Nobody ran into the gap

17 September 2026

Watching my own match engine play, something was off. The attack looked busy. Phase after phase, passes going wide, forwards carrying. And it never actually threatened anything. Sides were grinding out twenty phases and finishing exactly where they started.

My first thought was the obvious one: the defence is too good. Every defensive line in the game was holding, so the numbers behind the defence must be too strong. Turn them down and the attack breathes.

That would have been a bad fix, because it was the wrong diagnosis. Before changing anything I wanted to know one thing: how often is there actually a hole in the defensive line?

Two thirds of the time

So I measured it. Not whether anyone scored, not whether a break happened. Just the geometry. At every moment of open play, is there a gap in the defensive line wide enough for a player to run through, close enough to reach?

Inside the opposition forty, a runnable hole existed about two thirds of the time.

The defence was not too good. The hole was there, most of the time, and the attack was not going anywhere near it.

Why not

The reason turned out to be a single assumption buried in how defenders pick who to mark. The marking logic worked on the backs. Every back had an opposite number, and that number knew where he was and drifted with him.

Forwards were not in that system. A forward standing in the line had nobody formally assigned to him, which sounds like it should make him easy to attack. It did the opposite. Because the attacking side worked out its options from who was marked, an unmarked forward was not read as space. He was not read at all. The gap next to him was invisible.

So the attack kept doing the one thing it could see: carry into the man in front, set the ball, go again. Twenty phases of honest work and no threat, because the thing worth attacking was never on the list of options.

Why this matters for the game you will play

If I had trusted my first instinct and softened the defence, the matches would have opened up. Tries would have gone up. It would have looked fixed, and the sim would have become slightly more of a lie. Sides would be scoring because the defence had been told to be worse, not because anyone had beaten it.

What you want, when your fly-half spots a hole and goes, is for that to have happened because there really was a hole and he really saw it. That is a different game from one where tries fall out of a dice roll. It is the difference I am chasing.

This is one of roughly sixty of these. I have been measuring the engine against real match data for months, and about half of what I was sure about turned out to be wrong. Those are the interesting ones, and they will end up here.

Prefer a reader? RSS feed. All posts: the devlog.