For NFL football, our simulator needs quite a bit of information. This information is provided by Run The Sims, based on years of experience and tweaking projection models. You, however, have the ability to change many of these inputs in the DIY Simulator HQ app. Let’s talk about the information that the simulations need.
First, the simulator needs to know the pace – that is how many plays we expect each team to run on average. We also need to know each team’s run-pass split. The simulator will take this info and inject a first level of randomness to establish the general flow of the game. Each simulation will begin with a new, randomly assigned degree of play volume and run-pass split, centered on our inputs. For our purposes, we do not include sacks, throwaways, spikes, or any other non-targeted throws in our pace and run-pass split calculations. We also consider quarterback scrambles as rush attempts.
Now that the simulator knows how many run plays and pass plays to assign to each team, we then start assigning opportunity based on rushing market share (RMS), target market share (TMS), and passing market share (PMS, sorry). The simulator randomly assigns a rusher to each team rush attempt based on the market share split provided. It does the same on pass plays by assigning a targeted receiver and a passer.
Once the rusher is selected, we then assign an outcome to that player algorithmically. The outcome of this play (typically yardage) is calculated based on a number of factors and does not allow for user input. What the user can control, however, is each player’s touchdown likelihood which is assigned through rushing and receiving touchdown market share.
That is about as much as I can tell you without giving away too much secret sauce. But basically, the simulator goes one play at a time through an entire game and then logs the results for each player at the end. So, after simulation number one – Patrick Mahomes was 35/48 passing for 312 yards, 3 TDs and 1 INT. Travis Kelce caught 7 of 11 targets for 119 yards and 1 TD. Etc, etc, etc. So all of this goes into a single simulation for a single game. We then run this entire process over again 9,999 more times, logging each player’s results in each simulation. We repeat this process for every game on the slate, which means we end up with a mountain of information that can help us make really intelligent and informed decisions.