Wednesday, January 14, 2015

Goalkeeper Stats Are (Mostly) Stupid


DC United's Bill Hamid had a great year in 2014, helping to lead a resurgent season for Ben Olsen's club.  Hamid led MLS in Save % and deservedly won the MLS Goalkeeper of the Year Award.  The other goalies receiving consideration also were near the top of the league in Save % (poor Jon Busch).

Tracing back through the recent history of the award this same trend can be seen: the "best" goalies have the highest Save %.  This seems reasonable; most of a goalie's worth to his team is measured by how good they are at stopping shots on target from reaching the back of the net (measured as Save %).  Therefore, it is intuitive that the "better" the goalie the higher their Save %.  If this is the case we would expect the best goalies to consistently outperform their counterparts year over year.  So we tested the persistence of Save % based on a sample of MLS goalies (2011-2014 seasons) who played a minimum of 2,000 minutes.  It turns out Save % is a horrible predictor, with absolutely zero relationship between Year 1 Save % and Year 2 Save %.  Breaking the stat into its components (Inside the Box Shots, Outside the Box Shots) did not help, either.

What does this mean?  For one, it confirms what most of the analytic community has known for some time: evaluating goalies statistically is really difficult.  It also, at least to me, confirms that luck/variance trumps skill when we evaluate goals scored (for strikers) or goals allowed (for goalies).


Wednesday, December 31, 2014

Measuring Referee Bias in BPL



Ask any fan their opinion of BPL referees and the responses will range from "mediocre" on the positive side of the spectrum to "#$%&..." on the more negative side.  These fans make their feelings known every match day, often vociferously.  It is possible that certain referees may hold biases against certain clubs, players or managers, but- in particular- the home/away bias has been noted in a multitude of prior studies.

Perhaps the most subjective decisions, and therefore most susceptible to bias, a referee makes is determining when a card should be shown to an offending player.  These decisions can often appear maddeningly inconsistent.  Indeed, we looked at the number of fouls committed per card for each BPL team (home and away) and found a pretty significant difference.  On average it takes a home team 6.6 fouls before they receive a card while for away sides it only takes 5.6 fouls.  Perhaps, just as a player might be said to "see red" in a fit of madness, a "yellow mist" might besiege a referee in front of thousands of baying fans.

See below for a team by team breakdown.  Chelsea is subject to the largest disparity in home/away fouls per card.


Tuesday, November 11, 2014

On Southampton's Stingy Defense

This article originally appeared in BSports



Southampton are enjoying a tremendous start to the season.  They are currently in second place on the table after 11 played, four points clear of Manchester City and have allowed only five goals, less than half a goal a game.  The Saints have been great defensively, though their astonishing goals allowed rate may be due to regress.  A major factor in the team’s defensive success has been their ability to limit opposition possession in their defensive 3rd.

Final 3rd Passes Allowed PG
Arsenal
101.2
Southampton
107.8
Manchester City
113.2
Manchester United
113.7
Newcastle United
121.3
Everton
122.2
Chelsea
124.5
West Ham
124.6
Stoke City
124.8
Tottenham Hotspur
126.5
West Bromwich Albion
126.6
Liverpool
130.4
Burnley
141.4
Queens Park Rangers
143.5
Swansea City
150.7
Leicester City
153.0
Hull City
157.0
Sunderland
159.1
Aston Villa
165.5
Crystal Palace
167.1

But the team’s defensive success starts further up the pitch.  Southampton leads the BPL in number of possessions won in the middle 3rd of the pitch, and the team has three of the top 10 individuals in the metric.  In particular, midfielders Victor Wanyama and Morgan Schneiderlin have done a great job of snuffing out opposition possession before it becomes dangerous.  Even striker Shane Long is getting into the act.

# Poss Won  Middle 3rd / 90
Victor Wanyama
Southampton
7.4
Charlie Adam
Stoke City
6.3
Fernandinho
Manchester City
6.0
Scott Arfield
Burnley
5.2
Tom Huddlestone
Hull City
5.0
Nemanja Matic
Chelsea
5.0
Alexandre Song
West Ham
5.0
Carlos Sánchez
Aston Villa
5.0
Shane Long
Southampton
4.8
Morgan Schneiderlin
Southampton
4.7


Monday, September 29, 2014

Who is the Most Important Passer for Every EPL Team?



The above chart shows the most important passer for every team in the EPL, based on pass usage rate.  Pass usage rate is calculated by taking a player’s passes per 90 min and dividing it by their team’s passes per 90 min.  It represents the share of their passing contribution to their team.


Typically, a team’s most important passer is either a central midfielder or a defensive midfielder (see: Yaya Touré, Morgan Schneiderlin, Gareth Barry).  Players in those positions should be the primary conduits for ball circulation.  Therefore it is troubling—though not surprising given their form—that  both Liverpool and Manchester United’s most important passers so far are defenders (Dejan Lovren and Tyler Blackett?!).  This is an indication that both teams, while strong in possession, have not done a good enough job of advancing that possession to more dangerous areas of the field.  It is still very early in the season, though, and the expectation is that Liverpool should be fine and Manchester United’s recent acquisitions Ander Herrera and Daley Blind will help.  Also of note, West Ham’s Mark Noble—who has been getting recent support for an England call-up—is leading the EPL in pass usage for the second consecutive year.

Tuesday, September 16, 2014

Usage Rates: A Primer

This article originally appeared in Stats Bomb




If there was one over-arching principle for analyzing soccer statistics, it might be “context is king.”  For example, Arsenal’s Bacary Sagna averaged 54.5 passes per 90 last year and West Ham’s Mark Noble averaged 53.2 passes per 90.  Intuitively, our first reaction is probably that both players exhibit roughly the same level of passing influence—with maybe the slightest of edges given to Sagna. But we are not controlling for the fact that Arsenal led the EPL with 569 passes per game while West Ham was second from bottom, averaging 326 passes per game.  To adjust for this disparity we take each player’s passes per 90 and divide it by their team’s passes per 90, thereby creating a pass usage rate for each player.

Passes / 90
Team Passes / 90
Pass Usage Rate
EPL Rank
Mark Noble
53.2
326.2
16.3%
1
Bacary Sagna
54.5
569.4
9.6%
142


Once adjusted, we now see that Noble was a significantly more influential passer for his team than Sagna and actually recorded the highest pass usage rate of any player in the EPL.  It should be noted that usage rates are not a predictive metric, nor are they meant to be, but they are a very useful tool to help us understand a player’s influence on their team and separate “team effects” from individual statistics.

General Usage Rates

Pass usage rate is—to this point—the most widespread usage rate.  Devin Pleuler recently published a nice write-up on the subject, which also introduced the idea of network centrality.  Pass usage rate is a “general” usage rate in that it does a good job of approximating a player’s general influence to their team.  Another general usage rate is the touch usage rate.  It differs from pass usage rate in that it measures more actions than just passes attempted (i.e. if a player receives a pass and shoots or turns it over before attempting another pass), so it is potentially a better proxy for general player activity.  We have also included Arsenal player’s pass and touch usage rates to further exemplify these differences.




 (Arsenal, 2013-2014)

Touches / 90
Pass / 90
Touch Usage Rate
Pass Usage Rate
Abs. Difference
Mikel Arteta
95.7
80.5
12.2%
14.1%
1.9%
Aaron Ramsey
98.5
77.3
12.5%
13.6%
1.0%
Mesut Özil
86.3
68.6
11.0%
12.1%
1.1%
Santiago Cazorla
89.8
67.1
11.4%
11.8%
0.4%
Tomas Rosicky
76.7
64.1
9.8%
11.3%
1.5%
Jack Wilshere
82.7
63.8
10.5%
11.2%
0.7%
Mathieu Flamini
77.4
63.1
9.9%
11.1%
1.2%
Nacho Monreal
82.0
55.0
10.4%
9.7%
0.8%
Bacary Sagna
82.1
54.5
10.5%
9.6%
0.9%
Per Mertesacker
61.8
48.8
7.9%
8.6%
0.7%
Kieran Gibbs
76.0
47.7
9.7%
8.4%
1.3%
Lukas Podolski
62.4
46.1
7.9%
8.1%
0.1%
Laurent Koscielny
57.9
41.7
7.4%
7.3%
0.1%
Olivier Giroud
50.4
32.5
6.4%
5.7%
0.7%
Wojciech Szczesny
39.7
18.0
5.1%
3.2%
1.9%





Attacking Usage Rates

You can measure a player’s attacking influence by looking at their shot usage, key pass usage, and general shot contribution usage.  These were the top 10 in the EPL last year.


Shots/90
Shot Usage (%)
KP/90
KP Usage (%)
Shot Contr./90
Shot Contribution Usage
Luis Suárez
5.5
32.1%
2.6
20.4%
8.14
47.5%
Christian Benteke
2.8
25.0%
2.0
22.7%
4.82
42.5%
Marko Arnautovic
2.9
25.5%
1.9
22.2%
4.79
42.5%
Wayne Rooney
3.7
26.8%
2.1
18.6%
5.77
41.7%
Jason Puncheon
2.8
25.3%
1.8
21.3%
4.51
41.3%
Robert Snodgrass
2.5
20.2%
2.3
25.9%
4.83
39.3%
Wilfried Bony
3.9
30.1%
0.9
9.0%
4.84
37.1%
Kevin Mirallas
3.1
21.2%
2.4
18.2%
5.48
37.1%
Philippe Coutinho
3.6
21.2%
2.5
24.3%
6.13
35.8%
Rickie Lambert
3.3
23.4%
1.7
16.3%
5.02
35.7%




Defensive Usage Rates

Defensive statistics remain a relatively under-researched domain in soccer analytics.  When people do talk defensive statistics, usually only tackles and interceptions are discussed.  This is a mistake, as tackles and interceptions combine to only comprise 24.4% of turnovers.  Any overall defensive usage rate should also include clearances and recoveries. 

2013-2014 EPL Turnovers (By Type)
Tackle
            11,153
12.6%
Interception
            10,435
11.8%
Clearance
            23,459
26.6%
Recovery
            43,236
49.0%

            88,283
100.0%


Here are the top 10 in overall defensive usage rate in the EPL last year. 


Turnovers Forced / 90
Defensive Usage Rate
Nemanja Vidic
20.0
15.9%
Phil Jagielka
17.8
15.9%
Marcos Alonso
16.0
15.4%
James Collins
17.3
15.3%
Nemanja Matic
17.2
15.1%
Martin Skrtel
18.7
15.0%
Martín Demichelis
16.3
14.7%
Laurent Koscielny
17.7
14.7%
Curtis Davies
16.9
14.7%
Youssuf Mulumbu
17.5
14.5%


It would also be informative to look at tackle, interception, clearance and recovery usage rates, respectively to get a sense as to a defender’s tactical responsibilities.  For example, Nemanja Vidic does most of his work with clearances (11.4 of the 20.0 turnovers he forces) and is responsible for nearly 31% of all of Manchester United’s clearances when he is on the field.