Dafuge & Data: Striker Health Check
It’s time for a striker health check. In previous posts I’ve covered our longball approach, fixing our right-back problems and end of season one recap. Now it’s time to check whether our strikers in season two are firing on all cylinders. We’re about 40 games into season two, so it’s almost over. But it’s not too late to check and tweak. Remember we must do well to appease the Seagulls.
Striker Dream Team: Kayode and Ethon
But why do we even need to check? We had a striker dream team in season one with Varian and Kayode netting 47 goals in all competitions between them. They were immense, and worked really well as a team. Why fix what isn’t broken? Well because it did end up broken.
Striker Sorrow
Like all beautiful things this striker partnership was fleeting. After the playoff loss at the end of season one against Farsley Kayode announced he was going to retire. In three days. This was despite having a contract until the end of the following year.
I begged, but to no avail. And Kayode retired to become a fitness coach. He then refused to sign for me because being a fitness coach in Scarborough, chip and candyfloss capital of the UK, is probably akin to rolling a massive rock up a hill.
We won’t hold this against Kayode though. I hear he is a thoroughly decent and pleasant man in real life, and he did give his all when he played for us. One day, one day we will tempt him back.
The Replacements
We needed someone new to cover the TM role. And ideally further cover as we had no idea whether Varian would be the same man after his ankle break. We ran the rule over three possible players – Blissett, Donaldson and Pressley (worst boy band ever).
It wasn’t to be though. We missed out on all of them. They were either lacking in interest, or we couldn’t pay them enough (the prize tickets for the arcades were not seen as being a suitable goal bonus). Painfully Blissett ended the season on 15+ goals for fellow Vanarama North team Gateshead. Instead, we had to look for youth, and look to strengthening the PF position instead. With a view to sliding Varian into the TM role. He’d moonlighted there in the past.
Small and skillful(ish). Not really what we like but the best we could manage. Cannonier arrived at the start of the season to take over the PF role. Salisbury arrived in February to bolster the front line. We still had back up in the form of Stanley and McClellan as well.
Striker Doubts
The season actually has been going well. And I think the issue less me doubting the strikers, and more me doubting whether I’m using them correctly. Our xG feels a little low compared to this stage last season. So it’s all about whether I could get more out of them. I’m 40 games into the season at this point. We have 6 left to try and cement our place in the playoffs again.
At this point, Varian has 26 goals, Cannonier 16, and then Stanley and Salisbury are drawing on 3 each. We are scoring, but we need more.
The Data
Back to the data we go for our striker health check. We need to numbers so we can see where the fine-tuning should occur. In the squad view below I’ve focussed on some of the important attacking statistics, and I’ve then added them to a spreadsheet that lets me calculate a few extra things like per 90 or per shot values.
Something to note here as well is that whilst Stanley and McClellen are on this view I’ve not included them in my deeper data dive as they’ve not played many minutes (351 and 278 respectively, so 3-4 games each).
In the above figure we have the xG value per 90 for each player plotted against the average xG value per shot taken. Letting us see which players get good xG totals per game, and which are getting good shots off (xG per shot). It lets us see who has an xG per 90 that might be inflated by taking lots of poor shots.
And here it tells us…well not a great deal. Varian is above average for the xG per 90 and xG per shot (about .16, so 16% of his shots could be expected to result in a goal). But Cannonier and Salisbury aren’t massively behind. If anything Salisbury is just edging it but remember he has not played as many minutes as Cannonier yet.
Importantly they all have the exact same shot on target percentage. They are all getting 52% of their shots on target.
This figure breaks down the assists, headers won and key passes per 90 for each player. Again not seeing huge differences here apart from the suprising aerial ability (considering his 5 for heading and 6 for jumping) of Salisbury. Oddly this suggests that maybe he could do a job in the TM role. To date he has mainly been playing the PF role and competing for that with Cannonier.
Third time is a charm considering figures one and two didn’t separate the strikers a great deal. In this, we can compare the total goals, total xG and the percentage increase or over performance when we compare the actual goals scored to the xG value. Essentially an efficiency value this simple calculation tells us how much a player is outperforming their xG.
Varian’s goal tally is a 35% increase on what his xG value suggests, and Cannonier is not much behind with an increase of about 30%. Way back in third place we have Salisbury with a poor bump of 1.3%.
It looks like we are currently getting better value, more goals than expected, from Varian and Cannonier.
Marginal Gains?
So do we have a striker problem? Not really. We have two strikers in Oakley and Varian who are over performing, and another promising striker in Salisbury. We just have a conundrum. Whats the best combination to use and why?
Now if this were real life I’d be worried about Varian being in his 2nd season of outperforming his xG, and of Oakley’s over performance. I’ve mentioned it before but in stats values tend to revert to the mean. So this peak of over performance should eventually dip. But this isn’t real life, and I think xG is under calculated. Whilst I’ll keep an eye on them for any dips I don’t think there’ll be any major ones. I will be looking into xG over performance more in a future post.
Instead of a striker problem I think we just have a case of how best to increase our marginal gains. I think I’ve learnt a few things:
- Varian and Cannonier are the 1st choice pair – They both have a good record and are overperforming their xG by about 30% when it comes to goals. Though the sample size is smaller Cannonier just squeaks an extra 1.3% out.
- Salisbury might be worth a shot at TM rather than the PF the coaches recommend – He’s winning even more headers per 90 than our targetman Varian. Rather than using Salisbury as the sub/reserve for Cannonier in the PF role, we should maybe try him out for the TM when resting Varian. Or swapping Varian into the PF role. After all, he got 20+ goals as a PF last season.
- The difference isn’t due to shot accuracy – All three strikers have a shot accuracy of 52%. So the over performance from Cannonier and Varian isn’t down to them being really good at getting shots on target (just over 1 out of 2) or being better at it than Salisbury. My gut feeling here is that there’s a mental component, whether that is positioning, composure or bravery, that is having an impact.
With this in mind, the plan was to play Varian and Cannonier, with Salisbury to come off the bench. Ideally to replace Varian.
Did it work?
We had 6 games to go. Six games to squeeze all the points out we could. I immediately cocked it up by forgetting to swap Salisbury and Cannonier around against Needham Market. It worked out though and we still won 1-0.
Then followed the 5-4 epic win against Guiseley. Salisbury came on for Cannonier at half time, swapped into the TM position. He almost set one up with a flick on before tapping home the winner in the 93rd minute.
Against the then 1st place Darlington we made the same swap at half time, and Varian and Salisbury not only grabbed a goal each but managed to shithouse. They set each other up for a near-miss before Varian dived over a defenders leg, won the penalty and then rolled around pretending to be hurt to waste time. He’ll be getting a bonus. Salisbury then fired the penalty straight at the keeper. It ended 2-2 but a spirited point against the league leaders. Importantly this got me thinking that maybe I had made a mistake shifting Varian to TM for 40 games. Clearly, he is better as a PF.
We then face 3rd place Gateshead, who started missed opportunity Blissett against us. In the 3D match he looked like his spirit animal was an actual brick shithouse. I was most upset to have lost out on him. That said, it didn’t take long for us to stamp all over Gateshead. Again at half time Salisbury came on, and Varian swapped to TM. Leading to a brace for Salisbury, and a goal and an assist for Varian. We finished 4-3 winners.
Now it was clear that whilst the data said Cannonier should start the data was also based on Varian being TM, and Salisbury being a PF. With them switching because of suggestion two in the list above it was a different matter entirely. So I made the decision to start with Varian as PF and Salisbury at TM in the following match against Kettering.
Against 17th placed Kettering we came unstuck, but largely due to bad luck. Salisbury set up a goal, Varian set up a disallowed goal, and most set pieces involved the two of them sowing chaos. But it just didn’t work out. With this loss, any sneaky title hopes were dashed, and we were only in the playoff spots by a point. A very loose table.
It all came down to the last game against Chorley, 19th place. A cagey game, but we squeaked it 1-0 and made it into the playoffs.
So over the course of the last 6 games, our strikers scored 5 goals, and got 4 assists. As well as a cumulative xG of about 6. We can’t compare to what would have happened without our changes but knowing what we do about the average xG per game before we can work something out. With Varian, Cannonier/Salisbury (a 60/30 min split for ease) without any changes we would have got an xG of 4.5. We beat that by 1.5, and got a slight overperformance of 0.5 in terms of goals.
Fine margins indeed. Over a longer run of games I imagine these margins would have increased.
And the Playoffs?
We do have an extra run of games. I don’t want to reveal too much about the playoff results just yet but over the non-disclosed additional matches, my strikers brought that to a new total of 12 goals, got 6 assists, and an xG of 9.
If we go back to the original figures looking at xG per 90 and shot, as well as the percentage increase or overperformance we get a slightly different picture now.
Salisbury has shot up the rankings. He’s bagged plenty of goals from the TM position, has an accuracy of 62%, and has a relative increase in goals from xG of just over 40%. What a difference a little bit of extra data makes. He’s now the front runner to partner Varian next season rather than Cannonier.
Why did this work?
This striker health check made me think about the strengths and value of the three players differently. As a result, it got me out of the narrow focus I had on Salisbury just being a PF. By swapping him in and keeping Varian on I got to play two strikers who were better in the air, which suits the longball game.
Whilst there wasn’t much between Cannonier and Salisbury when both considered as PF’s, the gap widened when Salisbury got the nod as TM because of what the headers won per 90 stat revealed. A new partnership was formed that was a better fit for my style of play, even though Cannonier has had a pretty decent season as PF.
The stats helped me break out of my fixed thinking (functional fixedness for any Psychologists in the back) and try something a little different.
Pingback: Dafuge & Data: Stats and Summary Season Two - FM Tahiti