Archive for June, 2011

Shot Stopping vs Preventing Shots Revisited

June 16, 2011 1 comment

In a previous post I suggested that for a goalkeeper shot stopping is more valuable than preventing shots. By using some simple maths I showed that at real world save% levels (c. 60-80%) increasing save% by 1% is 1.5x to 4x more valuable than decreasing SoT/match by 1%. In this post I will build on that by looking at how often goalkeepers come to catch or punch crosses and how valuable those aerial contributions are using data from the first 16 rounds of the 2010/11 Bundesliga season. [1]

Goalkeepers face a lot of crosses every match. To be more specific, based on the data a goalkeeper faced on average 20.5 crosses per 90 minutes, or roughly one cross every four and a half minutes. Despite the high volume of crosses coming their way the average goalkeeper is passive in dealing with them, making only 1.24 catches and 0.83 punches per 90 minutes on average. Of all the crosses they face goalkeepers catch or punch only 10.1%. [2]

Of course some goalkeepers are more active in the air than others. The most active goalkeepers dealt with c. 15% of total crosses – Weidenfeller (15.2%) and Neuer (14.9%) – while the least active dealt with c. 6% – Sippel (6.4%) and Mondragon (6.2%). If we look specifically at the most dangerous crosses, the crosses into the six yard area, the average goalkeeper deals with only 32.8% of them and even Manuel Neuer, arguably the strongest goalkeeper in the world when it comes to aerial ability, deals with only 52.5%.

The relevant point is that even the most active goalkeepers deal with only a small minority of crosses. In addition to the low volume of aerial contributions there is also the question of how valuable each contribution actually is.

Let’s look at it from another angle and ask how valuable is a cross on average? If we simplify the question a little we could say that the value of a cross depends on three things:

1. What percentage of the time the cross finds a teammate
2. What percentage of the time the teammate is able to convert the cross into a shot on target
3. What percentage of the time the shot on target results in a goal

Intuitively it seems clear the value of the average cross is very low. The three-way parlay of finding a teammate, getting a shot on target and the shot resulting in a goal is very unlikely. According to a stat from @Orbinho via @11tegen11 in the Premier League the average cross leads to a goal only 1.6% of the time, so the value of a cross is 0.016 goals.

If the value of a cross is so low then the value of a goalkeeper dealing with that cross must also be low. If the average goalkeeper claims two crosses per 90 minutes and the most active goalkeepers claim just under three (Neuer 2.75), then the value of being elite at claiming crosses is c. 0.015 goals per match. Contrast this to shot stopping were the value of being an elite shot stopper is c. 0.2-0.3 goals per match.

In summary, crosses are a high volume low value action. Even the most active goalkeepers deal with only a small minority of total crosses, and each contribution has only a small value. The value of preventing shots by dealing with crosses pales in comparison to the value of shot stopping.

Something I didn’t address is the psychological benefit of having a goalkeeper who is good in the air on your team. I can’t speak from the perspective of a high level player but as a fan I know having a goalkeeper who is poor in the air fills me with fear and worry, and having a goalkeeper with a strong aerial presence has some positive value beyond simply the value they create by dealing with crosses. However I find it very difficult to believe that value could ever be enough that I would rather have an elite aerial contributor than an elite shot stopper.

I would also add that my view on the psychological benefit is that it comes not so much from dealing with a lot of crosses but more from simply not making any mistakes when you do deal with them. There are few things more frightening in football then seeing your goalkeeper flap at a cross, and that can have a negative effect on a team’s confidence. I don’t think a goalkeeper who mostly stays on his line but when he does deal with crosses does so confidently would have anywhere near the same kind of negative effect.

[1] I would have used the first 17 rounds to get the full first half of the season but the data for the 17th round was unavailable. The sample is also missing the match between Mainz and Stuttgart from the first round due to incomplete data.

[2] Goalkeepers do also drop some crosses (roughly 1 in every 20 matches on average according to the data) and sometimes come for a cross and miss entirely which is not reflected in the data (some keepers more than others of course) which would increase the actual value by a few tenths of a percent.

Categories: Theory

A Closer Look At Luka Modrić’s Passing

June 9, 2011 7 comments

While Gareth Bale may have won the PFA Player of the Year award and Rafael van der Vaart impressed with his goals and assists, for Tottenham fans it was the diminutive midfield maestro Luka Modrić who deserved the Fan’s Player of the Year award. Sir Alex Ferguson is a known admirer of the Croatian and many Manchester United fans see him as the ideal replacement for Paul Scholes, while Tottenham fans who know how important he is will be desperate to keep him at the club. In this post I take a closer look at his passing in the 2010/11 Premier League season using data gathered from the Guardian Chalkboards.

For this analysis I have considered only matches Modrić played in central midfield. Altogether the sample includes 2418 minutes played in 27 matches. 1338, or 55.3%, of those minutes were played in home matches, meaning his passing numbers may be slightly inflated from what they actually are. On the other hand he couldn’t have faced a more difficult range of teams. In the 27 matches in the sample he faced every team from the top nine in the final league table (excluding Tottenham of course) both home and away. He also faced teams 10-13 and 15 at home and teams 12 and 14 away. The only “easy” matches he played were Blackpool and West Ham at home and Birmingham and Blackpool away.

Modrić was the fourth most active passer in the league; only Danny Murphy, Michael Essien and Barry Ferguson attempted more passes than he did. In the matches he played in central midfield he attempted a total of 1715 passes in open play of which 1483 were successful. Per 90 minutes he completed 55 of 64 passes.

55.5% of his passes were played in the opponent’s half, 44.5% in his own. Of his passes in the opponent’s half 47.1% were forward passes, 34.2% sideways passes and 18.7% backwards passes. Of his passes in his own half 42.7% were forward passes, 38.9% sideways passes and 18.4% backwards passes.

His pass success rates in the opponent’s half for forward, sideways and backwards passes were 77.7%, 91.1% and 95.5% respectively. His pass success rates in his own half for forward, sideways and backwards passes were 75.2%, 96.3% and 97.9% respectively.

The accuracy of his different passes and the total number of successful/attempted passes of each kind are summarised in the following table.

Opp Half | Frwd Pass | 77.7% | 348/448
Opp Half | Side Pass | 91.1% | 296/325
Opp Half | Back Pass | 95.5% | 170/178
Own Half | Frwd Pass | 75.2% | 245/326
Own Half | Side Pass | 96.3% | 286/297
Own Half | Back Pass | 97.9% | 138/141

His lower forward pass success rate in his own half compared to the opponent’s half could be explained by the higher proportion of long passes he attempts. In the opponent’s half long passes make up 5.5% of his passes whereas in his own half they make up 8.5%. The long passes in his own half also tend to be more difficult forward passes whereas the long passes in the opponent’s half are slightly easier sideways passes. In the opponent’s half he completed 35/45 long passes, a success rate of 77.8%. In his own half he completed 30/57 long passes, a success rate of 52.6%. Overall he completed 2.42/3.80 long passes per 90 minutes.[1]

In 27 matches he attempted exactly 27 crosses with a success rate of 14.8%. Over the course of the season (including matches he played on the wings) he delivered 31 corners into the box, finding a teammate with 29.0% of them. His total cross success rate (including all crosses and corners) was 22.6%.

Over the course of the season he created 2.06 goal scoring chances per match (source: but only 2 of those 64 chances were converted into goals. Possible explanations for this could be that Tottenham need to buy some better strikers, that for some reason the kinds of chances he created weren’t particularly good goal scoring chances or it could just be bad luck. Without more detailed data it’s hard to know what the reason is.

Performance Against Top Sides

To see how he performed against the strongest opposition I looked at the matches he played against Manchester United, Chelsea, Manchester City, Arsenal and Liverpool. He played all five teams both home and away and played the full 90 minutes in each match, and for the most part his numbers are similar to his overall numbers.

On average he completed 51/59 passes of which 51.4% were in the opponent’s half. His ratios of forward, sideways and backwards passes and their corresponding success rates in the opponent’s half were 43.0% (79.2%), 39.1% (89.0%) and 17.9% (96.3%) respectively. The corresponding numbers for his passes in his own half were 37.9% (71.3%), 42.8% (94.3%) and 19.3% (100%). He also completed 2.4/4.4 long passes per 90 minutes.

As you can see he attempted slightly fewer passes and more of his passes were in his own half, he was slightly more conservative in his passes in terms of forward and sideways pass ratios and he attempted slightly more long passes, but for the most part the numbers stack up well which suggests he doesn’t wilt under pressure and is capable of performing at his usual level even against the best opposition.

[1] If you compare my long pass numbers to the numbers from for example you will see they differ quite a lot. I think this can be explained by me being a lot stricter in my definition of a long pass when gathering the data.

Categories: Players

A Closer Look At Mario Gómez’s Shooting

June 5, 2011 3 comments

Mario Gómez finished the 2010/11 season as the Bundesliga’s top scorer with 28 goals in 32 appearances. Per 90 minutes he scored 1.03 goals, or put another way he scored a goal every 87 minutes he was on the pitch. In the top European leagues only Cristiano Ronaldo (40) and Lionel Messi (31) scored more goals than he did (Antonio Di Natale equaled his tally of 28). In this post I take a closer look at his shooting and goal scoring using statistics gathered from the official Bundesliga website.

In total Gómez took 95 shots or 3.50 per 90 minutes. Of those 48 were on target, 26 off target, 16 were blocked, 3 hit the post and 2 hit the crossbar. The majority of his shots, 79, came from open play. 2 came from penalty kicks, 4 from counter attacks, 4 from corners, 1 from a throw-in and 5 from free kick crosses.


One of Gómez’s strengths is his ability to score goals with either foot or his head. With his right foot he scored 13 goals from 49 shots, with his left 9 goals from 24 shots and with his head 6 goals from 21 headers. His goals-to-shot ratios were very good with all three; 0.27 with his right foot, 0.37 with his left and 0.29 with his head.

Shot Locations

To see where his shots came from I divided the pitch into seven sections: Six Yard Left (orange), Six Yard Middle (green), Six Yard Right (purple), Penalty Left (yellow), Penalty Middle (blue), Penalty Right (red) and Outside Area.

Gómez is a real penalty box predator; only 5 of his shots came from outside the area. 51 of his shots (53 if you count the two penalty kicks) came from Penalty Middle and 19 from Six Yard Middle, so altogether 75.8% of his shots came from directly in front of the goal. His other shots came primarily from Penalty Left and Right, 9 and 7 respectively, with 1 each from Six Yard Left and Right.

Given most of his shots come from directly in front of the goal it’s no surprise most of his goals do as well. 14 of his goals were scored from Six Yard Middle and 11 from Penalty Middle. 2 were scored from Penalty Right and 1 from the penalty spot. His goals-to-shot ratios from Penalty Middle and Six Yard Middle were 0.22 and 0.74 respectively.

Shot Destinations

Gómez tends to favour his left side when aiming his shots. Of the 48 shots that were on target 18 were aimed at the left third of the goal, 23 at the center third and 7 at the right third. Of the shots that were off target to the side of the goal 10 went to the left and 6 to the right. He hit the left post twice and the right post once. In total he shot to his left more than twice as often as to his right; 38.0% to his left versus 17.7% to his right.

When choosing the height of his shot his range was more balanced. 27 (34.2%) of his shots were aimed high (the top third of the goal and everything off target above the goal), 19 (24.0%) to the middle third and wide on either side and 33 (41.8%) to the bottom third and wide on either side.

In the following table I have summarised the destinations of all of his shots on target. I split the goal into nine sectors – top left, center, right, middle left, center, right, bottom left, center, right – and listed how many shots ended up in each sector and how many goals those shots resulted in.

Destination | Shots | Goals
Top Left | 3 | 1
Top Center | 3 | 0
Top Right | 3 | 3
Middle Left | 3 | 3
Middle Center | 7 | 3
Middle Right | 1 | 0
Bottom Left | 12 | 7
Bottom Center | 13 | 8
Bottom Right | 3 | 3

The bottom sectors were the most productive for him with 18 of his 28 goals coming from there. The left side of the goal was more productive than the right with 11 goals scored to his left versus 6 to his right.

To any Bundesliga goalkeepers reading this I would offer a suggestion: the next time you come up against Gómez and he is about to shoot stay low and get ready to dive to your right. Whatever you do don’t dive to your left.

Finally, what would this post be without a video of Gómez in action, so here is a video of all his goals from this season, including not just his 28 Bundesliga goals but also his goals from other competitions. Thanks to jackmcrobert for the video.

Categories: Players

Manchester United Central Defenders By The Numbers

June 4, 2011 1 comment

In this post I will look at the performance of Chris Smalling, Jonny Evans, Rio Ferdinand and Nemanja Vidic in the 2010/11 Premier League season using data gathered from the Guardian Chalkboards, but before I do so I feel the need to make a few points.

1. These numbers are not meant as a definitive measure of a player’s performance. Defending is a very complex art and it would be silly to suggest that it could be summed up with a few simple metrics. As is always the case, these numbers should be used to supplement the eyeball test, and in cases where the results differ from your observations they should be used as a test – is there a reason the statistic could be wrong? is there a problem with how the data is gathered? is it possible the statistic is right and I’m wrong? – to make sure your observations are sound. I will touch on this point again later in the post.
2. The sample sizes used here are very small so everything should be approached with the appropriate level of skepticism. The total minutes included for each player is as follows: Vidic – 3118, Rio – 1710, Smalling – 1037, Evans – 871.
3. I include in these numbers only minutes played at center back which means Evans and Smalling are missing some minutes they played at left back and right back respectively.
4. I have omitted from the data all passes made in the last ten minutes of the Blackburn away match since it was a highly unusual situation and to include those passes would pollute the sample with data that is not a true reflection of a player’s game.
5. I haven’t included Wes Brown’s minutes because he only played one match at center back.


Evans was the most active passer, completing on average 43/50 passes per 90 minutes. Vidic and Rio were next with 38/47 and 37/46 passes respectively while Smalling brought up the rear with 30/40 passes. Note however that Evans played a disproportionately large 61.5% of his minutes in home matches, so you would expect his passing numbers to be slightly inflated. Smalling played 45.5% of his minutes at home so his actual numbers may be slightly higher than they seem. Vidic played 48.0% of his minutes at home while for Rio the corresponding number was 52.6%.

In the graph above I have separated each players passes into forward, sideways and backwards passes. As you would expect each player plays the majority of his passes forward. Vidic is the most frequent forward passer with 62.3% of his passes going forward. Rio is close behind with 61.5%, Evans is third with 56.8% and Smalling last with 52.8%. When it comes to sideways passes Smalling is clearly ahead of the rest with 37.3% of his passes directed sideways. Evans, Vidic and Rio are all close together with 26.9%, 24.5% and 26.5% respectively. Smalling plays the fewest backwards passes, 9.8%, while Evans plays the most, 16.3%. Rio and Vidic are close together with 12.0% and 13.1% respectively.

The success rates of each player’s forward, sideways and backwards passes are shown in the graph below:

In addition to attempting the fewest forward passes Smalling is also the least accurate forward passer, completing only 60.7% of his forward passes. Evans is the most accurate at 77.2% while Vidic and Rio are very close with 71.4% and 72.0% success rates respectively. When it comes to sideways passes Evans, Vidic and Rio are all close, completing 96.2%, 95.2% and 96.1% of their sideways passes. Smalling is slightly less accurate at 91.8%. Each player’s backwards pass success rates are very high, as you would expect and hope.

Needless to say these numbers don’t tell us everything. For example they tell us nothing about the ratio of long, medium and short passes a player played. They also tell us nothing about how difficult each attempted pass was, or about how much pressure the player was under when they attempted the pass. This is where combining these numbers with your own observations from the eyeball test becomes important.

Defensive Actions

The graph above shows each player’s successful aerial duels, successful tackles, interceptions, successful clearances (both headed and non-headed) and blocks per 90 minutes.

Smalling was the most active aerial dueler, attempting 5.2 aerial duels per 90 minutes and winning 3.8 or 73.3%. Rio was the least active, attempting 2.7 and winning 2.0 or 75.0%. Vidic won 2.9 out of 4.3 attempted aerial duels for a success rate of 68.4%. The weakest in the air was Evans who won only 2.3 of 4.6, a success rate of only 50%.

On the ground Vidic was both the most active and most successful tackler. He won 1.6 of 2.0 attempted tackles, a success rate of 78.9%. The second most successful was Rio who won 76.0% of his tackles, 1.0 of 1.3. Evans won 1.0 of 1.4, 71.4%. The weakest tackler was Smalling who won only 58.8% of his tackles, 0.9 of 1.5 attempted.

Evans made the most interceptions, 3.6 per 90 minutes. Vidic, Rio and Smalling won 2.9, 2.4 and 1.8 respectively.

Smalling made 5.1 successful clearances per 90 minutes out of 8.8 attempted, a success rate of 58.4%. 68.3% of his clearances were headed. Vidic was successful in 5.0 of 9.7 attempted clearances, or 51.6%. 60.8% of his clearances were headed. Evans attempted 8.0 clearances per 90 minutes but succeeded in only 3.4 of them, a success rate of 42.8%. 64.9% of his clearances were headed. Rio attempted the fewest clearances, 6.9, and was successful in 3.6 of them, or 51.9%. 54.2% of his clearances were headed, 45.8% non-headed.

Smalling made the most blocks, 0.87 per 90 minutes. Vidic, Evans and Rio made 0.84, 0.72 and 0.42 respectively.

Smalling won and conceded 0.95 and 0.87 free kicks per 90 minutes. The corresponding figures for Evans were 0.93 and 1.24 and for Vidic 1.01 and 1.15. Rio won 0.53 but conceded only 0.16 – just 3 free kicks conceded all season.

On The Usefulness Of These Numbers

Are these numbers meaningful? Can we use them to answer questions like who is the better player, who performed better over the course of the season? Take Evans and Rio. Evans completed more passes, won more aerial duels, won more tackles, made more interceptions, made the same number of clearances and made more blocks. The logical conclusion then is that Evans is the better player, yet anyone who has watched Manchester United play could tell you that Rio is by far the better player, and was also much better over the course of the season.

These numbers alone are not enough to make judgments of relative worth and value, but that doesn’t mean they are useless. Used in the right context and as a supplement to the eyeball test they can have value. For example, knowing that Evans won only 50% of his aerial duels is valuable because we then know this is an area he has to improve on. I don’t need the stat to know that Evans is poor in the air – I can see that with my own eyes – but having the stat reinforces the point.

Goals Allowed

If simply looking at the stats isn’t enough to tell us who is the better player we need to find other ways of answering the question. One way is to look at how many goals a team concedes when the player is playing. While this method also has obvious flaws and limitations I decided to include these graphs anyway. The first graph shows how many goals the team allowed per 90 minutes when each player was on the pitch. The second graph shows how many goals the team allowed per 90 minutes when a certain center back partnership played. Even though they do gel well with the eyeball test I wouldn’t put too much stock in these.

For reference the number of minutes played by each partnership is as follows: Vidic/Evans – 673, Vidic/Rio – 1594, Vidic/Smalling – 851. Evans/Rio, Evans/Smalling and Brown/Smalling also played together this season but the sample sizes were so small they were of no value and so I omitted them.

Categories: Players