tech should win 90% of the time by that.  Sorta.

See, the model says, “Okay, Tech’s going to ‘win’ 90% of the time, i.e. they’re going to get 3.6 points per weekend.”  And that expected value is what’s used.  Why?  I’ve never gotten the sense that college hockey games are independent events, which is to say that what happened on Friday night will drive what happens on Saturday night (injuries, benchings, etc.).  This may be a failing of the model — I haven’t tested it extensively, but it worked reasonably well last year. But it’s the model that I’ve chosen.

So if Tech is supposed to pick up an average of 3.0 points per weekend, rounding that up means that, on average, they sweep.  In fact, if you use breakpoints in determining what is a sweep (for the model, it’s 2.95 /4, which gets pretty close to the historic average of ties produced in WCHA games), Tech sweeps every time.

The math — standard deviation

We’ve all heard the term “outlier”.  We probably know one.  Shoot, pretty much every NHL player is an outlier in some form or fashion, a man so uniquely skilled at hockey that people pay him vast sums to do so.  But even in the NHL, some guys are simply better than others.

In statistics, this concept is variance.  Tied closely to variance is the concept of a standard deviation, which is to say how wide the distribution is.  In your standard academic exercise, 20 kids take an exam where the average is 68: 18 kids make an 80, one makes a 100, and one makes a 0 because they tried to cheat off of the kid acing the exam.

You can run this if you want, but the standard deviation is about 17 points, which is to say that the kid acing the exam was more than two standard deviations away, while the kid who cheated was four.  In statistics, the former is expected variance — there’s nothing unusual about a kid who aces an exam when the vast preponderance of the class barely passed it.  However, the latter is significant and should prod a question as to why (even though we know why in this case).

The application

In the Tech-UAA case — thanks for staying with me — the mean is shifted so close to the 4-point limit that centering a bell curve here also mean that we have to consider the width of said curve.  I set that width — the standard deviation — to the width of the mean in comparison to its distance from the edges.  As such, the closer you get to 0.0 or 4.0 expected values for the home team, the more likely a sweep will happen.

You can see this with UAH @ LSSU, which is nearly even in terms of KRACH (48.19 UAH, 43.97 LSSU).  This is why a split is most likely: the mean is very close to 2.0, and the variation is pretty wide, so the answers to to spread evenly between the center and the poles.  This is largely a pick-’em series, but the numbers say that UAH is slightly better from a comparison standpoint.

But again, that’s why they play the games.

The cascade

If you simulate all 22 remaining series — and I have — 10,000 times, you get results that look like the below.

B9sGuWHIQAA2hbf.png-large

 

It’s a shame that Alaska is ineligible, because three teams fighting for one spot would be far more desperate than three fighting for two.

My intent for the next couple of weeks is to come up with a way to set up a table that shows how many times that, say, the above order comes into play.  If I were doing this with a database and not an Excel spreadsheet, this would be simpler, but I ran out of time to do anything else.

 

Thinking on Next Season — Recency, KRACH, and Goal Differential

My summer project is to learn Python so I can put the prediction information out on a Web app.  Also, the programming that I want to do lends itself to a scripting language rather than using Excel as a brute force Monte Carlo simulator.  I’m going to simulate all 140 games and their results, and that would be one huge, unwieldy spreadsheet that would be a bear to create.  Using Python and an SQL database is a saner approach.

I’ve been thinking about what makes sense for a predictive model.  Here are my thoughts — feel free to chime in.

  1. Even though KRACH is reactive rather than proactive, I think that it’s a reasonable predictor of future performance — say 60% or so.  KRACH gives you a strong indicator of team quality.
  2. I think that KRACH suffers because it uses win-loss data instead of goals for/goals allowed data.  With the WCHA having a logjam in the middle of the rankings, I think that GF/GA would’ve been a good discriminator.  Let’s consider two schools: Alaska and Michigan Tech.  Late in the season, the Nanooks went into Houghton and blitzed the Huskies.  KRACH saw those simply as two road wins; a goal-differential ranking system would see 7-3, 7-2 as an indicator that the Nanooks were demonstrably better rather than just clearly better, and perhaps their home sweep of Ferris State the next weekend wouldn’t have been as big of a surprise.
  3. There’s something else to consider: recency.  Let’s consider Alaska and Lake Superior.  The Nanooks were 14-12-2 for the season, but they started off 3-7-0.  Did they face a tough WCHA schedule in that stretch?  They played Northern (home split), Lake (away split), Ferris (swept away), Tech (swept at home), and Anchorage (away split).  They caught each of those teams when they were better than they were all season (save Ferris, who was hot all year).  Given how the Nanooks finished the regular season (5-1-0), you can argue that recency matters.  (And yes, Anchorage, you beat them three-of-four to end their season.)
  4. A recency view also makes sense when you look at Lake Superior. They started 4-1-0 and were fairly highly ranked, and they finished out of the playoffs, getting swept by the top two teams in the league the last two weekends.  But you can also look at January, where the Lakers went 2-6-0.  The wheels eventually fell off in the Sault, and recency matters.  Also, you can make a goal-differential argument as well: yes, the Lakers beat the Chargers on the road, but both games were one-goal margins.
  5. One more argument for recency and goal-differential: UAH clearly improved as the season went on, as the blowouts were fewer and farther between.  The Bemidji road win seemed impossible at Christmas, but it was merely improbable by February.

I think that I’ll end up with a not=so-secret sauce for the model: won-loss KRACH, a paired-comparisons goal-differential system, and recency (probably last 10).

If you’re thinking this through, though, you’re anticipating my next step: each predictive run will predict each game, with the results driving subsequent predictions.  You’ll play a night’s games, re-calculate the model, then play the next slate of games.  It’s an iterative approach, and it’s possible that one team will go off on a tear on any individual simulation, just as would happen in real life.  Running a Monte Carlo simulation will even that out, but you’ll be able to see how many times a team streaked or slumped, because we’ll have each individual simulation.

I’ll also be able to adjust the weights of the model as the season goes by, as I’ll be able to compare the model to reality.  The recency weights in the model should temper my urge to tweak the weights of the other two parameters very far.

No Math: How to Do a 10-Team WCHA Playoff: Play-IN

I’ve been thinking about how to pull off a 10-team WCHA playoff with minimal perturbation to the first-round / Final 4ive.  I don’t think that a five-team or five-game playoff should be an end goal, as I like the current setup.  But if we’re going to go to an everybody-in system, I’d like to propose an alternative.

The Play-in

Seeds 1-4 get home ice.  Seeds 5-6 get certainty with their travel plans to 3/4.  Seeds 7-10 get sent to a 1/2 site with the #1 seed choosing the pairing whose winner they would play in the full first round.  But they’re only choosing this by picking who they host for a play-in game: 7-10 or 8-9.  7-10 and 8-9 play a single play-in game on neutral ice, and the winner then plays the host school (i.e., no re-seeding).  Finishing 7-10 now means that you have to win three games to make it to the Final 4ive.

Let’s look at this with how this season finished.  Ferris State, as the #1 seed, would’ve been able to pick a pair: Northern Michigan & Alabama-Huntsville or Bemidji State & Lake Superior.  Ferris State would have had until 3:00 p.m. on Sunday, March 9th to decide whom they would face in the opening round.  A play-in means that there’s even more reason to get the #1 seed, because you’ll get to handicap your best competitor.

Ferris could have taken BSU-LSSU and been pretty happy with it:  the Bulldogs did well against both teams, and they really held serve at home, so they would have felt confident in facing their opponent.  Or they could have taken the NMU-UAH pairing.  They were 3-0-1 against the Wildcats.  But maybe UAH knocks off Northern Michigan in that 7-10 play-in game!  Then Ferris would have an even easier task in its first-round series, as it’s pretty much a given that the McNaughton Cup winner can pick up two wins in three games no matter how many headstands Guerriero and Larose do.

Think about it for a second: two play-in games would be really exciting, probably as exciting as the WCHA’s final night was.  And even that would still have been exciting, as UAH was the only assured seed at #10.  Three teams (BSU, LSSU, NMU) could’ve been the 9th seed, four the 7th and 8th (UAA, BSU, LSSU, NMU in some combination).  After all, 66 of the 243 (27%) scenarios for how the final day came out would’ve had UAA in either 7th or 8th, and of course they could’ve gone as high as 5th.  If nothing else, we would’ve not heard caterwauling about how UAH made LSSU miss the playoffs by not playing them four times.

Don’t get me wrong: I like the top-8-or-go-home scheme.  I’ll like it next season if UAH is scuffling to be in the playoffs (don’t scoff or it might happen!  Wait, please scoff), too.  For a team at the bottom, I think that the push to fight and claw for getting into the top 8 means a lot — because making the playoffs is an accomplishment, not a given.  But if you’re really hell-bent on having all ten teams involved, send 7&10 and 8&9 to the venue of the #1 seed’s choosing.

What say you?  Comment below or reply to me on Twitter at @wchaplayoffs.

2014 WCHA Playoff Race: Playoff Branches

Are you ready to rumble?  Now, you would think that this would be pretty easy: five games.  But those five games can go three ways: a win for each team or a tie.  (You have to consider both teams because it factors into the standings each way.]  We will consider the scoring from the perspective of the home team as convention.

As we know, we have five games tonight.  I’m going to refer to them by Game #s, much the same way that the NHL does their games.  The WCHA plays 140 games: each team hosts 14 games over the course of the season.  As such, these are Games: #136-140.  We’ll go by the puck drop time and then by alphabetical listing by the home team to differentiate the games.  So we have:

  • Game #136: Bemidji State University (BSU) at Bowling Green State University (BGSU), puck drop 7:07 p.m. Eastern.
  • Game #137: Lake Superior State University (LSSU) at Ferris State University (FSU), puck drop 7:07 p.m. Eastern.
  • Game #138: Northern Michigan University (NMU) at The University of Alabama in Huntsville (UAH), puck drop 7:07 p.m. Central.
  • Game #139: Michigan Tech University (MTU) at Minnesota State University, Mankato, puck drop 7:07 p.m. Central.
  • Game #140: University of Alaska Anchorage (UAA) at University of Alaska (UAF), puck drop 7:07 p.m. Alaska.

Now, anyone who’s been following this fun WCHA playoff run (USCHO, UAHHockey.com) knows that the final tally of this season will come down to Game #140.  Will Anchorage stay in the playoffs?  Will either school bump teams out for home ice?  All of these things were possible from my perspective starting this post before Friday’s games — I needed the time to set everything up!

There are 243 ways that this whole sordid mess can be sorted out tonight.  Yes, 243: three possible outcomes for each of the five home teams is 3 * 3 * 3 * 3 * 3 = 243.  Yeah, this breakdown seemed a lot more fun when it was 2 * 2 * 2 * 2 * 2 = 32.

I’ve attached a PDF an Excel spreadsheet at the end of this post.  It’s the clearest way to present the situation, because there are branches off of the root of “Saturday’s games that have not been played”.  Let’s look at how this might work:

Winners: BGSU, FSU, UAH, MSU, UAF.

That is one of 243 overall branches, and it’s also one of 81 branches where the Falcons start the night off with a win.  If they tie or lose, you can throw those 81 winning branches away and focus on the branches that have their root with the result of the Beavers-Falcons matchup.

I think that you can see where this is going: when 9:30 p.m. Eastern rolls around, we’ll be from 243 branches down to just 27: the NMU-UAH result, the MTU-MSU result, and the UAA-UAF result.  An hour later, it’ll be down to just three.

But I wanted to do the hard work for you so that you could look things over yourself.  Print this out, or look at it in your browser — you’ll be able to follow this.  Just read it left to right.  The labels will work with you to help you narrow things down pretty quickly.  By the time the first games are over, you’ll be down to just one of the pages.

Let’s use that BGSU, FSU, UAH, MSU, UAF result again, which is #243 in the table.  That gets us the following table:

Seed Team Record Points
1 Minnesota State 21-7-0 42
2 Ferris State 20-6-2 42
3 Alaska 15-11-2 32
4 Bowling Green 13-11-4 30
5 Michigan Tech 12-12-4 28
6 Alaska-Anchorage 11-13-4 26
7 Northern Michigan 12-15-1 25
8 Bemidji State 10-14-4 24 1
  Lake Superior 12-16-0 24
  Alabama-Huntsville 3-25-1 7

BSU wins the A) tiebreaker, 3-1-0 in head-to-head play with the Lakers

Now, the PDF doesn’t have that listed tabularly, but it does give you a 1-10 listing of the teams.  You can follow the tiebreakers and standings yourself from my post over on UAHHockey.com earlier this week, but in short: A) if (all of) you played four times, it’s the head-to-head record(s); B) Conference wins; C) Winning percentage against teams from the top to the bottom of the table.

What we know here on Saturday afternoon:

  • UAH will be 10th.
  • UAF will be 3rd.

That’s it.  That’s what’s so awesome about this!

Check out the PDF.  I hope that this makes for an enjoyable companion to your Saturday night scoreboard watching!  I have audited it reasonably extensively, but it is entirely possible that I missed something.  If so, you have my apologies.  I felt that we needed a different way to visualize this information, and that was my aim.

2014 Playoff Branches 0

2014 WCHA Playoff Race: Probabilities Predict Standings

Sometimes, the obvious stuff doesn’t occur to you right away.   The probabilistic model could be extended: take the most likely scenario and use that to predict the final standings!   Gosh.

That means:

  1. BG and BSU split.
  2. FSU sweeps LSSU.
  3. NMU sweeps UAH.
  4. MSU and MTU split.
  5. UAA and UAF split.

That gives you the following:

Seed Team Record Points
1 Ferris State 20-6-2 42
2 Minnesota State 20-8-0 40
3 Alaska 14-12-2 30 1
4 Michigan Tech 13-11-4 30
5 Bowling Green 12-12-4 28 2
6 Alaska-Anchorage 12-12-4 28
7 Northern Michigan 13-14-1 27
8 Bemidji State 11-13-4 26
  Lake Superior 12-16-0 24
  Alabama-Huntsville 2-25-1 3

1 A) The teams went 2-2-0 this year, and B) the Nanooks have one more conference win.

A) The Falcons were 2-1-1 against the Seawolves this season.

Our bracket would be:

  • #8 Bemidji State at #1 Ferris State
  • #7 Northern Michigan at #2 Minnesota State
  • #6 Alaska-Anchorage at #3 Alaska
  • #5 Bowling Green at #4 Michigan Tech

The Alaska Plan ended up not becoming a reality, but here it is in all its splendor.  Oh my.

2014 WCHA Playoff Race: The Final First-Order and Probabilistic Models

Welcome to the new blog, everyone!  It was unfair to have this take over UAHHockey.com for much longer, and I have longer-term plans on it, anyway.  New blog, new Twitter handle, same nerd.

Here are the final predicted standings from the first-order model:

Team Record Points
1 Minnesota State 21-7-0 42
2 Ferris State 20-6-2 42 1
3 Alaska 14-12-2 30 
4 Bowling Green 12-12-4 28 2
5 Alaska-Anchorage 12-12-4 28
6 Michigan Tech 12-12-4 28
7 Northern Michigan 13-14-1 27
8 Bemidji State 11-13-4 26
9 Lake Superior 12-16-0 26
10 Alabama-Huntsville 2-25-1 3

1 The MacNaughton Cup would be awarded to both teams, but for seeding purposes, the B) tiebreaker goes to the Mavericks, who have  one more WCHA win.

The three-way log-jam at 12-12-4 is broken by the C) tiebreaker, given that all three schools did not play the other two squads four times.  They all have 12 wins, so you have to go to winning percentage down the table.  Against the Mavericks, the Huskies have the worst performance (.000), while the other two were .500.  At the next level, the Falcons went .250 against the Bulldogs, better than the winless Seawolves.

I call this a first-order model because it uses KRACH to create an expected value parameter that is then used to make a reasoned prediction.  It’s effective, to be sure, but it’s pretty simplistic.  As such, it’s been wrong this year on a number of occasions.  There are some limitations to the model, and let’s just look at where we work in Week 17:

Team Record Points
1 Ferris State 20-0-8 48
2 Minnesota State 16-10-2 34
3 Bowling Green 13-9-6 32 
4 Alaska-Anchorage 13-11-4 30
5 Northern Michigan 13-13-2 28
6 Lake Superior 13-13-2 28
7 Bemidji State 11-11-6 28
8 Alaska 11-14-3 25
9 Michigan Tech 9-12-7 25
10 Alabama-Huntsville 2-25-1 3

Yes, the team that the model currently expects to be #3 was in 8th, and the #9 team, Michigan Tech, is now very likely to be in the home ice race, too.

Clearly, in-season modeling is hard.  Hockey teams aren’t single numbers.  Just this past weekend, the expected value model said that Alaska would only take one point.  They, uh, swept Ferris.

Oh, and the first predictive model I had said that they had nearly no chance of this.  But I can make my model better, and I have.

Screen Shot 2014-03-06 at 6.24.48 PM

We’ve gotten somewhere!  Here’s the source of my problem: I was under-estimating UAH’s chances at winning at Bemidji.  I famously said that UAH had a 1.6% chance of winning, but you know, I was underestimating them.  Also, I had too much room for teams to get ties.

Now the model does the following:

  1. The standard deviation range goes in the right direction.  If the teams are rather unequally yoked, the SD needs to be larger (squashing the bell curve down); when you’re near 2.000 points, you have a smaller SD.  Why?  You don’t need to widen the spread as much to give the unequal teams a chance to have all possible results.  After all, UAH gets points about 1/9 of the time, and you have to account for that properly.
  2. There is a weighted standings component.  It doesn’t have an “if this happens, these are the results with the tiebreakers applied” yet, but that’s planned for the summer.  No, doing an average of the 10,000 runs will give you the estimated number of wins, and that should be a good proxy for the tiebreakers.

Screen Shot 2014-03-06 at 6.25.11 PM

As you can see, the FSU-MSU thing really is a tossup: 55 comparisons in 10,000 results are all that separate them. Also, in nearly every scenario, equal points favors the Mavericks.  (The alternate scenario: MSU ties MTU twice.)

Bemidji State fans are unlikely to be happy with me here, but losing to UAH really hurt their KRACH.  (It breaks my heart.)  It’s reasonable to think that LSSU and NMU can pick up points (see above), and it’s hard for BSU to pick up more than a point, on average.  They’re on the road against a good team that is very good at home.

(About that: there is no home-ice bonus.  I need to study that NCAA-wide to figure out how to best model that.  I have an idea, but I didn’t want to enter more noise into the system that I already have with an unproven model.

I’ll run this again on Saturday to see where we are in the finals.  I also have a nice toy coming for you on Saturday.  #branches