In my recent post suggesting a new approach to squad usage, I referred to the benefits of squad stability – essentially, giving players a chance to build stronger connections during matches to improve performance. Because I’ve received some questions about it, I thought I’d offer a taste of some of the research that went into that recommendation.
Measuring the effects of squad stability is not easy, because there are so many countervailing factors. To understand what I mean, consider what would happen if a club used the same players again and again in the first half of the season. Would their performance in the second half be better than expected?
If the players got better at working together, it might. But if it just tired them out, it might not. Moreover, the fact that the club used the same players regularly might suggest that it was lucky in avoiding injuries; if the injury rate returned to a more normal level in the second half of the season, performance would be expected to drop. And it’s even possible that the club was using the same squad for league games while deploying another set of players for cup competitions; once the club exited those competitions, the manager would have more players to choose from for the league. In that case, performance in the league might improve during the second half of the season, but not because of squad stability in the first half.
All in all, research into the effects of squad stability at the macro level – the team as a whole – seems like a ship doomed to crash upon the rocks of causality and bias. But we can still look at the effects on the micro level. If the hypothesis that playing together builds connections is correct, then pairs of players who interact frequently should become more productive as they spend more time together on the field. And this is indeed what I’ve observed.
To measure these micro effects, it’s important to look at players who haven’t spent much – or indeed any – time in the same squad before. That said, I don’t want players who are new to the league; then any improvements in performance might be the result of getting used to the playing conditions. So for a given league and season, I’ve been examining pairs of players where one of them just joined the club after previous experience in the league.
A potential case in point is the pairing of Eden Hazard and Cesc Fabregas. Chelsea made headlines (at least among analysts) for its extreme squad stability in the English Premier League last season. Both players were on the pitch for more than 3,000 minutes in league play. Fabregas had played for Arsenal, so he knew the style and rigors of the Premier League, but his Chelsea teammates were new to him. In particular, he left Arsenal before Hazard even came to England. So what happened when they started playing together?
I used NYA’s non-shots-based model to measure the value of the passes from Hazard to Fabregas in terms of the expected goals generated, on average, by each one. The figures are lumpy, but the trend is pretty clear:
As the season wore on, the passes between the two players became progressively more valuable. And then I measured the average time between passes from Hazard to Fabregas. Again, there was a somewhat perceptible trend:
More time playing together apparently meant more frequent interplay between the two Chelsea stars. By the end of the season, 90 minutes of playing time would have led to about 27 passes at 0.005 expected goals per pass, for a total of around 0.14 expected goals just from passes from Hazard to Fabregas.
Naturally, there could be other explanations for these phenomena. Jose Mourinho may simply have instructed Hazard to pass more often to Fabregas, and to wait until the Spaniard was in a dangerous position before doing so. But the more I see patterns like this, the more I believe the intuitive point that prolonged interaction helps performance.
A caveat is that what’s true in one context may not be true in others. So it’s just as important to analyze other combinations – goalkeepers and defenders, midfielders sharing the center of the pitch, and perhaps groups of three or more players – to understand where stability is worth pursuing.
And as always, if the supposed benefits of stability don’t translate into better results, then it might be time to measure something else. The uptick in the productivity of the Hazard-Fabregas partnership also coincided with a downturn in Chelsea’s overall attacking production. Were Diego Costa’s absences (or better marking by the opposition) directly responsible for this trend? Or did the same factors force Fabregas to play further forward, automatically increasing the value of passes from Hazard? In any specific situation, it’s worth ruling out this kind of alternative hypothesis before drawing conclusions.