Season Preview: Which teams will see an increase (or decrease) in goal scoring next season?
written by: @TheGersReport
Last month I introduced a new stat called Projected Goals which simply applies the league average Shot Accuracy & Shooting Percentage (Sh%) to the unblocked shots a player takes to project what his goal numbers would be if he simply matched those league average rates.
For example, Scott Sinclair took 56 unblocked shots last season & based on the kinds of shots he generated, his Expected Shot Accuracy would have been 57%. In actuality, 55% of his unblocked shots were on target - so his accuracy was only slightly below the league average for the kinds of shots he took (i.e. there is a different Expected Shot Accuracy for a shot taken 30 yards out then there is for one taken 10 yards out from the heart of the goal). Now, his Expected Shooting Percentage [xSh%] (the rate of shots on target that 'should' have been goals) was 36%, but his actual Sh% was only 26%. What that math looks like translates into a Projected Goals per 90 rate:
*(Unblocked shots x Expected Shot Accuracy) x Expected Shooting Percentage = Projected Goals
Scott Sinclair: (56 x 0.568) x 0.363 = 11.55 Projected Goals*
When you apply the Expected Shot Accuracy to the shots he took, 31.81 of those shots 'should' have been target (in real life he put 31 shots on target). Then when you multiply that by his xSh%, you get his Projected Goals of 11.55.
But in actuality, he only scored eight goals as his finishing was nowhere near the league average rate.
My theory is that if Sinclair hypothetically repeats (or comes close to) these shot totals next season, he is a very safe bet to score more goals next season.
With the new season right around the corner, I figured it may be interesting to apply Projected Goals to both teams & players to predict who will see an increase or decrease in scoring next season. I implore everyone who reads this to benchmark this post, so we can all revisit it as the season goes on. I really want to see how accurate the projections will be.
Now, you may wonder why you would use last year's data, with last year's squad to make predictions about next season. Firstly, it is a way to model how data should drive recruitment & tactical adjustments. If a team is expected to see a decrease in goals, they need to know that as they look to make adjustments to the squad for the upcoming season. Also, when dealing with the hundreds of shots taken by teams - it will be interesting to see how collective regression (& even luck) influences results.
Applying this stat to a team analysis can help target potential deficiencies for a team based on their squad's results from the season before. For example, a team like Aberdeen projects to see a slight decrease in goals (-1.64), however that's based on their squad from last season. That squad included Adam Rooney & Kenny McLean who combined for eleven non-penalty goals. What will the team do to replace those goals? To what extent were those players expected to score more or less goals given their underlying numbers?
Note: teams are ranked from highest expected increase in goals to the biggest projected decrease. Infographics have been created that identifies the top goal scorers for each club (non-penalty goals only) along with how their goal rates compare to their Projected Goals is then broken down to per 90 rates. I also highlight the player who will likely have the biggest influence on the change in the club's goal scoring rate. Alternatively you can also click on the clubs badge to go direct to their preview.
Side-note: This is why I only include non-penalty goals...because Sofien Moussa's goal totals would be the same as Leigh Griffiths' nine...if you included penalties.
Five goals from your main forward is putrid & most observers would agree that the 'eye test' shows us a really bad finisher whenever you watch Moussa. Look at that Shot Accuracy on unblocked shots: 38% of Moussa's shots were on target when if he simply shot the league average from the shots he takes it would be 49%. It's hard to score when you can't even force the goalie into having to react!
I'm not a believer in Moussa....but the data might be. His Shooting Percentage numbers are (gasp) kind of inspired. When he did get shots on target (which wasn't all that often) he scored 46% of the time, & even more impressively his xSh% is 49%. Part of those rates is a small sample size given how often Moussa's shots were sent sailing into arms of the supporters behind the goal, but the numbers should also provide some hope for Dundee that Moussa might have more goals in him this season.
Interestingly, as a team, Dundee's Shot Accuracy was basically equal to their expected rate but you can see their finishing was shocking. A Shooting Percentage of 23% was the league's worst & the next best team is no longer in the division (Partick Thistle had a Sh% of 25%). But there xSh% was actally among the best rates in the league at 32%.
The question becomes where will those goals come from?
Below is a table of all the Dundee players who played at least 900 minutes, along with their Projected Goals per 90 minutes & their actual goal rate.
If Dundee is to see a bump in goals:
- A lot of it will fall to Sofien Moussa again. This will be a nice test for this statistic. It suggests Moussa will be a better goalscorer (if he can get the same kinds of shots at a similar volume), while the eye-test tells us he sucks.
- El Bakhtaoui & Leitch-Smith are not in the squad for the upcoming season. Last week I wrote about a few young prospects who deserve a chance to step into those roles. Neil McCann has options, it will be interesting to see who he gives the opportunity to.
- You can see a lot of the players in the 0.10 to 0.15 Projected Goals per 90 range are defenders - which makes me believe that Dundee creates some decent chances off of set pieces. If that's true, they might be due to see more of those chances turning into goals.
Last season, Celtic were victims of regression. All of their top scorers from 2016-17, saw a decrease in scoring. It was a decline in scoring rates that the stats would have flagged for anyone paying attention.
Look at the goal rates alongside the Shooting Percentages from one season to the next.
That's what regression looks like. Those top four scorers were riding unsustainable finishing rates & all came back to reality last season.
The scary thing is that regression led to Celtic only winning the league by nine points instead of 30.
As a team, they are poised for a pretty significant spike in goals next season & there is little evidence to suggest their defence is going to all of the sudden become porous.
The numbers suggest that, if given the chance, Scott Sinclair will have a bounce back season after seeing his goals per 90 drop form 0.73 in 2016-17 to 0.31 last season. In the infographic you can see that his xSh% was 36% while his actual Sh% was a woeful 26%.
If Sinclair matches his Projected Goals next season that would translate to 3-4 more tallies to his goal totals. That may not sound like much but imagine if Kris Boyd scored four more goals last season, or if Alfredo Morelos did. Their totals would leap to 20 & 18 (& that doesn't include penalties).
Ultimately, for a team as good as Celtic to have a Shooting Percentage as low as 27% is scary...because it's a pretty safe bet it's much closer to 30% or above next season.
For a reference, their Shooting Percentages in the previous two seasons? 43% & 32%.
Yeah, I don't think 27% is happening again.
St. Johnstone will be a very curious side to follow this season. Obviously there has been some real turnover in talent given that their top two goal scorers are no longer at the club. But there are some real elements of hope with the low-risk investment in Tony Watt & the fact that they have two of the top rated forward prospects in Scotland. Another reason to be hopeful if you're a Saints fan, is that David McMillan is healthy. He showed flashes of being a very competent striker in his extremely brief appearance last season, which I highlighted in this recent post.
But, honestly the biggest reason for hope is the fact the club is collectively due for some luck. They only scored on 29% of their Shots on Target last season despite an Expected Shooting Percentage of 35%. Now, they don't generate tons of shots, but their counter attacking style tends to lead to quality chances (compare their expected rates to the likes of Celtic, Rangers & Aberdeen).
Let's start off by acknowledging the fact that replacing the long-term impactful player that was Louis Moult will not be easy. He produced at an elite level in two of his three seasons in Scotland & was able to flip that into a move down south. He scored a lot of goals to help keep Motherwell competitive & £450,000 still feels like a bargain price for a striker in his prime but alas...thus is the market for players from the Scottish Premiership
It should be noted that Moult's scoring rate from last season was looking pretty unsustainable. His Shooting Percentage of 41% didn't come close to aligning with his xSh% of 29% & in turn his actual goal production outpaced his projected rate. Then again, his Shooting Percentages in his previous two seasons go against the idea that rates over 40% will regress.
In 2015-16, his Sh% was 46% & the next season it was 56%. Technically, the regression was beginning as it dropped down to 41% last season...but still - maybe these are indicators that he was too good for this level of football. Checkout his Goals Above Average rates from his time with Motherwell to his limited minutes with Preston North End last season.
The fact he was scoring at ever increasing rates, while riding three straight seasons of "unsustainable" Shooting Percentages is evidence that Louis Moult had outgrown the Scottish Premiership. Rumors have been floating out there linking him to a loan move to Aberdeen - but for his own development he's better served being challenged in a better league. He's kinda conquered Scotland already.
That was a long way of saying - Motherwell shouldn't worry about replacing Louis Moult's production. It's kind of rare to find players who are too good for this level (especially when you're Motherwell). Rather, they should hone in on the player the stats suggest he should have been. Use his Projected Goal rates to flag potential targets.
That won't be easy, though given that Moult had the 11th highest Projected Goals per 90 rates. The player at #10 was available this summer but given his position & injury history, David Templeton isn't necessarily an ideal player to target. Instead, Motherwell have brought in Connor Sammon, whose Projected Goals per 90 was 0.30 last season (his actual goal rate was 0.18).
Sammon rarely seems to pass the 'eye test' as a quality goal scorer so let's see what the numbers from last season show us.
This is more hopeful that I would have expected. Sammon's Expected Shooting Percentage is pretty off the charts good, which may mean he's due for a bump in goals this season.
Why is that xSh% so high? I mean, he played for Partick Thistle last season...well, 57% of his shots were Scoring Chances (kicked shots from heart of box &/or kicked/headed shots from within six-yard box). For a comparison, Moult's rate was 33%.
Will Sammon replace the goal scoring of Moult?
Do I really need to ask that question?
Again, the more realistic expectation is whether he can come close to Moult's Projected Rate of 0.39 goals per 90?
That seems a little more doable.
But then again...he looked pretty bad struggling to get that 0.18 goals per 90.
The real key will be Craig Tanner. Earlier this summer, I tabbed Tanner to be among the top breakout players next season & his Projected Goal rate supports that. However, it's Tanner's ability to create quality opportunities for teammates that will also help Motherwell's chances next season. He really can be a dynamic player, who will help players like Sammon & Curtis Main chip away at making up for the loss of Louis Moult.
Ok, so this one is a little weird. It's pretty rare to go into a season in which you will be relying upon a 35-year old forward to replicate the kind of season that Kris Boyd had last year...& then expect to score even more goals as a team.
First, even though Kris Boyd scored at a noticeably higher rate than his Projected Goals rate...ummm...that projected rate is really, really good. Only Leigh Griffiths, Moussa Dembele, Odsonne Edouard, & Florian Kamberi had higher Projected Goals per 90 than Boyd (whose rate also ties him with Alfredo Morelos).
Can he replicate his form in this upcoming season? Would you really be all that surprised if he did??
So why do the numbers suggest Killie will score more goals this season? I mean Eamonn Brophy is also expected to see a little dip in his scoring. What gives?
When you dig through the numbers you see that the team was very reliant upon the goal scoring of Boyd & to a lesser degree, Brophy. Boyd accounted for 36% of the team's non-penalty goals, while the two combined made up 51% of the goals.
Kilmarnock weren't getting the kind of secondary scoring that you'd expect from a top six team.
The numbers suggest that will change next season. If you subtract Boyd & Brophy's shooting statistics from the club, Killie had a woeful Shooting Percentage of 22%. Their Expected Shooting Percentage? Well, it was 31%.
Killie were good at setting up quality chances for other players....they just weren't all that great at finishing off those opportunities. I'd expect the numbers to level out a little more next year, which will help make up for the expected dips in scoring for Boyd & Brophy.
Earlier this summer there were some whispers that clubs may be looking to sign Kyle Lafferty after he scored 13 goals last season (11 non-penalty goals & two penalties). Here's what I wrote when Craig Levein publicly addressed the rumors:
Lafferty's Shooting Percentage nearly doubled his xSh%. Whenever a player gets higher than 40% they're due for regression...& when your rate should have been 24% - the bottom may fall out.
The good news for Hearts is that adding Steven Maclean may help salvage some of Hearts projected dip in scoring. Maclean averaged 0.36 goals per 90 last season. That's equal to Lafferty's output. The difference is that Maclean's Projected Goals per 90 was 0.35.
A big part of that is the fact that Maclean is your prototypical poacher who makes his living in the six-yard box. The former St. Johnstone forward had 12 shots from within the 6-yard box last season (only Alfredo Morelos had more with 21). Kyle Lafferty? He had 4.
Lafferty scored five goals from outside the box (three from free kicks), which led to Lafferty taking 20 non-free kick shots from long range last season. That's what drove his expected rates down & it's highly unlikely that he will have a Conversion Rate of 16% from outside the box ever again in his career.
[This is awkward...I originally wrote this before the news that Rangers put in a bid for Lafferty. Clearly, I'm not a fan of this news & feel that it is pretty lazy recruitment on the part of the team I have supported all my life. Sigh...]
Aberdeen is an interesting case given the fact that they won't have Adam Rooney & Kenny McLean (or their 11 goals) in the lineup next season. Before we address that, let's see what the data from last season tells us.
A decrease of 1.64 goals isn't too extreme & you can see that there was very little difference in their expected shot rates & their actual results. It should be noted that those expected rates are pretty low & it would be advisable for Aberdeen's analysis team (if it exists) to look for ways to generate better chances.
Now, let's walk-you through how these kind of stats can help recruitment. Adam Rooney is gone. He averaged 0.34 non-penalty goals per 90 last season, while his Projected Goals per 90 was 0.28. If the goal is to simply replace Rooney based on his Projected Goals a player like Connor Sammon would have been the ideal replacement (too late obviously but he was available this summer). Sammon's Projected Goals per 90 was 0.30.
Of course, Aberdeen's goal should be to look for ways to make up for their expected dip in goals. What if we used Rooney's actual goal rate as a bench mark? Three forwards who were available on a free transfer this summer had Projected Goal rates that were close to Rooney's actual goal rate - Steven Maclean (0.35), Anthony Stokes (0.36) & Alex Schalk (0.38). Each of these forwards is no longer available & other factors made these options less appealing for Aberdeen (Maclean's age & Stokes propensity to being a fool). Alex Schalk would have been a fantastic signing for Aberdeen but alas he'll be playing in Switzerland next season.
But again, this is just an example of how Aberdeen could use stats to flag potential replacements & in reality they are not limited to the Scottish Premiership. If they could build a database for leagues around Europe that uses these kinds of stats, they could generate a list of potential replacements to be scouted pretty easily.
The fact that David Templeton is now playing football in England's League One really puts a dent in Hamilton's chances next season. Now, factor in that his Sh% was 41% & his xSh% was 33% & you have an indicator that even if he stayed - he wasn't going to average 0.46 goals per 90 again.
Hamilton will have a physical replacement for Templeton in the lineup...but it's highly unlikely that they will find someone riding a hot streak like Temps was last season.
Couple that with the fact that their two main forwards, Bingham & Ogboe, have abysmal Projected Goals per 90 rates.
It really won't be a surprise if Hamilton end up in the bottom two or three in the league next season.
Initially, there was some hesitation on my part to rely too much on the attacking numbers from Hibs last season given the drastic changes that accompanied the additions of Jamie Maclaren, Florian Kamberi & Scott Allan back in January.
However, the scintillating form of Jamie Maclaren is going to be really difficult to replace this season for Hibs.
Look at those numbers, he hit the target on 74% of his unblocked shots (expected rate was 53%, while he beat the keeper half the time on those shots on target [xSh% was 42%]). Beyond the fact that his accuracy & finishing rates are unsustainable (for whomever takes his place in the lineup)...it's more important to realize that those expected rates will be nearly as difficult to replicate.
Of the players who played at least 900 minutes & averaged at least two shots per 90 minutes last season, the only players to replicate Maclaren's Expected Shot Accuracy were Kamberi, Odsonne Edouard, James Forrest, & Scott Sinclair. Nobody was able to match his xSh% if they averaged 2.0 shots per 90 or more. Craig Tanner & Edouard were the only ones who came close with 40%.
I like Oli Shaw a lot & have advocated for him to take Maclaren's spot in the starting eleven. His Projected Goals per 90 was only slightly less than Maclaren's rate & his xSh% of 41%was actually almost identical to Maclaren's rate. The big "but" here is that Shaw only played 515 minutes last season. Can he really sustain those rates over an entire season?
Oh boy, that's not good.
Before you react to the names on the list, look at those Shooting Percentage numbers. Rangers overall success rate from Shots on Target was 39% - which was by far the highest in the league. The next two best Sh% rates came from Hamilton (ok?) at 33% & Hibs with 31%.
The fact is that Rangers are due for some real regression in their scoring next season.
"But...but...Josh Windass is being sold to the English Premier League! Alfredo Morelos isn't going to be our main striker! Daniel Candeias is going to be replaced!"
One, I'm not 100% sure any of that's going to happen & even if it does...the replacements are going to have to replicate numbers that are pretty unsustainable.
Maybe Umar Sadiq can walk in & replicate Morelos' goals per 90 of 0.50:
Yes, he averaged 0.58 goals per 90 last season....he also averaged 0.28 goals per 90 in the other three seasons he played before that (in very limited minutes). We don't have the data to predict if that 0.58 is sustainable but half of his brief career he's been an elite goal scorer, while the other half he's been a below average goal scorer. Sounds like recent commentary on Alfredo Morelos.
Even if Morelos & Windass do actually stay in the team & even if they regress to their Projected Goals numbers - those rates of 0.46 & 0.36 aren't bad. They're actually still pretty good & in actuality only represent about three of the projected decrease of 10.72 goals.
We're talking about a total of 180 shots on target, Morelos & Windass accounted for 65 of those. If you subtract the contributions of those two players, the rest of the team still had a Shooting Percentage of 37% & an xSh% of 32%. Hoping that replacements for Morelos & Windass will replicate their rates, while counting on the rest of the team outscoring the expected rates by that kind of margin just isn't realistic.
Rangers will score less goals next season....the hope needs to be that their defense improves enough that the overall goal differential doesn't take that big of a hit.
- Obviously Livingston & St. Mirren aren't listed here. However, Kenny Miller could be a very interesting contributor this season on the pitch. His Projected Goals per 90 was 0.35 (similar to the likes of Josh Windass, Craig Tanner, & Steven Maclean), but his actual goal rate was only 0.23 per 90 (his company this time was Kris Doolan, Rakish Bingham, & Stevie May). If (& it's a big "if") given his advance age, he could score at that 0.35 rate it would basically equal the likes of Kyle Lafferty & Adam Rooney's output.
- I don't recommend writing season previews to any of my peers....they take forever.
- Lastly, whenever I addressed recruitment in this post - it was limited to Scotland. That's not ideal. As the new seasons start across Europe, I will look to mimic some of this kind of data (Projected Goals) & apply it to leagues on the continent. To get a sense of what that may look like check out the new series Christian Wulff has been writing for the site.
- This was written under the influence of The Out Crowd, Chain & the Gang, Neil Michael Hagerty, Pleasure Forever & Television.