Watching the European Cup is not great fun for those whose standards are the Barcelonas of this world. How could it? These are national teams barely playing together and most of the teams know something about defending and little about attacking.
In former times people always invoked the number of goalless draws or goals scored per match as indications of (un-)attractive matches. With the advent of big data, we all have become more sophisticated and talk about possession statistics, PDOs etc. One often heard metric is shots on target, i.e. how many shots either go into the net, or would hadn’t the keeper or a defender deflected or stopped the ball. Football experts everywhere now use it as a tool for bitching: in this game there wasn’t a single shot on target for X mins. Just as an example, yesterday’s match Portugal vs. Croatia only saw two shots on target and only very late in the game. While I agree that this match was a bit dull except for the extra time, using this metric also shows what’s dangerous in relying on one statistic only, and especially considering the arbitrary decisions that lie underneath it.
If you think about it: Isn’t a free kick hitting the bar or just going slightly wide much more exciting than a lame long-distance shot easily plucked out of the air by the goal keeper. As in all human activities there is a margin of error which is acceptable if you look not for the final result (a goal), but for the attractiveness of the game (number of exciting opportunities created). In this sense, the shots-on-target metric, as useful as it may be for other purposes is too harsh if you talk about how attractive a match is.
It’s also important for football data scientists to understand that the social use of statistics differs from the experts (nerds’?) intention. There may be ‘objective‘ problems with shot on target measures. But the real issue is a sociological, or perhaps even psychological one. What does the metric do to us, and how does it makes our perceptions of the game change? As with any metric used as a objective target it looses some of its value, once implemented.
I would therefore argue that a more sensible shots-on-target measure should change the definition: include a reasonable perimeter around the goal (say 1 extra meter, or use a standard deviation of the average difference from goal) which counts as effective shots on target. I think this would be a good metric to see how many times in a match a football fan’s blood pressure significantly rises. After all, this is the real measure we want to come close to with match statistics, don’t we?
If you don’t believe me you should watch this short motivational video.