Getting two teams through a game

The last update was about making the players less interchangeable. I could have kept tweaking that generator forever. At some point, I had to give the players a game to play.

So I started connecting it to the Game Simulation Engine. One matchup was enough for now. Thinking about a whole league before I could make sense of one game would have been a very efficient way to overwhelm myself.

Naming it an engine was the easy part. I had mistaken a filename for a qualification.

What changed since the last post

Player data moved into team rosters. I was working with two sides in a matchup, rather than checking generated players one at a time.
I started the single-game simulation loop. The immediate goal was to get those rosters through one football game and produce a result I could examine.
Ratings started needing a football meaning. I began working through how the differences from the player generator should affect what happened in a matchup.
I looked past the final score. Player contributions and the route to a result became part of what I needed to inspect.
The two engines started feeding work back to each other. Trying to use a player field in a game showed me where the generator was still making assumptions I hadn’t thought through.

Getting to kickoff

Prototype captures of the matchup and the game itself.

Earlier pregame view showing the matchup, team emblems and coach portraits. — The earlier pregame screen. Two teams, two coaches, and a game waiting to start.Open full screenshot

Live playoff screen with a scoreboard, game clock and play commentary. — A later playoff game in progress. Having somewhere to watch the scores and commentary made the simulation easier to follow.Open full screenshot

A score wasn’t enough

A better roster should usually have an advantage. It shouldn’t get handed a win. I understood that as a football fan; turning it into something useful in code was another job entirely.

The temptation was to see a plausible score and call the run a success. But a believable final number could hide some very questionable decisions underneath it. I needed to look at how the players contributed, especially after changing who was on the team.

This was progress, but I was nowhere near ready to declare that I’d cracked football simulation. Getting a result gave me something to investigate. It didn’t tell me the result deserved to be trusted.

That’s where the next stretch went: working out why a game ended the way it did.