I just recently launched another fun interactive chart for your enjoyment, my 2016 season projections. These projections track each team's overall record, conference record, and per game scoring. The projections are meant to give people an idea of how much uncertainty there is in my model's ability to accurately predict season-long outcomes, because merely summing per game win percentages to arrive at a single projected win-loss record can be slightly misleading.
The model runs 10,000 simulations of the season, and tracks the final results for each trial run. I am running my model "hot," meaning each team's rating is updated after each simulated game within a simulated season. This means that in one of my simulations, St. Thomas was able to fall from the #2 team in the country to 4-6 because of a string of bad luck early in the season. By running the model in this fashion I'm able to more accurately display the inherent uncertainty in a preseason rating.
The graphs you see in the interactive graphic are histograms (they're actually bar charts, because I'm using a free web hosting service and it crashes when I try to upload 10,000 full simulated seasons, but they're essentially histograms). The tallest bar of each graph represents the probability of that variable occurring throughout the simulation. The blue bars are good things: wins and scoring points. The orange bars are bad things: losses and allowing points. More blue on the right and orange on the left is a good thing, while orange on the right and blue on the left is a bad thing. If the chart looks almost entirely like a big brown bulge (look at Simpson College's projection for an example), the model thinks your team is pretty average relative to their schedule. The relative height of the bar is the singular probability of that outcome on a 0-1 scale. If the bar's height is 0.25, that means there's a 25% probability of that outcome. You can also add the total heights of a series of bars to find the probability that a team finishes with at least that many wins, or averaging at least that many points. Want to know the odds Carnegie Mellon finishes at or above 0.500? Add the height values for every bar from 5-10. Every win-loss graph is also going to be symmetrical. For a team that plays in 10 DIII games, the height of the bar for nine wins is going to be the same as the bar for one loss.
My goal is to update my projections every week of the season, in addition to my regular ratings. But I do have a full-time job, and a girlfriend who isn't exactly thrilled by the prospect of me spending two hours on my laptop every night after work, so we'll see how streamlined I can make this process in the future. I hope you enjoy the projections!