Obviously Cornell/Brown tomorrow HUGE (more on that below). Brown can clinch a 2 way tiebreaker vs. Cornell w/ win tomorrow. Dartmouth still holds 2 way tiebreaker over Brown for now. Princeton can clinch 2 way tiebreaker vs. Dartmouth tomorrow.
Leverage for Saturday’s Games
Brown 38% (56% if win, 18% if lose)
Cornell 34% (98% if win, 64% if lose)
Dartmouth 22% (92% if win, 70% if lose)
Princeton 15% (> 99% if win, 85% if lose)
Harvard 6% (9% if win, 3% if lose)
Leverage on games a team is not involved in:
Brown 6% (in favor of Penn over Harvard)
Harvard 5% (in favor of Princeton over Dartmouth) and 3% (in favor of Brown over Cornell)
Cornell 4% (in favor of Princeton over Dartmouth)
Princeton/Dartmouth 3% (each, in favor of Cornell over Brown)
Tomorrow, I’ll try and put together some info about the most common 2 and 3 way ties
When writing your post you can click the little cog icon and insert tables and do all sorts of formatting stuff using the toolbar. Let me know if you have any questions and we are happy you are sharing all this info with us!
Curious if you have any reads on why Cornell is favored to make it over Brown conditioned on a Brown win tomorrow? Weaker schedule, but they would need to outplay Brown by 2 games, or I suppose they could beat them out in some 3-way tie scenarios perhaps?
Your intuition that ties don’t favor Cornell when Brown is involved is the right idea. 2 games sounds like a lot but just using Torvik as an example here are the win % for each of their final 3 games, respectively.
Brown:
Cornell:
Even if Brown wins tomorrow and goes 1 game up + H2H tiebreaker, those Torvik win % imply the following:
Cornell (3-0) & Brown (0-3) [11.5%]
Cornell (3-0) & Brown (1-2) or Cornell (2-1) & Brown (0-3) [31.5%]
Cornell (2-1) & Brown (1-2) or Cornell (1-2) & Brown (0-3) [33.6%]
Same record or Brown better [23.5%]
So that’s already ~43% chance Cornell exceeds Brown on wins w/out needing a tiebreak.
The other variable here is what happens in a tiebreaker w/ Dartmouth. Right now, my numbers suggest that when Cornell/Dartmouth tie for any position w/in top 4, Cornell wins the tiebreakers 70% of the time. On the other hand, in Brown/Dartmouth head to head ties, Brown wins only about 52% of the time.
The intuition here is that Cornell has already secured a split w/ Dartmouth and then has done better against top teams (Princeton). On the other hand, Brown still has to win another game against Dartmouth to secure the tiebreaker, so any path where Brown beats Cornell but then loses to Dartmouth ends badly for Brown if it finishes tied w/ Dartmouth.
I said I’d throw something together on the most common ties so here we go. Ran the seems deeper than usual this AM (10,000 simulated seasons) to get enough data in the weeds of infrequent ties, and here are all 2 and 3-way ties that occur with at least 1% frequency, and then the breakdown of the ordering in those ties.
I’ve just summarized it for any tie in the top 4 places; some of these ties I suppose may cover cases where teams tie for the 2-3 seed (not very interesting) or 3-4 seed (interesting), in addition to the most interesting cases where a team/teams losing the tiebreaker are sent home.
To give some background on how to read this chart, Cornell/Dartmouth tie for a spot in the top 4 just under 18% of the time. In those cases, Cornell wins roughly two-thirds of the time.
Similarly, Cornell/Princeton tie 17% of the time, with Cornell basically guaranteed to win the tiebreaker. This is because if Princeton and Cornell end up in a two way tie, it’s almost certainly is because Cornell beats Princeton, and thus wins the H2H 2-0. If they end up in a tie with Princeton beating Cornell, Princeton would need to drop 2 games in the standings to Cornell across the other 3 games of the season, so not very likely (and even still Princeton might lose that tiebreaker).
The take home here is that Cornell in decent shape if it ends in tiebreaker scenario w/ Dartmouth or Princeton, and bad shape if it ends in tie w/ Brown. Princeton not in a great tiebreaker situation, w/ Brown or Cornell, slightly favored vs. Dartmouth, but also the cases Princeton is ending in ties are more often those 2-3 or 3-4 seed ones which don’t affect whether or not they make Ivy madness.
Finally, here are the 2 most common 3-way ties which happen at least 1% of sims.
I’m terrified of Cornell and the ILT runs through Yale one way or the other, so the #4 seed for Princeton isn’t the end of the world. If Pierce can get himself right by then, then they have a puncher’s chance.
Thanks for everything you do. That all makes sense from a modeling perspective.
Re: Brown/Cornell odds, my biggest issue is with the Torvik game-by-game win probabilities (which I know you have nothing to do with). If Brown wins today, they’ll have won 4 straight, while Cornell will have lost 5 straight. Given that context, it seems absurd that Cornell would be 84% against a frisky Penn team and 69% against Princeton—even if both games are at home. Likewise, Brown would seem to have higher odds in its 3 games than the model gives it credit for. Maybe I’m wrong!
Reminds me of last year when Brown covered the spread in 9 (!) straight games to end the season. Vegas and the models never adjusted.
I think it’s completely reasonable to wonder whether any model (Vegas, Torvik, KenPom, myself) updates fast enough for any particular team. These systems are generally optimized to minimize error in spread across all NCAA, rather than a particular team or league.
That being said, we have to be careful not to “overfit” mentally either. Across all of NCAA, even in a perfectly efficient market, you’d expect some teams to cover often and some teams to never cover. If you and 350+ friends flipped 30 coin flips each, one of your friends will get a run of 10 heads in a row (eg. cover the spread 10 times in a row even if it’s correct).
When we live in the Ivy League bubble, this of course seems to us like the model is wrong/not updating fast enough, but zooming out it could be still consistent w/ expectations across Vegas/NCAA. In reality, however, I think it’s pretty hard to know for certain. As a Yale fan, a hot Brown team scares me a great deal.
Another thing that I am not sure how well any model deals w/ is the back to backs, which are unique to the Ivy. An old analysis found that Ivy road back to back teams covered basically exactly 50%, and at the same rate as KenPom spreads, but it’s possible its different know that they are less common…maybe something for me to look into this offseason.
As an aside, KenPom has a old article showing that the counterintutitvely, the pre-season AP poll was actually a bit better for predicting NCAA tournament success than the final AP poll of the year, in part I think because of this mental overfitting. https://kenpom.com/blog/the-preseason-ap-poll-is-great/
Are the tiebreak percentages with Brown involved weighted significantly on a Brown win against Yale, as this is also the most straightforward way for Brown to get in such an advantageous tiebreak position? If Brown went 2–1 down the stretch, losing to Yale, and Princeton and/or Cornell went 1-2, what then separates those teams?
I think that’s definitely part of it. Conditional on Brown being tied w/ Cornell in the standings in a 2 way tie, there’s > 50% chance that Brown beats Yale (and thus would win the H2H 2 way tiebreaker).
Of Brown/Cornell ties ~18% go down to NET. As I mentioned in the M/W Playoff Scenarios thread these are just “broken” alphabetically in my code, (Brown > Cornell) but in reality it’d be next to impossible for Brown to surpass Cornell in NET. The 93% of tiebreakers Brown wins > Cornell in the previous post is thus a bit overestimate; it’s probably closer to 75%.
For any Brown/Princeton 2 ways ties which are more common, it’s a similar story, except that Brown can also win a 2 way tie at 7-7 where Brown goes 2-1 w/ loss to Yale, Princeton goes 1-2 w/ loss to Cornell, and Brown wins because it will be 1-1 vs. Cornell (Princeton 0-2).
Finally, if (again, big if) Brown/Dartmouth end in a two way tie, I know the previous post says 100% to Brown; again about 12% of the time it goes to NET. Brown and Dartmouth are only separated by 2 places so it is very much in the air. In every other case, Brown wins the tiebreak without NET. In order to tie w/ Dartmouth in a 2 way tie, there are only 2 real possibilities
Tie at 8-6: Brown goes 3-0 (including wins vs. Yale and Dartmouth), Dartmouth either wins vs. Harvard or Yale. If against Harvard, Brown wins by having win against Yale. If both beat Yale, then Dartmouth’s loss to Penn comes back to bite it in the ass.
Tie at 7-7: Dartmouth loses out (including to Brown) and then Brown wins by virtue of better record vs some team (either Yale, or Penn).
All this to say, Brown is in the mix if it can get into some 2 way tiebreak w/ some teams. But even getting into a tie is going to take some difficult wins + some help.
One other thing that’s fun if you enjoy these kinds of things is that no team can be eliminated on Friday. Namely, the Harvard/Brown loser on Friday still has some hope on Saturday (though they surely need to win out + get some help).
Dartmouth can clinch a playoff spot w/ a win, and can clinch avoiding Yale in R1 w/ a win + a loss by Cornell & Princeton.
thanks for the very comprehensive explanation. I did not know that the Ivy was using NET as a tiebreaker, but I guess I should have figured that out. Hopefully Brown takes care of business + gets help elsewhere so that a loss to Siena isn’t the reason why we’re locked out lmao.
Also unrelated but what do you think of the WAC’s metric-based seeding system. Seems like something more conferences will adopt in the future. Personally, I like placing tangible value on non-con games but at the same time the system feels too nebulous for the average fan to follow along with.
Princeton @ Columbia
Princeton 15% (95% if win, 80% if lose)
Brown 7% (19% if COL, 12% if PRIN)
Dartmouth 5% (98% if COL, 93% if PRIN)
Harvard 4% (6% if COL, 2% if PRIN)
Penn @ Cornell
Cornell 8% (98% if win, 90% if lose)
Brown 6% (19% if PENN, 13% if COR)
Harvard 2% (5% if PENN, 3% if COR)
Yale @ Dartmouth
Dartmouth 5% ( if win, 95% if lose)
Princeton 3% (91% if YALE, 88% if DART)
Brown 2% (14% if YALE, 12% if DART)
Brown @ Harvard
Brown 24% (24% win, <1% lose)
Princeton 9% (96% if HARV, 87% if BRWN)
Dartmouth 7% (98% if HARV, 91% if BRWN)
Harvard 5% (6% if win, 1% if lose)
Cornell 2% (98% if HARV, 96% if BRWN)
Kind of crazy that a Harvard win is more beneficial for Princeton/Dartmouth’s playoff odds than Harvard’s own chances, but I guess their baseline chance is so low at this point one win can only swing so much (even winning out, Harvard would need some help to get in)
Thanks for your response. It’s hard to zoom out when the team you’re rooting for is on a streak, but you’re right—randomness plays a huge role and we can never truly know how much.
Kind of funny that Brown is a game back with two to play yet still controls their own destiny not only to get in but even to get the 3-seed. (If they win out they’ll be at worst tied for third at 8-6, H2H would be even and they’d have best win over Yale.)
Yale could have a very interesting choice on the final day. I haven’t hashed out all the scenarios but…
They could rest/limit their starters to prevent injury and potentially eliminate Princeton. But doing so could allow Brown into the ILT. Playing the Bears at the Pizz is not a good scenario for a #1 seed.
OR
They could go all out and eliminate Brown, but that risks injury and could let Princeton into the ILT. Despite their Ivy struggles, they still have Lee and Pierce (and a week to rest the latter).
Summary
Dartmouth and Cornell each can clinch w/ a win tomorrow. Princeton clinches with a win and any win by Harvard/Columbia/Dartmouth. Dartmouth and Cornell each guarantees a spot in the 2/3 game w/ win. Princeton’s path to clinch the 2/3 seed game tomorrow is W + Columbia W + Dartmouth W.
Brown/Dartmouth game tomorrow is huge leverage. Brown still probably needs to win out to realistically make Ivy Madness. Given that the loser of the Cornell/Princeton game gets Columbia/Penn, respectively, in last game, rather than Yale, just getting to 8 wins is going to be a lot easier for them than for Brown.
If Brown finds itself in 2 way tie w/ Dartmouth, it will win that tiebreaker (up to some small chance of it going to NET). If Dartmouth wins tomorrow, they can not mathematically end up tied.
Same w/ Brown/Princeton; if they end up tied either Brown won out (holding Yale win as trump card) or Princeton lost out (and Brown’s Cornell win is the trump card). As such, Princeton has a strong rooting interesting in favor of Dartmouth tomorrow.
Dartmouth/Brown
Brown 53% (56% if win, 3% if lose)
Dartmouth 35% ( if win, 65% if lose)
Princeton 17% (97% if DART, 80% if Brown)
Cornell 4% (> 99% if DART, 96% if Brown)
Harvard 3% (3% if Brown, < 0.1% if DART)
Princeton/Cornell
Princeton 10% (> 99% if win, 90% if lose)
Cornell 7% ( if win, 93% if lose)
Brown 3% (21% if Cornell, 18% if Princeton)
Yale/Harvard
Brown 10% (20% if Yale, 10% if Harvard)
Harvard 9% (9% if win, if lose)
if win, 95% if lose)
if lose)
