Old Gord, New Tricks

Craig Gordon has been 1^st choice Celtic keeper since the start of the Deila reign.

He left his previous club Sunderland in 2012 and for 2 years battled knee injury without a contract. It is remarkable for a professional player in any position to have, effectively, a 2 year “career break” and come back to the very top level of professional football. So, at the human level, we have a special individual. He should, rightly, speak to Chelsea about their offer as he owes it to himself and his family. He appears highly regarded by Rodgers but it was not always so.

Rodgers prioritises the ability of the ‘keeper to distribute and maintain possession. Ewan Murray in the Guardian provides insight. Rodgers is quoted as saying “I need to have, for how we work, a goalkeeper that means when we have possession we have 11 players; that he can distribute the ball and pass it,”.

Is it more important to have an excellent shot stopper or a fantastic passer? Clearly and ideally, both, but if you had to prioritise? The game after which Gordon was dropped was away to Hapoel Be’er Sheva when he completed 12 passes and gave away 24 on that unbearably fraught evening – he had saved a penalty. Rodgers brought in De Vries for next game home v. Aberdeen.

As a parallel, Guardiola at Manchester City shipped out Joe Hart – media darling and undoubtedly a great stopper of shots – for Claudio Bravo – supposedly an excellent footballer and passer. This has a basis in logic in that whilst a ‘keeper like Gordon faces 2 or 3 shots on target per game, he faces at 30+ opportunities per game to pass the ball. The quid pro quo, playing the numbers, is “I’ll sacrifice some goals ‘against’ over the long term for the improved probability of creating goals ‘for’ through maintaining possession”.

Bravo receives a lot of criticism in England because a) he has let in many goals and b) he is not (England’s Number 1) Joe Hart. Mainly a) to be fair. But Guardiola is nothing if not persistent. (NOTE – Guardiola demands the game data on his desk 1^st thing Sunday morning after the Saturday game from the Manchester City analytics team, so CBN likes Pep!).

De Vries did not last long as Celtic’s number 1, letting in 11 goals in 405 minutes from 18 shots on target. Gordon was reinstated after De Vries was beaten by a wonder strike from Coulibaly of Kilmarnock with their only on target shot of the 1^st half. You can hear the crowd reaction on video suggesting displeasure but it is a wondrous, dipping and fiercely hit strike. De Vries departed at half time, allegedly with a strain, and has not been seen since in the starting XI. Gordon came back in, and, to the subjective eye, is playing differently.

There were groans the last 2 home games as Gordon attempted, rigidly, to maintain possession and play out from the back. He seems to be playing to order, including adopting a higher starting position when Celtic are in possession. So, does the data suggest Gordon is meeting his manager’s objectives and has the old Gord learnt new tricks?

Gordon Passing

This shows completed passes per 90 minutes. Since Gordon returned from being dropped this season, he is completing around 8 more passes per 90m than at any time in his Celtic career.

Additionally, his pass accuracy has increased by 5% on his best season 14/15 and 11% on last season and the start of this.

Gordon has clearly heeded Rodgers requirements and greatly improved his passing out from the back both volume of passes and their accuracy. Consequently, his PEI (Possession Effectiveness) has increased by 11% over last season which is huge.

What all this shows is that Gordon has adapted his game to comply with Rodgers’ wish to have a ball-playing ‘keeper playing possession football out from the back. It also shows how Gordon regressed in this respect in Deila season 2, and carried on that form into 16/17.

Passing and securing possession was a key criterion for Rodgers as the Guardian article referenced above illustrates. Was De Vries a better passer of the ball? (NOTE: Small sample alert plus the opposition was the 2^nd and 3^rd in the league, Barcelona way and Inverness away)

2016/17	De Vries	Gordon
Pass90	22.67	25.61
Pass Complete %	74%	76%
PEI %	71%	74%
Minutes Played	405	3015

De Vries, on the evidence of 405 minutes, was not an upgrade on Gordon regarding the narrow criterion of passing and keeping possession. In addition, he let in 11 goals from 18 shots on target and committed 1 defensive errors and 2 critical errors.

Let’s consider Gordons’ non-passing contribution – i.e. the goalkeeping fundamentals. But before we do……

Regarding goalkeepers and data analysis.

Analysis of goalkeepers using objective data is, well, hard. See sportsnet for an eloquent explanation, and also an article on 11tegen11. If, on the rare occasion a sports broadcaster provides goalkeeping data, it is usually around save %. Unsurprisingly, it is ‘keepers from the teams lower in the league that often have the best “stats” – i.e. they have most saves to make and make many of them.

So much of a ‘keepers job is based on decision making, communication, positioning – all factors the “data” is poorly equipped to draw out. I collect more data points for the ‘keeper than the outfield positions but much of it is unused during a game – e.g. I collect the different “types” of save (e.g. tip over, caught, saved with foot). Goals conceded and saves made are partly a function of the shot quality and position of the shot as well as the actions of the ‘keeper. A fantastic tweet emerged regarding the Southampton v Arsenal FA Cup tie showing shot position to perfectly illustrate that. It is as important how the defenders react and their positioning – are they exposing the ‘keeper to high % shots?

(NOTE – I don’t have a “solution” to this but what I will attempt to do it present a range of data points to form (hopefully) a nuanced view on performance. Also, since I have data for the last 2.5 seasons, variance across those data points can be observed. )

Being a Celtic ‘keeper means you don’t have as many saves to make. This season, Celtic have faced 128 shots on target and conceded 37 goals. We have scored 85 goals from 265 shots on target. Therefore Celtic ‘keepers have failed to save 29% of on target shots, and opposition ‘keepers have failed to save 32% of shots on target.

Gordon has played the equivalent of 137.4 matches since 2014. What has Gordon’s performance been over the last 3 seasons?

Gordon ‘Keeper 2014/15

As noted, I don’t have a database of comparative ‘keeper data to know whether this is ”good”. Subjectively, I believe Gordon was considered to have had an excellent 1^st Celtic season – he won the Scottish Football Writers’ Association Player of the Year award and was selected in the PFA Scotland Premiership Team of the Year. So, this is the bench mark.

Gordon ‘Keeper 2015/16

A clear difference can be seen, as more goals were conceded (the Van Dijk/Deanyer affect?); less on target shots saved (4% reduction); less interceptions won; less key saves and more defensive errors. Across virtually all KPIs, a “worse” season.

Gordon ‘Keeper 2016/17

This season, the performance appears to be getting back to 2014/15 levels. All KPIs are an improvement on 2015/16 EXCEPT defensive errors, which are higher than the last 2 seasons. None of the KPIs are quite as high as 2014/15 EXCEPT Intercepts Won and On Target Shots Saved is back to those levels.

In summary, Gordon is not quite reproducing the performance of 2014/15 when he had Van Dijk and Denayer in front of him. 2015/16 was clear across-the-board regression in performance. This season, Gordon is getting pack to the 2014/15 benchmark, but not exceeding it.

‘Keeper Comparisons

The other ‘keepers who have played during this time are:

Player	Season	Games	Minutes
Forster	2014/15	4	360
Zaluska	2014/15	6	425
Bailly	2015/16	5	450
De Vries	2016/17	5	405

This is too small a sample for meaningful individual comparisons. But, if I aggregate all those individual performances it comes to about 20 matches in total (ok – 17.5), so how does Gordon (complete Celtic career) compare versus the rest?

Gordon ‘Keeper 2014-17

All Other ‘Keepers 2014-17

Gordon appears to outperform all the other ‘keepers who have played the equivalent of 17.5 matches over 2.5 seasons. Clearly this is a small sample and the games player by the other ‘keepers include some of our easier fixtures (East Kilbride, Stranraer, the Glasgow Derby at home) but also has outliers like Legia Warsaw and Barcelona away. So – fair warning!

The headlines are he makes more saves, conceded less goals and is less error prone. Interestingly, the other ‘keepers have a fraction better passing stats – as we end where we started!

Summary

Rodgers is reported as adamant Gordon is not leaving. We have shown he has taken direction from his manager and improved his distribution significantly, developing improved skills at 34 years old. Whilst not reproducing his stellar 2014/15 form, his performance has improved markedly over last season. Looking at the performance of the other ‘keepers, Gordon is clearly out performing them albeit no one has had a run in the side longer than 4 games.