smarterscout SPECIAL REPORT: What can data say about black coaches?
Despite the presence of players of virtually all ethnicities in professional football, non-white coaches – and especially black coaches – are startlingly underrepresented. We wanted to spotlight the black coaches who have broken through to manage in the top tiers of the game, and of course we wanted to do it with data. What we found was encouraging, but our results may also suggest just how far the sport still has to go.
First of all, a disclaimer: We classified a coach as black if he had self-identified as black or had dark skin that would generally be called black in the geographic area where he coached. If we have misclassified anyone, we are truly sorry and will make any necessary changes as soon as possible.
The most damning statistic that resulted from our research was the simplest one. Across all the leagues that we track on smarterscout, going back to the 2017 or 2017-18 season, we found only 30 black coaches. That's right – at more than 1,000 clubs, over three complete seasons and more, just 30 black coaches. There are probably more whom we missed, but the number is downright appalling.
Among the black coaches we found, we wanted to examine a very specific aspect of their performance: whether they made the whole greater than the sum of the parts on the pitch. In other words, given the squads they had at their disposal, we tried to figure our whether their clubs' results were better or worse than we would have expected.
To remove luck from the equation as much as possible, we decided to measure these results in terms of expected goals for and against. For every match, we began by creating an expectation of the results using the players' performances in the previous season*. Because of our league adjustments, it didn't matter which league they had played in during that season. Then we compared the actual results to our expectations in order to see whether the coaches had over- or under-performed.
We've grouped the coaches into tiers based on the outcomes of our research. For each coach, we've reported the average difference they might have made to expected goals for (xGF) and against (xGA), and our levels of confidence that they were superior to all the other coaches in their leagues. At 50% confidence, we would rule out the notion of being better or worse with equal confidence. Higher numbers are better for xGF (more expected goals scored), and lower numbers are better for xGA (fewer expected goals conceded). We required a minimum of 13 matches – roughly a third of a season in most leagues – to report our estimates.
Tier 1
Costinha. He's a free agent now, but as coach of Nacional for 34 matches in Portugal's top tier Costinha appears to have boosted the club's attack and may have helped its defending a bit as well.
Primeira Liga (POR1)
+0.21 xGF, 100% confidence
-0.02 xGA, 60% confidence
Keith Curle. One of the pioneers in English football, Curle has average numbers for attacking but excellent numbers for defending.
League Two (ENG4)
+0.01 xGF, 56% confidence
-0.06 xGA, 92% confidence
Nuno Espirito Santo. He ran the table with Wolves in the Championship and continues to excel in the Europa League and even the Premier League, at least for defending.
EFL Championship (ENG2)
+0.14 xGF, 98% confidence
-0.28 xGA, 100% confidence
Premier League (ENG1)
-0.02 xGF, 47% confidence
-0.23 xGA, 100% confidence
Europa League (UEL)
+0.10 xGF, 77% confidence
-0.34 xGA, 99% confidence
Henk Fraser. He's one of the best coaches in the Eredivisie, at Vitesse and Sparta Rotterdam. The latter was his first big club as a player and returned to the top tier under his guidance.
Eredivisie (NED1)
+0.22 xGF, 100% confidence
-0.28 xGA, 100% confidence
Robin Fraser. It's early days for Fraser at Colorado, but so far, so very good – at least on the attacking front. For defending, he looks average.
Major League Soccer (USA1)
+0.16 xGF, 89% confidence
-0.02 xGA, 47% confidence
Alberto Gamero. Without a doubt, Gamero has the strongest numbers of any coach we examined; he's almost a guarantee to boost a Colombian club's fortunes.
Categoria Primera A (COL1)
+0.10 xGF, 100% confidence
-0.08 xGA, 99% confidence
Hernan Medford. Medford has been very strong for attacking and average for defending at Herediano and Cartagines.
Liga FPD (COR1)
+0.07 xGF, 87% confidence
±0.00 xGA, 53% confidence
Darren Moore. Some fans were surprised when Moore was replaced at West Bromwich Albion, and his performance at Doncaster has continued to show his quality on both sides of the ball.
EFL Championship (ENG2)
+0.14 xGF, 97% confidence
-0.04 xGA, 67% confidence
League One (ENG3)
+0.06 xGF, 77% confidence
-0.08 xGA, 81% confidence
Jose Morais. His time at Barnsley was much better for attacking than defending, and we don't have quite enough data to appraise his time at Karpaty, but with Jeongbuk he's really hit his stride.
EFL Championship (ENG2)
+0.23 xGF, 97% confidence
+0.13 xGA, 15% confidence
K League 1 (KOR1)
+0.37 xGF, 100% confidence
-0.19 xGA, 100% confidence
Daniel Thioune. Thioune was the first black manager to break through in the top tiers in Germany, and he brought Osnabruck to the 2. Bundesliga – which has been tougher than the 3. Liga.
3. Liga (GER3)
+0.04 xGF, 74% confidence
-0.13 xGA, 98% confidence
2. Bundesliga (GER2)
+0.01 xGF, 57% confidence
+0.07 xGA, 20% confidence
Tier 2
Hubert Bodhert. At Once Caldas, Bodhert has done wonders for the squad's attacking but comes up short for defending.
Categoria Primera A (COL1)
+0.05 xGF, 85% confidence
+0.06 xGA, 10% confidence
Omar Daf. By contrast with Bodhert, the Sochaux coach looks like a real talent when it comes to defending, not so much for attacking.
Ligue 2 (FRA2)
-0.23 xGF, 0% confidence
+0.19 xGA, 100% confidence
Alexis Garcia. We only have data from Garcia's most recent jobs in management, at Deportivo Pasto and La Equidad, with better numbers for defending.
Categoria Primera A (COL1)
-0.05 xGF, 18% confidence
-0.04 xGA, 74% confidence
Roger Machado. The Brazilian former fullback is another coach who looks much better on defending than attacking.
Brasilerão (BRA1)
-0.05 xGF, 13% confidence
-0.01 xGA, 62% confidence
Sandro Mendes. Mendes's time at Vitoria Setubal was just a touch below average in our numbers.
Primeira Liga (POR1)
-0.02 xGF, 44% confidence
+0.04 xGA, 39% confidence
Clarence Seedorf. The man who won the Champions League with three different clubs as a player had more flair for attacking as manager of Deportivo La Coruna.
La Liga (SPA1)
+0.26 xGF, 98% confidence
+0.09 xGA, 26% confidence
Lito Vidigal. He's already taken two jobs in Portugal this summer, but his tenure at Aves, Setubal, and Boavista makes him look pretty average.
Primeira Liga (POR1)
-0.01 xGF, 42% confidence
+0.01 xGA, 45% confidence
Patrick Vieira. The former Serie A and Premier League star performed much better in MLS than in Ligue 1, suggesting that the quality of coaches in Ligue 1 might be higher, or perhaps that the transition from Manchester City's Elite Development Squad to New York City was easier.
Major League Soccer (USA1)
+0.06 xGF, 87% confidence
-0.12 xGA, 88% confidence
Ligue 1 (FRA1)
-0.16 xGF, 0% confidence
+0.06 xG, 15% confidence
Tier 3
Sol Campbell. Campbell looked average for defending with Macclesfield but struggled through his short spell at Southend.
League Two (ENG4)
-0.22 xGF, 1% confidence
-0.01 xGA, 52% confidence
League One (ENG3)
-0.19 xGF, 4% confidence
+0.27 xGA, 1% confidence
Jimmy Floyd Hasselbaink. His time at Northampton seems not great.
League Two (ENG4)
-0.18 xGF, 2% confidence
+0.22 xGA, 0% confidence
Chris Hughton. Our only data on Hughton was from his time at Brighton.
Premier League (ENG1)
-0.18 xGF, 0% confidence
+0.12 xGA, 1% confidence
Michael Nsien. The Tulsa native has been in charge since before his hometown club changed its name.
USL Championship (USA2)
-0.03 xGF, 30% confidence
+0.09 xGA, 15% confidence
Chris Powell. We didn't have data on Powell except for his stint at Southend. As in Campbell's case, there may have been greater problems in the background.
League One (ENG3)
-0.05 xGF, 23% confidence
+0.07 xGA, 14% confidence
We also identified seven coaches with insufficient data: Alex Dyer, Thierry Henry, Radhi Jaidi, Vincent Kompany, Eddie Newton, Jamison Olave, and Franck Passi.
Important caveats go along with these results. We took the players' minutes as given, so a coach's ability to select the starting XI and make effective substitutions is not included here. Also, a coach who has younger players would usually see their performances improve versus the previous season, and our estimates would give him credit for that change. By the same token, our estimates for a coach with consistently older players might punish him for their natural declines in performance. And of course, there are many factors other than a coach's ability that could affect overall performance – problems at the club, position in the table, cup competitions, etc.
Moreover, our estimates are an attempt to measure only one aspect of a coach's quality. In addition to their tactical choices, most head coaches and managers also have substantial input in recruiting. We're not measuring their ability to find good players or stick to a budget, nor are we gauging their talent in helping their players develop or even in facing the media.
There is much more to the job than what we're measuring here, but we can say one thing that seems important. As a group, these coaches are on balance as good or better than the other coaches working in their leagues; indeed, there are twice as many who excel than there are with subpar results. This could corroborate a prejudice against black coaches, since they would have to offer better results to get the same jobs.
Someday, we hope that prejudice will no longer exist, and black coaches in the top tiers of global football will be just as likely to succeed or fail as anyone else. In the meantime, club executives ought to take a hard look at some of the coaches listed above – as well as the ones who still haven't gotten their chance.
* Technical note: For each league, we have several seasons of data. For each season, we used all the other seasons of data to measure the relationships between results and the historical performance of the players involved. Then we predicted the results of the season in question. So for season 2017, we would use the relationships from seasons 2018 and 2019 to predict the results. Then we compared the predicted results to the actual results. The differences – or prediction errors – were the basis for our measure of coaching quality. We averaged the differences for each match across our two measures of expected goals. Then we compared the differences for all of a coach's matches in a league versus all the other coaches in that league, using a test for statistical confidence.
