About a month and a half ago, I briefly covered how teams win points in Serie A. The common narrative is that defense reigns, and I wanted to dive into that a bit more. The goal of the first piece was to determine whether or not that the adage about defense winning championships is true. Through that analysis, I concluded that there isn’t a significant difference between goals forced and goals conceded in predicting match outcome. I wasn’t quite satisfied with the analysis, though, so I continued to push forward.
My first thought was to do a sensitivity analysis. In theory, this would tell us which input changed the output the most. If we could figure that out, we would know which input was most impactful on the point total – and thus most important overall. After spending some time thinking on it, however, I decided against doing this sort of analysis.
This analysis wouldn’t tell us anything about the specifics of the results. It would tell us which attribute was most sensitive, but a further analysis would have to be done. I thought there may be what I’ve started calling a break point – a point where value started changing direction. It seems obvious that at 0-0 there’s going to be more value in scoring and moving to 1-0 than there is in avoiding moving to 0-1. If that’s true, it may be true if we keep the goal difference constant (i.e. moving from 1-0 to 2-1). A sensitivity analysis wouldn’t tell us this, however, so I decided to create a table of the expected points from every combination of goals forced and conceded between 0 and 7 goals.
This table tells us a few things:
- The obvious is accurate. If a team is tied, there’s more value to be gained by for going the win than there is value lost by preventing the loss.
- There’s more value in winning by multiple goals. As goal difference increases, so does point expectations (up to a 5-goal win).
The table also tells us about strategy depending on game state. For example, if a team isn’t winning (i.e. the score is tied or the team is losing), there’s no point that scoring more goals doesn’t add value. The maximum point expectation in a draw comes at 7-7 (1.387 points) and the max point expectation in a loss comes at 6-7 (0.459 points). This tells us that teams playing from behind should absolutely value offense more than defense. That doesn’t mean teams should commit crazy numbers forward, or that they should bring back the 2-3-5. It just means that there should be an emphasis on scoring if the team isn’t winning.
However, the same is not true for winning scenarios. In these scenarios, there are actually multiple break points. The first occurs for teams winning by 1-goal. Up to 3 goals forced, there’s value in scoring more. That means if a team is up 1-0, there’s more value to be added by going for another goal, still. The same is true for a team winning 2-1. However, if a team is up 3-2, there’s more value in goal prevention. That’s our break point. From there, every additional goal forced and allowed results in fewer expected points – to the point that a team at 5-4 expects fewer points than a team at 1-0.
If a team is up by multiple goals, the break point is different. When winning by multiple goals, the break point occurs between 5 and 6 goals. At 5-3, a team would expect 2.759 points, whereas 6-4 expects 2.757 points. If a team is winning by 3 (or more) goals, that break point doesn’t occur (at least, not through 7 goals forced).
This is all to say that there is a point that defense is more valuable, but that point doesn’t come until a team has already allowed 2 goals. Similarly, there’s a point that offense is more valuable, but that point stops after a team has scored 3 goals. So long as a team can avoid extreme results, there’s very little value to be gained by emphasizing either side of the ball.