2019 Bracketology Thread

[calourie]

Again agreeing with Hawkeye, at least moderately so, in this case over Admin. A Towson loss to Umass in the CAA finals, combined with a Hopkins loss to Maryland in BIG round 1 would leave Towson with a H2H win over HOP as well as (I believe but am not totally sure) a higher RPI than the BlueJays. The RPI difference won't be more than 1 in any case and the committee might well (and I think they should) include a 10-5 Towson over a 7-7 Hopkins given that scenario. But hey, I'm not on the committee.

[admin]

Loyola potentially not winning the Patriot League is the biggest monkey wrench out there that I could see upsetting the scenario in your article.

Agreed. Besides RU or Brown winning their Tourney, that's the other moving part. If Loyola doesn't win the Patriot League, then Loyola takes JHU's spot. Bottomline, there aren't a lot of moving parts.

If Towson loses to UMass, they're done. The committee likes these "you had your chance" scenarios. They won't see JHU having to beat UMd and PSU as a "you had your chance". JHU losing to UMd twice will barely effect JHU's stature.

[Homer]

Historically the committee does not give at-large selections to teams with RPIs below #12 or so. Hopkins is #11 currently and unlikely to fall below #12 even with a loss to Maryland. Based on LaxBytes I don't see how Denver can finish better than #14 if they don't win the Big East. Hop's average RPI win, average RPI loss, and SOS are all much better. Hop plays a vasty harder schedule, which would account for the worse record, and would have the best win of the bunch over Maryland. Obviously I'm advocating for my guys here but as best I can tell objectively, the committee will favor a resume like Hop's over Denver's if it comes down to that.

Oh, I get it that there's a strong case to be made for Hopkins > Denver as things currently stand.

I just think at this time of year -- especially nowadays that people work so much with predictive formulas -- we sometimes don't distinguish enough between "Team A and Team B are really close, but A is just slightly better on every metric so there's no case for B" versus "Most people looking at A and B would put A clearly ahead, but it's just possible a reasonable person could see it the other way depending on which metrics they choose to stress."

I feel like the Hop-Denver comparison is closer to the latter than the former.

[admin]

..."Best 5 wins" is a very reasonable way of looking at things, but doesn't correspond to how things are actually set up for the committee...

Homer, not sure if the "5 wins" were directed at us. If yes, we're looking at more than that. Along the same lines, Our RPI, which is a relatively simple formula so I think we have it right, has JHU ahead of Towson.

[Hawkeye]

..."Best 5 wins" is a very reasonable way of looking at things, but doesn't correspond to how things are actually set up for the committee...

Homer, not sure if the "5 wins" were directed at us. If yes, we're looking at more than that. Along the same lines, Our RPI, which is a relatively simple formula so I think we have it right, has JHU ahead of Towson.

I think it was directed at me since I listed them out in my post. It was moreso for reference/a refresher for everyone as to who each team had beaten than me saying "this is a metric that matters"

[HopFan16]

Again agreeing with Hawkeye, at least moderately so, in this case over Admin. A Towson loss to Umass in the CAA finals, combined with a Hopkins loss to Maryland in BIG round 1 would leave Towson with a H2H win over HOP as well as (I believe but am not totally sure) a higher RPI than the BlueJays. The RPI difference won't be more than 1 in any case and the committee might well (and I think they should) include a 10-5 Towson over a 7-7 Hopkins given that scenario. But hey, I'm not on the committee.

Agree with this, which is why if Hopkins doesn't beat Maryland it should root for Towson to win the CAA.

If Towson wins the CAA (and Loyola wins the Patriot), then I think both Cornell and Hopkins get in.

Hopefully the Jays beat the Terps again and none of this matters!

[Homer]

If Towson loses to UMass, they're done. The committee likes these "you had your chance" scenarios.

This would be the same committee that put Villanova in after they "had their chance" in the BE semis and lost to G'town a second time? Or maybe you're thinking of the committee that put Hopkins in after they "had their chance" in the B1G semis in 2016 and lost again to Rutgers?

I think maybe you like "had your chance" scenarios a little more than they do.

[Hawkeye]

Cornell

Could you give some insight on why your model thinks Cornell's at-large chances are so low? It's the one thing that your model seems to have very different than most.

[admin]

JHU lost to Towson in the first game of the season and just beat UMd, their biggest QW, in their last. Started weak, finishing strong.
Towson's biggest QW is against Loyola (in the 4th game of the season) and the next is JHU (in the first game of the season). Started weak, finishing weak(er). And finishing the season with a loss to an off-radar UMass... I dunno.

[Hawkeye]

JHU lost to Towson in the first game of the season and just beat UMd, their biggest QW, in their last. Started weak, finishing strong.
Towson's biggest QW is against Loyola (in the 4th game of the season) and the next is JHU (in the first game of the season). Started weak, finishing weak(er). And finishing the season with a loss to an off-radar UMass... I dunno.

There are certainly people out there who know more about the selection process than I do who may correct me here...

I am under the impression that timing does not matter to the committee - that all games count equally, regardless of their temporal placement.

[admin]

... a reasonable person could see it the other way depending on which metrics they choose to stress."

Homer, I hear you. With this said, the only metric I see Denver ahead of JHU on (besides numerous Rankers and Computer Rankings) is their Winning Percentage, #16 vs. #30 or 69% vs. 54%. And, since JHU's SOS is ranked #2 while Denver's is #17, I don't think they'll win on that one either.

[Homer]

JHU lost to Towson in the first game of the season and just beat UMd, their biggest QW, in their last. Started weak, finishing strong.
Towson's biggest QW is against Loyola (in the 4th game of the season) and the next is JHU (in the first game of the season). Started weak, finishing weak(er). And finishing the season with a loss to an off-radar UMass... I dunno.

I cannot emphasize enough that recency/trend of results has never* mattered one iota to committee selection. Not understanding this is the single biggest cause of confusion/shock/outrage every single year when the brackets come out. It should even be a sticky thread at the top of FanLax: IT MAKES NO DIFFERENCE WHEN THE GAMES HAPPENED. February is the same as May -- not in reality, but in the selection process absolutely.


* By never I really mean since 2011 or so when they fixed the process somewhat; before then things were weirder and it was sometimes hard to figure out exactly what they were going on. Also, just because it never happened before doesn't mean they couldn't suddenly change, but it'd be an explicit departure from the stated criteria and a dramatic departure from past practice.

[admin]

This would be the same committee that put Villanova in after they "had their chance" in the BE semis and lost to G'town a second time? Or maybe you're thinking of the committee that put Hopkins in after they "had their chance" in the B1G semis in 2016 and lost again to Rutgers?

When Villanova lost to G'town, they went to 10-5 and had beaten several good teams: Yale, PSU, Penn... A different comparison.

[admin]

In terms of timing, I hear you. Though... I'd have to look to find examples i seem to remember cases when teams were in the yes-maybe-no zone and they lost in their Conference C'ship and "worse" teams supplanted them. This isn't a 'when the game was played' issue but more a 'you had your chance', 'you're limping', scenario. Again, I'd have to look but I think Bucknell last season was one of these.

[Homer]

... a reasonable person could see it the other way depending on which metrics they choose to stress."

Homer, I hear you. With this said, the only metric I see Denver ahead of JHU on (besides numerous Rankers and Computer Rankings) is their Winning Percentage, #16 vs. #30 or 69% vs. 54%. And, since JHU's SOS is ranked #2 while Denver's is #17, I don't think they'll win on that one either.

If you're more of a line-up-their-wins type of person in terms of how you look at team resumes, you might see it like: each has a win over UNC. DU beat Towson and lost to Princeton; JHU beat Princeton and lost to Towson. So all of that cancels out and you're left with Denver's win over #17 Georgetown as the main differentiator. You could look at Denver's better record vs. a lower SOS and Hopkins' worse record vs. higher SOS as basically even, but DU shows somewhat better against top-20 competition. Plus some people may find a 7-7 record inherently gross and might be looking for a plausible case for another team, which I think DU at least borderline has.

[admin]

Homer, what you're saying is true. One could make that argument.

[calourie]

Timing of wins has been a factor sometimes and sometimes not. The ever changing committee is left to its' own discretion on a yearly basis. We will all know in 6 + days the how and why of whatever the decisions are they make in its' present iteration. There will always be some squawking no matter what they come up with.

[admin]

We will all know in 6 + days the how and why of whatever the decisions are they make in its' present iteration.

You're an idealist. Our only squawking, in general, is a lack of objectivity and transparency.

[Mr3Putt]

This isn’t a , I think this, I think that. Rpi, SOS, Head to head then they make a few exceptions. Towson has a higher RPI, SOS then Hopkins & Cornell. Cornell has win over Towson, Towson w a win over Hopkins. If Towson beat Loyola on New Years Day it’s still the 2019 season. Umass is very good. Trust me their goalie could do cartwheels over any goalie in the BIG10, ACC, or IVY. Towson has to post this week and certainly not perfect. It will be a close call!!

[laxreference]

Cornell

Could you give some insight on why your model thinks Cornell's at-large chances are so low? It's the one thing that your model seems to have very different than most.

Great question. It's 2 parts really.

First of all, they play Yale, and the Lax-ELO ratings suggest that Yale wins that game 71% of the time. Since my model looks at each team's projected resume at the end of the year, 71% of the simulations have Cornell with a loss to the Bulldogs and the resulting RPI implications of that. It's worth noting that the projections have Cornell in 76% of the time if they beat Yale. So it's not that it's down on Cornell's chances of getting in; it's more than it's not expecting them to beat Yale and it's down on Cornell's chances assuming they lose that game.

But that pales in comparison to the 2nd point, which is that the model currently expects them to be the last team in. Hopkins, by virtue of their win over MD, jumped over Cornell in the RPI standings. Since the model expects both JHU and Cornell to lose their next game, it assumes that Cornell would remain the last team in. I'm relying heavily on RPI in the model, which may be a mistake.

Which brings me to the main point about why my projections for Cornell are lower than most other projections. If Cornell is indeed the last team in, then they only get in when there are no bid-stealers. This year, there are relatively few conferences with the possibility of a bid-thief, but "few" does not mean "none". You may disagree that Cornell is the last team in, but my model assumes they are. So that means that if any of the below scenarios happens, Cornell is out:

Loyola loses the PL (34.7% chance)
Towson loses the CAA (69.4% chance; and RPI suggests that they would probably get in over a Cornell team that loses to Yale)
Brown wins the Ivy (4.8% chance)
Rutgers wins the Big Ten (4.1% chance)

Thinking back to our probability lessons, since all those events are independent the chance that none of those things happen is (someone check my math to make sure I'm explaining this right - looking at you CU77):

(1-.347)*(1-.694)*(1-.048)*(1-.041) = 18%

So a Cornell team that loses to Yale is basically hoping for a sequence of events that happens only 18% of the time. Even if you were just talking about the Big Ten/Ivy events, there is a 10% chance that at least one of Rutgers/Brown with their tournament.

So bottom line, I suspect that people are underestimating the chances that there is a bid-thief out there, waiting to reveal themselves. Since the model accounts for that, that's how you end up with so many simulations where Cornell gets left out of the field.