Keeping Fine

In the second of the 2016/17 season reviews, I start at the back, with the goal keeper. I have already produced a feature on Craig Gordon, and I will not repeat it but request you read “Old Gord, New Tricks”.

Gordon has been the Celtic first choice for the three seasons I have been collecting detailed data. Evaluation is, on the one hand, comprehensive and complete and on the other, difficult on the basis there isn’t much of a peer group with which to compare. Long story short, I showed in “Old Gord, New Tricks” he outperformed the cumulative efforts of all the other ‘keepers across all the key indicators.

But what of his individual performances? Gordon suffered his first Celtic set back, being deposed by Rodgers signing Dorius De Vries. After letting in 11 goals from 18 shots on target in 405 minutes, Gordon was reinstated. And that is the last time I mention De Vries. Gordon completed 4905 mins, his highest for Celtic by 195 mins and the equivalent of 54.5 matches. Deila rested him occasionally given his previous injury record. With Rodgers, his number 1 plays every game.

He was rumoured to be interesting Chelsea in the January transfer window and either a bid never arrived or Gordon’s stated happiness at Celtic prevailed. The rumour sites are currently (June 2017) silent on signing targets that are ‘keepers and so it appears that entering his 4th Celtic season and his 35th year, Gordon will remain the Celtic number 1.

Before we start a reminder: Analysis of goalkeepers using objective data is, well, hard. See sportsnet for an eloquent explanation, and also an article on 11tegen11.

So, to his performance scorecard.

Shot Stopping

It is what he is there for, keeping the ball out the net. A core skill. Of course, statistically quite difficult to evaluate as the better ‘keepers will tend to play for the better teams and have less shots to face.

This season is Gordon’s best in terms of save % of on target shots. He managed to keep 80% of shots on target out. This compares to Celtic scoring with 37% of their shots on target – 63% of shots being saved. Most Celtic games are in Scotland, and Celtic are vastly dominant in them. Shot quality for the Scottish opposition may be inferior due to Celtic having better defenders. Overall Celtic had an xG against of 49.11 and conceded 50 goals. Therefore, Celtic conceded almost precisely the number of goals you would expect given the chance quality the opposition created.

He has had less shots to face than under Deila and produced less volume of saves.

Highlight saves included my candidate for his save of the season at home to Barcelona from Luis Suarez:

Also notable was the last minute save from a Sviatchenko interception against Aberdeen that sealed a psychologically important early season 1-0 away win:

Both big saves in big games. Conversely, the one shot he faced resulting in a goal that I felt he should have saved was also a big game, away at Manchester City. Iheanacho’s high and hard shot seemed to go through Gordon who may have dived a fraction early.

Gordon has racked up 33 key defensive saves over the season at 0.61 per 90m. This is lower than 2014/15 (0.85) and 2015/16 (0.73). However, there has been significantly less saves to make (14 less than last season) behind a more dominant defence than in recent seasons.

In terms of highlight matches, three times Gordon has produced 3 key defensive saves in the same match:

  • Away to FC Astana in the Champions League qualifier where a hard-fought draw was achieved;
  • Home to Manchester City where he saved 6 of 9 shots on target. Only Barcelona (10) managed more in any one game;
  • Away to Ross County, where although Celtic won 4-0, Ross County had 14 shots, 5 on target; and
  • In the 1-1 derby match against The Rangers, Gordon mitigated defensive slackness with 3 key saves.

As a stopper of shots, Gordon has had a near exemplary season with few errors. His performance is marginally better than previous years against fewer opposition efforts, as he is now playing in his strongest Celtic team.


This is the second evaluation criteria as Rodgers clearly values it so highly. The thinking being that as ‘keepers for a team like Celtic (domestically dominant) have so much possession and face relatively few shots (7.7 per 90m in all games), a strong footballing ‘keeper will maintain possession and even help the team create chances. This will outweigh a few avoidable goals at the other end if the ‘keeper is a great shop stopper but average footballer.

I showed in “Old Gord, New Tricks” how Gordon had adapted and improved in this regard, and here is the refreshed data.

Gordon’s passing in the final Deila season was carried over until Rodgers dropped him. “Post drop” shows the improvement in pass accuracy, an increase of 11%. This is the equivalent to playing a new position – i.e. if a striker suddenly played a midfield role. The number of passes completed per game went up from 19.22 last season to 26.7 this. He is completing more passes, and passing more accurately.

Since being reinstated, four Celtic goals have had the passing chain heavily involve Gordon:

  • 29/01/17 McGregor’s opener in the 4-0 win over Hearts;
  • 05/02/17 Involved but did not commence the passing chain for the Dembele team goal away to St Johnstone in the 5-2 win;
  • 02/04/17 commenced the move leading to Sinclair’s second goal in the title clinching 5-0 away to Hearts; and
  • 18/05/17 started the break that led to Roberts being fouled and Griffiths scoring the opener in the 5-0 win at Partick Thistle.

He has also created two chances directly. The long ball pass to Griffiths in the final minute of the derby match versus The Rangers in March would have resulted in a late winning goal or penalty had Mr Madden not been imperfectly positioned. He also created a chance against the same opposition in the Scottish Cup semi-final.

A statistical oddity is that, despite being under less pressure this year (less saves to make) and focussing on building from the back, his clearance rate has increased. A clearance is basically a hoof up the park to clear danger as opposed to a targeted pass. He clears 2.18 times per 90m compared to 1.82 last season. Potential explanation: Celtic have 65% possession this season compared to 60% last. Celtic players are clearly under instruction to maintain possession where possible. They complete on average 535 passes per 90m compared to 464 the previous season. That may mean that Gordon is given more pass backs than previously, to maintain possession. Some of those pass backs may put him under pressure thus requiring more clearances.

The collateral damage in this approach is the higher risk of errors. Gordon’s error rate has gone up markedly compared to the last two seasons.

As we have seen above, Gordon’s handling is excellent. His positioning is also sound as there are very few goals conceded he could perhaps have saved. The increased error rate is almost wholly down to attempting riskier passes out from the back. Two recent examples are the 4-0 home game against Hearts and the 5-0 away win against Partick Thistle. In both games, he twice gave the ball away allowing the opposition to create a scoring threat. Two defensive errors are the highest Gordon has been awarded for any single game this season. It has happened three times, the home game against FC Astana being the third, where he came out to clear with a header straight down the middle to their forward who scored.

Would Gordon have been involved on the pass chain that resulted in four goals or created two clear chances without the mentality to play out from the back? Although his defensive error statistics have increased, Celtic have not conceded any goals directly from his misplaced passes. It may therefore be a risk worth taking.

Overall, Gordon has clearly adapted his game to meet his managers expectations. New tricks indeed.


Over the season, Gordon has caught 47 crosses and only failed to catch 1, at home to St Johnstone in January. A defender cleared the danger. His cross catching rate of 98% is in line with previous seasons.

63 saves have been made where the ball has been caught safely, out of a total of 121 shots faced on target (52%). Gordon has only fumbled one save, away at Hearts on the opening day of the league season. The loose ball fell to a Celtic defender.

Overall there have been no goals conceded due to handling errors.

Decision Making

No, I am not a mind reader but we can infer the quality of decision making from certain actions recorded. As the ‘keeper is the last line of defence, the impact of a wrong decision can be the loss of a goal or loss of a player due to a red card.

As part of Gordon adapting his game, his starting position when Celtic have possession is generally higher. To play the role of “sweeper ‘keeper” requires being available for the back pass and to be positioned close enough to the outfield players such that move momentum is maintained. He must also be near enough to action a short pass to the next player in the chain.

Like the “pass out from the back” paradigm, there are greater risks. There have been several occasions this season where Gordon has made the decision to affect the play despite this taking him away from his goal, encouraged by a higher starting position.

  • At home to Alloa Athletic he came out and collided with Spence in the right back area and picked up a yellow card;
  • Away to Ross County a very similar incident resulted with another caution for bringing down Boyce outside the box;
  • Away at Partick Thistle, Gordon was fortunate Azeez was booked for diving when there appeared to be contact outside the box; and
  • Away to Aberdeen he collided with Shinnie wide within the area without any contact on the ball.

There were no adverse outcomes from any of those incidents which is perhaps fortunate. Gordon is still adapting to being a more proactive, front foot ‘keeper. Also, those were the only fouls conceded by Gordon this season. This is how Rodgers wants him to play and I will monitor this aspect in the new season.

Positively, Gordon has achieved the most interceptions this season:

The number of intercepts is slightly up on his initial season but he is noticeably more successful executing them.

As discussed above, dealing with crosses – including the decision to come and collect, is virtually faultless.

Gordons other (of two) critical defensive error occurred away to Hapoel Be’er Sheva and could have been fatal to Celtic’s season. Already a goal down and under pressure, Janko and Gordon collided resulting in a simple tap in putting Celtic 0-2 down, one goal from Champions League elimination. Hesitancy and lack of communication from both players contributed. I feel that since Gordon has come back from being dropped, his more assertive and decisive performances, warts and all, suggest this was a rare aberration.


In January when I wrote about Gordon, the evidence suggested he had returned to near but not quite matched, the 2014/15 season performances after a noticeable dip in 2015/16. Considering his scorecard:

His yearend performance is therefore more impressive than I anticipated in January. My confirmation bias when starting this article was that I would be writing about succession planning and the need to have a new ’keeper put pressure on him. Instead, Gordon’s performance rough edges are a direct result of managerial instructions. His test for next season is to absorb the learning whilst maintaining exceptionally consistent overall performances.

Leave a response

Your email address will not be published. Required fields are marked *