Here's the latest outlook. We're sitting in 5th place by mean seed, but not far behind PSU and MSU (less than half a seed line), with just over a 50% chance of a double-bye.
Also strange to observe that IU seems to have a weirdly good shot at a top-4 seed. The actual numbers there are very small - out of a million runs, they were the 4 seed 163 times (about 1 in 6,000) and the 3 seed only once. Comparing IU and, say Rutgers, I think we see this behavior because:
- Both teams sit at 9 losses, so their best finish would be 11-9
- Both teams would need 2 or 3 of the 4 11-6 teams to finish 11-9 in order to get a 4 or 3 seed.
- For IU, those two things are correlated — they play both Illinois and Wisconsin, so an 11-9 IU means a higher likelihood of other 11-9 teams.
- For Rutgers, those two things are uncorrelated — they play Maryland and Purdue, so an 11-9 Rutgers means nothing about 11-6 teams falling to 11-9.
- The lesson — correlation is important!
(obviously those scenarios are
very unlikely to occur, but still fun to think about if you like stats)
View attachment 5769
We've been talking about tiebreakers a lot this week. The next two charts show 1) what the forecast would look like if all ties were broken by coin flip, and 2) the difference between the real situation and that scenario (where positive indicates better odds with the
real tiebreakers).
As you can see, we do slightly better in the coin flip TB scenario, with about a 60% chance of a double-bye. The B1G tiebreakers are lowering our shot at the 1, 3, and 4 seed, but helping us out at the 2. This is, again, averaged over a million simulations, so that's not to say there aren't times when we're tied for 2nd and tiebreakers put us 3rd — just that, on average, tiebreakers are tending to bump us towards the 2 and 6 seeds and away from the 3 and 4 seeds.
View attachment 5770
View attachment 5771
Some other notes:
- Non-conference record was used in 269 of 1 million simulations, or around 1 in 4,000.
- I'm now counting championship tiesituations as well as total ties. We have about a 20% chance of a tie at the top, which breaks down into:
- 2 teams - 15%
- 3 teams - 4%
- 4 teams - 1.3%
- 5 teams - 0.5% (1 in 200)
- 6 teams - 0.03% (1 in 3,000)
- In terms of total ties, it looks like there's still a very remote chance for a 7-team tie, about 1 in 10,000. The simulation did not encounter any ties with 8+ teams, but that doesn't mean they're not still possible.
- Coin flip tiebreakers still seem to be in play, but only happened 4 times in a million simulations.
- A note on methods: With 20 games left, there are just over 1 million ways the remainder of the season could go. At this point it starts to be feasible to just check every scenario by going down the list. My code doesn't have a switch to go to that approach, though — it just simulates the remainder of the season N times, which should still produce something fairly close to the truth if N is large enough.
- Fun fact — assuming every game left has a 60% favorite and 40% underdog, 1 million simulations still only have a 1% chance of hitting the "all underdogs" scenario, which is over 3000 times less likely than "all favorites".