Chattanooga FC dominated eastern division foes Flower City Union Saturday night. A four-goal haul from Markus Naglestad highlighted the most dominate performance of any NISA team this season. This was the first time a player netted four times in a game for Chattanooga FC as well as the first of any NISA player.

Chattanooga deserved every goal, generating a whopping 4.35 xG. To put that in perspective, the next highest this season was 3.5 as Maryland blitzed Flower City in week 2. The average has been floating around the 1.5 range. Usually I would include possession too, but without the clock on the eleven broadcast it’s super difficult to track accurately. However, I’m rather confident the field was tilted in Chattanooga’s favor a majority of the match.

Chattanooga had no problem generating chance after chance. There seemed to be some extra impetuous in the CFC attack to possess the ball in dangerous areas and create chance after chance. You can see above that there weren’t that many swathes in the game without some attack, although the team slowed down in the second half having already put the game to bed by halftime.

Markus Naglestad was of course the standout performer, scoring 4 goals from 1.85 xG, the most xG generated by a single player in a NISA game, overtaking Ian Cerro on opening day (remember penalty save and tap-in). It was great seeing the whole team pour forward in attack. Earlier in the season it seemed like players at the top of the Christmas Tree were getting isolated when breaking forward, but CFC threw plenty of bodies forward and the front three felt like a real front three. Markus Naglestad sort of reminds me of Olivier Giroud, holding up the ball to distribute to players around and scoring bangers.

With this result and the extra points against Valley United, Chattanooga FC are poised to take first with a win over Bay Cities on Saturday. It seems like just a few weeks ago we were bottom of the pile, but as I noted then, the team is playing great, fluke results will come, but in the end the proof is in the pudding and this pudding has CFC rising towards the top.

A Note in Methodology

These stats missed the usual 48-hours-after-kickoff window as I realized my xG model was starting to devalue good shots. I was using logistic regression to predict xG, but several outlier shots started to stretch the scale of the regression.

To mix things up, I started using decision trees like the one below. Basically, R takes all the shot observations and starts choosing the best way to split them up into groups. After generating a thousand unique trees, I then find the average outcome to predict the xG of a shot. In the future, I may just use this to pull out the bulk of shots that were never going in and regress the rest. Data exploration like this is why I even started this project. If you have ideas to use feel free to reach out!

As always, thanks for reading!