Dafuge & Data: Dream and DNA (and season summary 5)
This one might be a little longer than you are used to. The season summary too. Alongside a recap of our first season in League 2 as part of the Dafuge challenge, I’ll also be covering a preliminary analysis of my team’s DNA. We already know it’s partially made of chips, rock (the eating kind), and violence. The blood of the Seagull pulses through our veins. But we will take a bit of a non-standard statistical dive into what attributes make our team tick.
If you want you can skip the season summary and get straight into it HERE. But I’d be tempted to stick around as it was a very bizarre debut professional season. If you are late to the Seagull infested party though check out the links below for more about the haddock and chips powered journey so far:
So we made it to the actual league. No more semi-professional nonsense for us. But with those changes came the dawning realisation that the league was going to be a lot more competitive. Like a lot. We had a tiny wage budget, a low reputation and lets be honest, not the most desirable of locations. It’s hard to recruit when you’re on the coast of the North Sea.
But we tried. With a slight increase in stature and wage budget, came a slight improvement in our recruitment pool.
Not by much though. We were still competing with non-league teams for players. Whilst we had a seat at the big (well bigger than lower league) table our chair was awkwardly placed and our cutlery was yet to be supplied in time for the summer transfer banquet.
We did raid a lot of former top-flight youth academies for their rejects. This was partly because so few decent senior players wanted to come to us. And partly because they were low risk. They wanted low wages, and if they met their potential then we would be laughing. If they didn’t then at least the cost was minimal.
We shored up the strike force with Sasaki and Fairs. As well as Winn. Though the cooly named Sonny Winn was also going to be making appearances as the wide targetman. It gave Varian some more choice upfront and if you’ve read the love letter to Varian published earlier you’ll know it went well for at least one of them.
Lehane and Carlo on the left came in to provide some cover and competition for the leftback and left wing roles.
Another promising player was Duggan in the attacking midfield spot. We don’t actually have an attacking midfield spot but he looked like he could work as a box to box midfielder.
Sales and Loans
We also shifted some players for a little bit of money. Nothing vast but it was the start of our wheeling and dealing as a seller. Or loaner.
Hardly big bucks but it gave some players who had been quite useful and loyal a chance to prove their worth elsewhere. We’d made a very quick jump to league football, and despite Mallon and Oakley being key to that they were increasingly looking like they needed more time to make the jump.
Last season we had introduced a range of tactics before sticking with the Klingon inspired Quapla. A lopsided 4-4-2 that used a wide targetman to create more space and even up numbers going forward. But would such an attacking formation work after our promotion? Well, we trusted the system and work it did. I’m not going to try and build up the suspense as basically there was none.
We got to 1st place after a few games, never dropped below 3rd after that and had promotion and the title wrapped up with games to spare. Varian was on fire despite some injuries and scored 27. Fairs stepped up to replace him well with 18 goals, and Sasaki got 15 goals and 13 assists. Win also chipped in with a healthy 11 goals and 10 assists from that wide targetman position.
We were helped immensely by Harrison in net as well. Despite being massively under pressure in terms of shots, and xG against we managed to keep the damage down. The conversion rate against us was low. I think this was partly due to his superhero efforts between the sticks (21 clean sheets) and our defence limiting the number of decent shots. The xG reflected a lot of poor shots, rather than a handful of good shots.
I think impenetrable sums up the defence. As well as dirty and needlessly aggressive. I know I’ve said I’m not a fan of some of the season comparison graphs but the one below kind of highlights that whilst we had lots of shots to deal with they either weren’t the greatest or our keeper was the greatest. Either is fine with me.
What this all means for us in terms of general stats is that whilst this was a dream season it was a dream due to over performance. We can’t be sure of repeating this success as this wasn’t even a success that the stats would predict. The expected points, taking into account xG and xG against had us much further down the table in 12th. Not even in the playoffs.
Well once we’ve come down from the promotion high, and recovered from our lambrini soaked hangovers there’s going to have to be some serious rebuilding. Or some serious future-proofing. Promotion is nice. Promotion is lovely. But if we punched well above our salary weight in league two then league one is going to be a whole different mountain to climb.
Most of our players are consider future Vanarama National players. Not even stars. Just you know, squad players. We’ve beaten the odds two seasons in a row. Flamingo Land Park, though expanded, is going to be one of the smaller stadiums in terms of capacity and attendance. Next season there’ll be teams like Huddersfield, Peterborough, Doncaster, QPR, Sunderland, Ipswich, Charlton… plenty of teams with much stronger backgrounds than the lowly Seadogs.
Lets get some self-plagiarism in with a bit of a quote about DNA.
The idea that there are certain properties or attributes you can reduce the team to. Essentially attempting to embody a particular football philosophy within your team by recruiting players with those attributes and traits. Whether that is the high press of Liverpool, the creative use of space of Man City, or the borderline aggressive psychopathy of the 1980’s Wimbledon Crazy Gang […]
However, how do you go about determining what the club DNA should be? There are various approaches. For example you can simply distil the philosophy down into the key attributes that logically fit that approach. If you want a hard-working, never-say-die club that don’t crumble under pressure then your players would logically need to have high work rate, team work, and determination. Perhaps even bravery in addition to this. Following this decision you can then just recruit players with higher values in these DNA attributes. Hoping that the attributes selected translate to on the pitch performances.
Yet this approach misses an important factor, an important variable some of us nerds might say. You might know what philosophy you want the players to live and breath but are they actually playing that way? Your tactics will play a part in this. Obviously, there’s no point having attacking DNA if your tactical set up doesn’t mirror this. Furthermore what you are asking your players to do tactically might throw up some surprises.Me, in that other post about DNA for Dictate The Game
As alluded to in the first FM19 article I did about it there are different ways of approaching it. Often people distil their team DNA down to a few different attributes they think sums up either how they play or what they want their players to embody. Fibra is often the one that is thrown up, and there are plenty of spreadsheets out there for helping you check if your players fit your selected DNA.
My approach though is to set up the style and tactic we want. Get it working generally, and then dig into the stats to see which attributes predict success in terms of the KPI’s for that playing style. So DNA driven by the data.
I do this by using something called a Regression analysis. It essentially checks whether a group of values can predict one particular outcome value. Do increases or changes in the predictor variables, significantly change the outcome variable. With this approach, I can take a stack of attributes from players, and see if any of these attributes are significantly linked to an increase in whatever KPI I choose. Goals, average ratings, key passes, and even now for FM21, xG.
Genuinely think I’ll find it hard to explain my approach to the game better than I did in my original DNA article on this blog. I really enjoyed writing it so please head over HERE if you want to check it out. If you don’t just imagine that 3 Men and Baby is being rebooted, but the 3 men are Big Sam, Dyche, and the late 1980’s Wimbledon side instead of Selleck & Co. And the baby is basically the football we try to play.
I’ve got near on 2 seasons of aggressive football data to use now. Since switching things up in the Vanarama National, and moving to the Klingon inspired tactics we’ve played a few games. We’ve not got an ideal amount of data as Regressions take a lot to work properly (you need about 80 data points or in our case 80 players with a seasons worth of data, plus 20 more for each extra variable you look at). We would need around 500 players worth of data, or almost 10 seasons worth for the ideal Regression. But you can make it work with less. We’ll have to otherwise the Dafuge challenge will be done before we get to look at the stats.
What do we already have in our DNA?
We had a hoof it harder hoof it longer tactic in FM19 that was uber-violent and direct. In FM20 we managed the same again with Belfast Celtic, beating many disciplinary records as we made our way up the leagues. The DNA test from those tactics revealed the following attributes as being key for a range of our KPI’s:
- Jumping Reach
- Positions specific – Marking & Finishing
The above is fully explained in the original DTG article but basically the bigger the number in brackets the stronger the relationship between the attribute and the KPI at the top. This was from the FM19 analysis. But in FM20 not much changed the same general relationships seemed to be in place for our hoofball approach.
New game, a new tactic, new KPI’s, new analysis. Our Klingon tactic is a little different from previous versions, and our match engine might be very different. We also have fun things like xG to look at. Though xA is still something that has to be done manually so…isn’t getting done by me here.
I ran the regressions again but with some of the extra KPI’s like xG. And with the older KPI’s just to see what had changed. If you want to run this sort of analysis yourself you’ll need software that can perform multi-linear regressions. So either SPSS, PSPP, JASP or even R if you really want to nerd out.
With this very attack focussed KPI comes a few new attributes to consider. After breaking down the overall xG stat to xG per 90 we get the below:
A few surprises in terms of what is not there. Finishing and heading, key attributes previously for a lot of our attacking KPI’s aren’t significantly related to xG. It sounds a bit counter-intuitive as that’s how the ball gets in the back of the net but what we’ve found is that for our tactic, and our team, there are much more important factors at play.
Whilst our favourite of aggression is still present, we’ve now got mental attributes like anticipation and off the ball combining with physical attributes like balance, agility and natural fitness. If that strikes you as odd for a route one side don’t worry. This is actually reflecting that we play a very direct yet counter-attacking style*.
The anticipation and off the ball movement help our players get in the right positions and react to the sudden change in play. This is also important with helping get to 2nd balls when you’ve got two targetmen and a pressing forward sniffing around.
The balance helps with recovering from aerial balls, and the agility with the changes in direction and pace needed in a counter-attacking game. Natural fitness is presumably there because it’s such a physically draining way of playing – putting the hurt on the opposition in our half and then sprinting the rest of the pitch as part of our counter-attacking moves every chance we get.
xG side note
*Also, being picky, xG represents chances in good goalscoring situations or positions, not whether they are actually scored just that they have a good chance of being scored. So heading and finishing, and whether the chance is on target or not, is largely under most xG models unrelated to the quality of the chance itself.
New DNA Strands
When looking at the rest of the KPI’s the usual suspects made an appearance, but the new strands of balance, agility and natural fitness kept cropping up.
Again this seems to represent that a lot of the roles and positions, to succeed, needed to be able to deal and recover well with aerial balls (balance), change direction and pace quickly in attack and defence (agility), and go the distance in what are intense games (natural fitness).
But what about other stats?
No other consistent surprises appeared. Unfortunately, I couldn’t dig into stats like clearances and blocks. I think these are key for my kind of game but whilst you can see them on the player/team stats in the league stats view you can’t get them in the squad view. Making it very difficult to use. Key tackles have only recently been fixed as well so I can’t really check on that until I’ve played another season and a half.
Expected assists would be great, as would packing. But neither exist in-game. You can try to work it out yourself, which I have done, but it’s too time-intensive to be a sane hobby. I’d also like to see winning the 2nd ball, rather than just possession won as a stat but until it appears in-game I just have to use the eye test to check on what’s happening. Unfortunately, the eye test is rubbish no matter how good you think your eyes actually are. Data is supreme, everything else is just varying degrees of error and opinion.
Much like a poorly managed pandemic response over time, we will have to deal with some DNA mutations in our team. As the level of play changes, the tactics change, and the tactics we face change what makes us successful will also morph.
Now we have the base DNA in place though we will be revisiting this every season, and every new (tactical) variant to see what is making us tick.