The Importance of Goal-Kicking Accuracy

So far this season, eight teams have lost after generating more scoring shots than their opponents and three more have been defeated despite matching their opponent's scoring shot production, which means that the outcome of over 15% of games might this year have been reversed had the losing team kicked straighter.

This fact has not been lost on coaches, who've often cited this factor in post-game pressers as a contributor, sometimes a major one, to team losses. 

Such practical demonstration of the importance of goal-kicking accuracy leads naturally to a desire to quantify that importance. That's what I've done for today's blog using the same model of team scoring that I discussed in this blog from 2014 and using the same assumptions about feasible Scoring Shot scenarios that I used in this more-recent blog (essentially, that both the Home and Away teams are assumed to generate Scoring Shots in the 15 to 40 range subject to the constraint that, between them, they generate at least 40 and no more than 55 Scoring Shots).

Now the base scoring model has the Home team and the Away team converting Scoring Shots at the historically-estimated level of 53%. We'll model the influence of variability in goal-kicking accuracy by considering that base model and three variants. Specifically, we'll model the full range of feasible Home team and Away team Scoring Shot scenarios with:

• The Home team converting at 50% and the Away team at 53%
• The Home team converting at 53% and the Away team at 53%
• The Home team converting at 56% and the Away team at 53%
• The Home team converting at 59% and the Away team at 53%

For each of these four combinations of conversion rates we'll generate 10,000 games for every feasible Scoring Shot scenario. The chart below then, which summarises the outputs, is based on almost 50 million simulated games.

View fullsize

Each curve tracks the estimated probability of a Home team victory as we vary the difference in the expected number of Home team and Away team Scoring Shots. On the left are scenarios where the Home team is expected to generate fewer Scoring Shots than the Away team, and on the right scenarios where the opposite is true.

As we jump from one curve to another we're mapping the relationship between the difference in expected Scoring Shot production and the Home team victory probability for different Home team Scoring Shot conversion rates. The red line, for example, tracks this relationship for a Home team that converts at only 50%, while the purple line does the same thing but for a Home team that converts at 59% instead.

At a macro level, the chart reveals that:

• Converting at a higher rate lifts the Home team's probability of victory for any given expected Scoring Shot differential (as we'd very much expect).
• The lift in the Home team's probability of victory in absolute terms is greatest for those games where both the Home and the Away teams are expected to generate the same number of scoring shots. In a game, for example, where the Home and the Away team are expected to generate 25 Scoring Shots, a Home team converting at 50% is expected to win only 45% of the time, while one converting at 59% is expected to win almost 58% of the time.
• The benefit remains significant even for games with moderately non-zero differences in expected Scoring Shot production levels. For example, in a game where the Home team is expected to generate only 18 Scoring Shots to its opponent's 28, converting at 50% implies a 12% probability of victory while converting at 59% lifts that probability to 19%.

Viewed in isolation, a 7% point increase in the probability of victory might not seem much, but if such a fillip could be achieved in each game of a full 22 round home and away season, that's an increase in the expected number of victories of 1.5 games. That, for most teams, matters a lot.

Another interesting way of using the chart above is to estimate the rough equivalence between Scoring Shot production and conversion rate, which you can obtain from the horizontal distances between the lines for a given probability.

So, for example, a Home team can achieve a 50% victory probability by:

• Generating, on average, about one more Scoring Shot than the Away team if the Home team converts at 50% (and the Away team at 53%)
• Generating, on average, about the same number of Scoring Shots as the Away team if the Home team converts at 53% (and the Away team at 53%)
• Generating, on average, about one Scoring Shot fewer than the Away team if the Home team converts at 56% (and the Away team at 53%)
• Generating, on average, about two Scoring Shots fewer than the Away team if the Home team converts at 56% (and the Away team at 53%)

The trade-off 1 Scoring Shot equals approximately 3% points of conversion holds for a wide range of Home team probabilities.

SUMMARY

If we accept that the scoring model used here - including the parameter estimates derived from 2006 to 2013 data - is a reasonable approximation of football reality, then the chart above can be used to estimate how a team's victory probability varies with changes in its goal-kicking accuracy, assuming it faces an opponent with an "average" level of accuracy.

That chart suggests that small but practically significant benefits would accrue to a team able to consistently kick more accurately than the average team such that it could afford to produce 1 fewer Scoring Shot for every 3% point increase in Scoring Shot conversion while still maintaining the same probability of victory.

(Note: Those of you with long memories might recall that I performed some calculations similar to those reported here, but using a much simpler model of team scoring over five years ago in this blog. The results then were similar though less-attractively illustrated than now. Progress can be measured in a variety of ways.)