Thursday, February 6, 2014

Evaluating Creativity



                My last post asked the question whether many creative players, specifically Ozil and Mata, were overrated because the basic metric used to rate them (key passes) doesn’t take into account turnovers. We would really need to look at something like the assist/turnover ratio in basketball to appreciate the full value of a player’s creativity. I concluded that since we don’t currently have accurate data, it is really hard to tell whether Mata and Ozil are overrated. But today, I’m going to take a stab at it, using a relatively simple formula as a proxy.

                The assist/turnover ratio is brilliant in that you are comparing the number of positive creative events (assists) a player has with the number of negative creative events (turnovers) he has. To develop a proxy, we need to come up with a positive event and negative event that we can easily measure. In coming up with this formula, I stuck with the idea that the key pass should be the positive event we look at rather than an assist. Since scoring is much more infrequent in football, using key passes gives us more data and lets us focus on the value the passer adds (rather than the shooter). For the negative event, I looked at the total number of unsuccessful passes a player has (that is, (1-Passing Percentage) * Total Passes). This formula allows us to look at the key passes versus unsuccessful passes, and gives us a better idea of who the creative players truly are.

                There are some obvious flaws in this. Many of a player’s turnovers are not the result of a misplaced pass, but rather by being caught in possession of the ball, or stopped when attempting to dribble around a defender. The key pass metric itself is not perfect, and counts passes leading to a shot on target and of target equally (at least Squawka, my data source for this project, does). Still, there are some key advantages to using this metric when comparing players. It’s not a counting stat, where the more you play the better you do, so there’s no need to adjust for playing time. Unlike passing percentage, it doesn’t inflate totals due to making more simple, short passes. And most importantly, it has some way to measure mistakes and hold them against the player who makes them.

                 Using the proxy, I looked at all midfielders with more than 20 key passes on the season. The leaders of the Key Pass list contain some familiar names, along with a couple odd ones: Eden Hazard, Ozil, and David Silva are examples of the former, Moussa Sissoko and Robert Snodgrass the latter. How do they stack up in this new metric? See for yourself (data through 2/1/14).





Assist/Turnover Ratio Proxy (key passes/unsuccessful passes)
Player Team Key Passes Unsuccessful Passes A/T
Mirallas Everton 42 85 0.49
Mata Man United 32 71 0.45
Nasri Man City 43 97 0.44
Townsend Tottenham 21 55 0.38
Silva Man City 45 118 0.38
Willian Chelsea 27 72 0.38
Jarvis West Ham 22 60 0.37
Navas Man City 35 96 0.36
Downing West Ham 32 90 0.36
Lennon Tottenham 23 67 0.34
Snodgrass Norwich 46 138 0.33
Ward-Prowse Southampton 29 88 0.33
Hazard Chelsea 55 169 0.33
Ozil Arsenal 44 140 0.31
Sessegnon West Brom 23 74 0.31
Sterling Liverpool 24 79 0.30
Ben Arfa Newcastle 20 66 0.30
Dembele Tottenham 25 83 0.30
Sissoko Newcastle 49 169 0.29
Amalfitano West Brom 31 107 0.29
Giaccherini Sunderland 22 77 0.29
Ki Sung-Young Sunderland 22 81 0.27
Johnson Sunderland 29 107 0.27
Holtby Tottenham 20 81 0.25
Eriksen Tottenham 26 110 0.24
Redmond Norwich 22 94 0.23
Westwood Aston Villa 34 147 0.23
Henderson Liverpool 37 164 0.23
De Guzman Swansea 30 133 0.23
Gouffran Newcastle 20 90 0.22
Larsson Sunderland 26 119 0.22
Puncheon Crystal Palace 29 133 0.22
Bannan Crystal Palace 20 92 0.22
Gerrard Liverpool 35 162 0.22
Whittingham Cardiff 36 170 0.21
Cazorla Arsenal 29 142 0.20
Lallana Southampton 30 148 0.20
Oscar Chelsea 27 134 0.20
Cabaye Newcastle 31 156 0.20
Davis Southampton 28 143 0.20
Wilshere Arsenal 25 130 0.19
Noble West Ham 38 201 0.19
Pienaar Everton 26 139 0.19
McCarthy Everton 24 134 0.18
Routledge Swansea 22 124 0.18
Coutinho Liverpool 29 164 0.18
Adam Stoke 26 152 0.17
Brunt West Brom 23 136 0.17
Barkley Everton 21 126 0.17
Lampard Chelsea 21 128 0.16
Fernandinho Man City 22 142 0.15
Huddlestone Hull 37 239 0.15
Fer  Norwich 21 149 0.14
Ramires Chelsea 25 178 0.14
Shelvey Swansea 26 186 0.14
Livermore Hull 22 175 0.13
Barry Everton 22 190 0.12
Ramsey Arsenal 20 198 0.10
Jedinak Crystal Palace 21 278 0.08



Observations:
1.       Mirallas tops the leaderboard! Didn’t see that coming, but he has been having a great year. On the other hand, Ross Barkley is near the bottom. Barkley has been getting a lot of praise, but I think this adds to the evidence that he’s experiencing the “British media fawning over an average young British player” phenomenon.

2.       Note that true central midfielders tend to perform a lot worse by this metric, as you’d probably expect. Lampard, Fernandinho, Barry, Fer, Huddlestone, and Adam all have a reputation as good passers, but their central position does mean they pass the ball more often, and thus misplace more passes. This may also be because wingers/attacking mids are more prone to losing the ball in possession, something not taken into account in this metric.

3.       Team shot preferences do play a role here. For example, Spurs players will probably be overrated since they take so many shots form distance, a shot which usually doesn’t require as incisive a pass. This could be fixed by having key passes defined as passes leading to a shot on target, something that is much harder to do from distance. West Ham are another team whose players might be similarly affected.

4.       Two that did better than I expected: Juan Mata and Willian. Hazard is the name people talk about in the Chelsea attack, but for all his speed his passing leaves a lot to be desired. Willian has both attributes, and is a relatively unsung player at this point in time. I thought Mata might drop a bit by this metric, but he did quite well. If you give him a free role, he can be very dangerous. Hopefully United will screw him up like Kagawa and Anderson.

5.       Aaron Ramsey is an odd case. Second from the bottom on this list, but he has six assists. His teammate, Mesut Ozil, does well in this metric but not spectacular. I think this provides a more appropriate reflection of his season.