Assists and Chances

Celtic scored 126 goals from 1120 chances or a goal every 8.89 chances. Sinclair led the way with 15 assists but does that make him the most creative Celt? Who were Celtic’s creative craftsmen sculpting the Double Treble?

The Glossary splits uncertainty with logic like a through ball from Rogic.

Assists

Not all goals have assisters, and x of Celtic’s 126 were unassisted.

I am generous with assists compared to (say) the Fantasy Football sites. I award assists for being brought down for free kicks when the resulting shot goes straight in. Also, if the same player is brought down for a penalty and then scores it, he gets a goal and an assist. This is on the basis there are two completely separate actions. This may reward the strikers more often than “official” stats – which you rarely get to see anyway! Otherwise I follow the FF principles such as if the ball comes off the goalkeeper from a shot or cross then the shooter/crosser gets an assist if the rebound is netted.

But you are here for the numbers so here is the assist table for all competitions.

Anything over double figures is a good effort. Last season Hayes had 24 for Aberdeen which is exceptional. The top Celtic assister by volume was Roberts with 18 (sigh). In 2016/17 five players achieved double figures for assists compared to the three this.

Assists Per 90m

You know the drill by now. Totals are interesting but not statistically meaningful when comparing the squad. To remove the vagaries of number of minutes played we need to calculate the per 90 minute averages. And I am here to serve.

Dembele had a very good season assist wise with 0.44 assists every 90m. Although no player got near last season’s leaders Griffiths on 0.66 and Roberts 0.60 per 90m. Despite his lack of appearances Roberts makes it to second on the list illustrating what we missed. Griffiths was fourth but nowhere near his 2016/17 productivity.

The impact of considering per 90m averages can be seen clearly on Sinclair’s numbers as he drops to third by this measure.

For the players who were here last season I can compare their per 90m assist rates between the two treble winning seasons.

Most players had significantly less assists per 90m. The one’s whose rate increased the most generally had low volumes of overall assists (Armstrong, Boyata and Bitton). Dembele and Sinclair were the only notable contributors whose rates went up from last season.

Chances Created

The quality of a pass can obviously have a massive impact on the outcome of the shot. But there are clearly many other factors such as the field position, the positioning and subsequent action of the defence, the weather, and so on.

By counting chances created as well as assists we can better assess individual player creativity as there is a much bigger sample. Although Celtic had 108 assists this season, 1120 chances were created although 178 of those had no creator (e.g. chance arose after action from opposition).

The total chance creation statistics again more reflect the number of minutes played.

Looking at Chances Created per 90 minutes provides a more accurate measure of creativity.

Did I mention I rate Armstrong highly? Once again he is top of the charts pushing on for nearly 3 chances created per 90m. Hayes follows him, then Rogic and Roberts. Spot a trend?

Hayes has been unlucky so far in that he is creating chances at a good rate but not getting reward with assists. And once again we can lament the lack of Roberts on the pitch. With a good run of fitness and form I am sure he would have been challenging that 3 chances per 90 marker.

Ntcham, Forrest and McGregor create between 0.5 and 1 chance per 90m less that the aforementioned.

Expected Assists (xA)

You also know by now you don’t always get what you deserve in football, especially over a relatively small sample of a complete season. Expected Assists (xA) tells us the quality of the chances created whether the goal is scored or not. A shot does have to result but it is the position the pass allows the shot to be taken from that we measure in reporting xA.

In many ways, therefore, this measures the true creative talents removing the vagaries of what the shooter did with the chance. And as you can see it paints a very different picture.

The margins are very small but there is a top four all creating over 0.300 xA per 90m. And of course we did not see enough of any of them due to injuries – Armstrong, Rogic, Hayes and Roberts. If you read my stuff regularly you will not be surprised that these four produce the highest quality chances on average. Removing all of them from the side therefore has a significant impact on creativity and partly explains the massive drop off in goals scored especially in the SPFL (106 down to 73).

Ntcham, Forrest and McGregor all had great seasons for Celtic and played very well in most games. But they all manage to provide 0.05 to 0.1 less xA per game that the top four players. Over a long season that could equate to about 10 goals – enough to turn a few draws to wins and defeats to draws.

But this is theoretical in terms of what “should” have happened based on the models. In terms of actual assist tallies versus the number expected the picture is this:

Those at the top of the table have “outperformed” the Expected Goals model. That might be due to excellent play or it may be luck due to good finishing from their chances proportional to others. Either way, I would expect regression over time to the mean. That is, the actual assists and xA will be very similar over the long term.

Of those at the top, again apart from Dembele and Sinclair we are talking very low samples. Boyata in particular benefitted from two goals where he passed a simple ball in central midfield and a player ran on to score.

Sinclair had four penalties awarded for fouls against him to boost his assist total. And Dembele was twice brought down for penalties that he himself scored. Overall Dembele had more than 4 more assists than expected.

At the other end of the scale, regulars Brown, Lustig, Tierney and McGregor all had more than 2 assists less than expected. Despite Tierney’s continuing excellence, he had 12 assists last season and 9 this. Although the xG model states he should have had about the same.

Chance Productivity

Next we can see the proportion of chances created that resulted in goals.

25% of the chances Dembele created resulted in goals. That’s seems either exceptional or very streaky, or perhaps both. The next three players on the list only contributed 8 assists total so I suspect their sample sizes are too low. Then Sinclair on 18%.

The main creative midfielders are around the 8-15% which “feels” right.

But let’s not take away from the fact both Dembele and Sinclair have been highly productive setting up high quality chances this season.

Or have they? We can measure the quality of the chances by looking at the average xG value for the chances each player has created.

The table confirms that the quality of the chances created by Dembele and Sinclair was 2-3% PER CHANCE better than the bulk of the creative talents. Roberts and Rogic do well by this measure. Tierney can expect greater rewards in another season.

With each chance and shot I also record whether the effort was inside the box or not. Chances inside the box are generally more likely to be successful although taking the shot centrally can be as important. Which player created the most chances inside the box?

This is fascinating! The bulk of the creators are bunched around the 45-58% mark in terms of chances created inside the box suggesting a value of about 50% is “normal”. But Griffiths and Tierney smash it. Both create 73% of passes inside the box.

I suspect with Griffiths that the number of set pieces he takes that is a factor – i.e. crosses into the box from corners. I’ll keep the whole set piece analysis for another article as this one is already getting long (again!).

With Tierney this is a real outlier but reflects that he does not thrash the ball into the box from wide left. He always keeps the ball until he can play a “pass” as opposed to a speculative “cross”. I will continue to monitor this as it is quite an outlier and probably a powerful indicator of Tierney’s attacking worth.

How many passes does a player play on average before setting up a chance? Taking out the defenders and more defensive players who play a lot of passes but not necessarily in the final third:

This seems to be allied to position with Tierney included but with a similar pass per chance rate as Ntcham who is a deeper midfielder than many. All the attacking midfielders are bunched around 15-25 passes per chance.

The real outlier is Griffiths – 7 passes between each chance! All killer no filler. Griffiths is not a player for the non-material pass – his every action is positive. He does give the ball away consequently but when the cliché goes “he makes things happen”. Here is the evidence.

Summary

I hope you can see you need to look at the data many ways to assess creativity across the squad. Simple assist totals does not tell the wholes story.

As has been the theme all season, the injuries to key creative talents has robbed the team of high performing talent in this regard. The lack of Roberts and Griffiths, as well as prolonged absences for Rogic and Armstrong and to a lesser extent Hayes, takes away players who can create at the highest rates in the squad.

However, Dembele and Sinclair have been highly efficient this season in their chance creation and I will be interested to see if they can maintain those rates.

Two further thoughts to finish:

1. McGregor, Forrest and Ntcham have had great seasons and I won’t take anything away from them. However, if Rogic, Roberts and Armstrong had been fully fit and firing the data suggests they would have created more chances. I suspect that is a large factor in the overall downturn in goals this season.
2. If Celtic lose both Roberts (for definite) and Armstrong (not looking promising) then that is a massive reduction in creative talent that I hope the money men will look to replace.