Read on Twitter
So a couple of months ago I wrote this thread about our beloved Insight product at @Twenty3sport . Well, this week its getting a major revamp, and I hope you’ve all got 5 minutes so I can tell you about it. https://twitter.com/dperdomomeza1/status/1234949388147056645
Mark ( @EveryTeam_Mark ) has already written a couple of pieces on Insight which you can check out here:
https://www.twenty3.sport/inside-the-upgrade-of-twenty3s-insight-tool/
https://www.twenty3.sport/discovering-footballs-data-driven-stories-with-the-twenty3-insight-tool/
https://www.twenty3.sport/inside-the-upgrade-of-twenty3s-insight-tool/
https://www.twenty3.sport/discovering-footballs-data-driven-stories-with-the-twenty3-insight-tool/
In a nutshell, Insight is about automating finding interesting or noteworthy narratives in the data for the user. The placeholder name when we first started working on it was “Story Mining”, which might give you an idea of how we were conceiving it.
That in itself is a cool concept, and I don’t think anybody else does anything similar in the football data space. But as with most cool ideas, the devil is in the implementation. Does it do justice to the initial objective?
I remember my boss asking me: “Will this actually be helpful to people? Will they really find interesting stories, with enough breadth and frequency to cover the needs of their job day in day out - or will it simply be more of a luxury?”. Let me explain why I answered “yes”.
Lets suppose that I, as an opposition analyst or a journalist, am interested in Arsenal today. I go to Insight and I type “Arsenal” in the search bar, and hit enter.
Lets click through some of these stories. Pepe isn’t dribbling as much lately
BTW, you’ll notice in all my screenshots below that I’m cutting off the “additional context” to fit into the screen, but it gives you extra info like this:
At the back, David Luiz is seeing more of the ball in first halves than second halves. Mustafi shows the same pattern.
Ceballos does more defensive work at home than away. In fact, Arsenal as a whole do.
The recent switch in formation shows Xhaka having less of the ball. Related, we see that with their recent run of away games playing in a diff. formation, whilst Xhaka was the main outlet in their 4231, the ball travelling across the back and out via the right is more common now
David Luiz’s pass distribution sonar tells the same story - he’s distributing towards the right much more away (I think the fact that they’ve played so many away games recently is also why this story shows up as a home v away thing).
Saka is finally playing in different positions than his makeshift left back role of most of the season.
As such, he's not doing as much defending anymore:
Finally, Arsenal fans are somewhat excited with a recent upturn in form with the change of formation to 3 at the back, but the should take it with a pinch of skepticism: They’re outperforming their goals per xG comparatively to before.
...
As well as their goals per shot if you’re not into xG (I suspect you are if you made it this far into the thread).
Remember, I didn’t do any research here. I only typed “Arsenal”, hit enter, and had all this information 1 minute later.
I answered “yes” to my boss because the scale of what we’re doing here, looking for statistically significant stories across home/away, formation, over time, whatever, is just massive - massive enough to ensure we’re finding everything there is to find. At scale. Automatically.
And that is the nature of the revamp we gave the tool - rethinking our concepts and implementation into this super-powerful generality. I’m extremely proud of what me & the team ( @girela_d, @EveryTeam_Mark and @ewheatcroft94) have achieved here - trust me, it was very challenging
I hope those of you who’ll get your hands on this quickly find its not a just a luxury. And those of you who haven’t yet, make sure to reach out - but hurry up, limited stock only (just kidding, unlimited stock)
