Still following this tutorial.

fn position_translation(windows: Res<Windows>, mut q: Query<(&Position, &mut Transform)>) {
    fn convert(pos: f32, bound_window: f32, bound_game: f32) -> f32 {
        let tile_size = bound_window / bound_game;
        pos / bound_game * bound_window - (bound_window / 2.) + (tile_size / 2.)
    }
    let window = windows.get_primary().unwrap();
    for (pos, mut transform) in q.iter_mut() {
        transform.translation = Vec3::new(
            convert(pos.x as f32, window.width() as f32, ARENA_WIDTH as f32),
            convert(pos.y as f32, window.height() as f32, ARENA_HEIGHT as f32),
            0.0,
        );
    }
}

The position translation: if an item’s x coordinate is at 5 in our system, the width in our system is 10, and the window width is 200, then the coordinate should be 5 / 10 * 200 - 200 / 2. We subtract half the window width because our coordinate system starts at the bottom left, and Translation starts from the center. We then add half the size of a single tile, because we want our sprites bottom left corner to be at the bottom left of a tile, not the center.