What modern day football analysis can teach us about the importance of niche metrics

When I sat down to watch Match of the Day recently, eagerly looking forward to watching the highlights of my beloved Arsenal play Leicester City in the Premier League, I was struck by something the commentator said at the end of the game:

"It ended 4-2, but that only tells half the story".

I watched the full match earlier that day and my view was that the scoreline doesn't even tell us half the story - and to validate this, I looked at the data, of course.

Looking at the stats that flashed up on MotD, you'd be forgiven thinking this looks like a routine home win by Arsenal, leading all the metrics by some margin. The reality was that despite Arsenal's dominance, they were a just few minutes away from dropping valuable points.

The final score is like the primary metric and the match stats are like the secondary metrics, however, even with all this data, we do not have the metrics to really explain how the match panned out.

Arsenal absolutely dominated the game but needed a couple of fortunate injury time goals to win it after a lucky deflected header and an absolute worldie of a goal (XG: 0.04) brought Leicester level. The second half was peppered with a string of incredible saves by the Leicester keeper as Arsenal threw the proverbial kitchen sink at the Leicester goal.

Some of this can be explained in the stats MotD displayed after the game - for example it's clear Leicester were lucky to even score those two goals, out-performing their XG by bagging 2 goals out of an XG of just 0.27. You may come to the conclusion that the goal keeping played an absolute blinder based on the number of big chances he faced (10) vs number of goals Arsenal scored (4) which checks out in this case, but in other cases the strikers may have had an absolute stinker - without more granular data, we couldn't know for sure.

This chart helps us to understand the game states better and proves how improbable it was that Arsenal needed two injury goals to win a game they dominated so comprehensively.

Whilst Leicester were objectively lucky, Arsenal had their fair share of luck too. Both those late Arsenal goals were arguably lucky to find the net, with one being an own goal and the other coming as a consequence of a quick break whilst Leicester pushing high up the field looking for another equaliser and the the keeper (for once) pushing out a weak effort that resulted in the fourth goal. Furthermore, Arsenal's Italian defender, Riccardo Calafiori, was lucky not to be sent off in the second half after escaping a second yellow for a clumsy challenge.

Whilst there are analytical measurements such as game tilt, expected assists and Passes Per Defensive Action (PPDA) that might shed further light on the performance, it feels like we need a new metric that measures good fortune (xluck?), and if anyone figures how to measure that, I'm sure they'll make millions! Personally, I would also be interested in someone coming up with a measure for the level of stress each fan may feel (xBPM?).

I didn't track my heart rate during the match but I imagine it looked something like this:

The point is that even in Football, where advanced analytics and niche metrics have permeated throughout the game over the last decade, there are still gaps in our understanding due to not being able to accurately measure everything we'd like to.

Let's now look at how this problem can skew the perception of an individual players' performance.

For the entire game, Arsenal's defender, William Saliba, expertly marshaled the experienced Leicester striker, James Vardy, barely giving him a kick. Indeed, according to one pundit, "It was like watching a terrier chase a Ferrari for 90 minutes".

However, you would come away with a very different conclusion if you look at the Fantasy Football points he accumulated for that match - 0.

William Saliba was punished despite not being particularly culpable for either goal and picking up a harsh yellow card. This particular model isn't smart enough to assign points accurately based on actual performance - the 'clean sheet' metric is easy to measure but not necessarily an indicator of individual performances, yet defenders are nonetheless over-indexed on that metric.

The phrase 'underlying metrics' have become synonymous in this new world of football analysis:

As a club, it is a way to achieve marginal gains in ways not possible in the past. You don't have to look far for examples of clubs completely changing their fortunes by taking a data-driven approach to everything from recruitment strategy to tactics - most notably Brighton and Brentford - it's no coincidence both teams have owners who also own sports betting analysis companies.

As a player, you have access to data to optimise your performance from decision making to tactical awareness while recovery and load management data helps them avoid injury and fatigue.

As a fan, it's become a way of objectively (to some extent) articulating what we can see with our own eyes without the biases, tribalism and partisan nature of any competitive sport.

As a pundit, you now have an unprecedented amount of data and insights to give viewers a nuanced and sophisticated analysis of the match they're watching. Why so few of them chose to do so is an enduring mystery.

So how does this relate to business metrics?

You've no doubt heard the saying "What gets measured gets managed". It seems reasonable enough but scratch below the surface and consider whether what you're measuring is actually meaningful.

V. F. Ridgway published a paper in 1956 on this topic in which he summarised:

"Not everything that matters can be measured. Not everything that we can measure matters."

In business, it is very tempting to measure only what is easy to measure. Conversion rates, bounce rates, time on site are all such metrics. They're out of the box metrics that give you a meta overview of whats going on. It's the business equivalent of just looking at the final score or number of shots each team generated.

It doesn't give you anywhere near like the full picture of what users are doing on your site. To do that, you need to look at the underlying metrics, and not just any metrics...the right ones for your business. Discovering these metrics don't come in any standard reports dashboard; you're going to have to identify them yourself through research and critical thinking.

Some of these metrics may not be easy to get your hands on, and may require a lot of data architecture work to pipe all the data together. I see this a lot around metrics like Life Time Value (LTV) and other longer term metrics which are incredibly important to track and measure accurately but are often ignored due to the complexity involved getting that data.

A Goal Tree Mapping exercise is a great way to identify those niche metrics for your business and really get an understanding of your users behaviour. These niche metrics are the ones that you should obsess over; not a site wide conversion rate.

What are your equivalent of the modern analysis permeating throughout the football world? Get in touch if you want help in discovering niche metrics for your business.