BOROCHAT
Newest article: Re: Match thread - Blackpool (A) by CloggerYesterday 21:20Yesterday at 21:20:53view thread
Oldest article: WATN by MerseyBoro2/8 08:51Fri Aug 2 08:51:13 2024view thread
Next thread: REVELL IN by Mr Fickle3/12 21:43Tue Dec 3 21:43:54 2024view thread
Our Attacking Issues (using Statistics!)
Views: 630
I’ve shared everyone’s frustrations with our attacking returns and performances recently, so I decided to try and figure out the problems using some of our stats from the league (and therefore didn’t include this weekend’s game).*
I had a working theory that Tyreece Simpson could solve our problems.** Now, since I can already hear you all typing furiously in disgust, I decided to use stats to preemptively prove the doubters wrong. What I found has vindicated a few of my pre-existing theories, but mostly shone some light on some of the underlying causes of our problems up front. Spoiler alert: the solution isn’t Tyreece Simpson.
Before that, I need to write a quick note on how to read correlations. Essentially, the way they work is that all relationships (say, for example, Reid’s minutes played vs. goals scored by the team) are graded on a -1 to 1 scale. A -1 would mean that when Reid gets a low number of minutes, we score more goals every single time he gets few minutes. A 1 would mean the opposite: when Reid gets a high number of minutes, we score more goals every single time he gets more minutes. A score of 0 would mean that Reid’s “minutes played” has no bearing whatsoever on us scoring more or fewer goals. Now, of course, this won’t happen in real life in such a cast-iron way, but there will be some degree of relationship between BGR’s minutes and total Stevenage goals that sits somewhere between -1 and 1.
Let’s look at strikers first. By assessing each of their playing minutes against the team’s Goals, Shots, Shots on Target, and Shots on Target as a % of total shots, we can get an overall view of how the presence of each of them impacts these statistics.
Tyreece Simpson
As I said at the start, I came into this thinking that Simpson would be the key**. However, his numbers just don’t back up my theory. Simpson has a near-0 correlation between his minutes and our Goals. That is to say, his presence or absence on the field essentially does nothing to (positively OR negatively) influence our goalscoring threat. Simpson getting more minutes does have a slight positive relationship with Points (0.11) but this is quite a small relationship so we shouldn’t give it too much weight.
Jamie Reid
Sadly, BGR suffers with the same conclusions as Simpson, also seeing a near-0 correlation between his minutes and our Goals. Reid does not even have the slight positive correlation with Wins as Simpson does. This is deeply concerning, as he sits 1st in our attacking players “minutes played” table with 939 minutes played (with Elliot List next highest with 552 minutes played).
Elliot List
Listy is where we start to see some positivity. As he plays more minutes, we can see a fairly weak, but still positive association with Goals scored (0.2) and more Shots (0.12). This is interesting because it suggests that List (and not Simpson) was the driving force behind our overperformances in the first 8 games (having played 55% of all possible minutes) compared to the most recent 8 games (only 21% of all possible minutes). However, a key caveat is that List has a fairly negative correlation with Shots on Target (-0.18), which may be an illustration of his wastefulness in front of goal when he does get an opportunity.
Jake Young
Young is where we start to see the most interesting results, as he specifically is an absolute powerhouse in the stats. He has by far the strongest relationship between his minutes played in a game and Stevenage scoring during that game (0.4), and this performance is mirrored by his influence on Shots (strongest relationship out of all strikers with 0.2), Shots on Target (strongest relationship out of all strikers with 0.37) as well as Points (strongest relationship out of all strikers with 0.31). For reference, the relationship between Young’s minutes played on us getting more Points is nearly 3 times stronger than Simpson’s (our next biggest contributor).
List + Young
With the knowledge that this pair were statistically our best-performing strikers, I looked across all our league games to see how often they had played at the same time (across the first 90 minutes of games) and I found that they have been on the pitch together for only 1 minute, at the end of our recent game against Reading. The performance (albeit not the result we wanted) against Mansfield has only worked to solidify my confidence in this attacking pairing.
Having looked at all this, I wanted to test another theory: that Roberts & Kemp do not work when played together.
Kemp + Roberts
I sort of see these guys as similar to England’s issues, with too similar players getting in each other’s way. Therefore, I totalled up the amount of minutes per game that Roberts & Kemp are on the pitch at the same time, and assessed this against the attacking metrics as I did our strikers. What I found was that Kemp & Roberts getting more minutes at the same time is the number 1 negative correlation with Goals (-0.35), Shots (-0.26), and Points (-0.19). This combination of players contributes more negatively to our attacking output than any other of our attackers on an individual basis.
(On a side note, I also investigated how often Kemp & Young play together in the league and it’s depressingly low: a combined 71 minutes across all our league games so far this season. Annoyingly, I therefore can’t assess their relationship with our attacking performance.)
Midfield
Lastly, let’s look at our midfield. In short? It is bad folks. This really shocked me as there hasn't been a lot of thought or discussion in this area, although you could argue that chance creation should really be originating from our midfield.
Looking at individuals, both Harvey White and Dan Phillips have a negative impact on Goals (-0.24 and -0.26 respectively) and Shots (-0.25 and -0.16). White has a particularly bad relationship with Shots on Target (-0.46), which is the single strongest correlation that I could find across all players and metrics. This suggests that White’s perceived ability to create is potentially completely false. A bright spot for Phillips is that he does show a strong correlation with our Shots on Target (0.32), suggesting that he is actually quite effective in chance creation. Louis Thompson is another story completely. If we look at the stats, he’s actually hugely effective for us going forwards, with us having more Shots (0.19), Shots on Target (0.42), Possession (0.15), and Wins (0.19) when he gets more minutes in a game. His relationship with Shots on Target in particular is powerful and suggests that he’s one of our biggest assets when it comes to creating more dangerous chances.
Conclusions
The statistics seem to suggest that we should attempt 2 specific fixes to try and turn the team’s performances around. First, List & Young should be given a run of starts together, to see how effective a combination of them would be, as individually they’re our best strikers (on paper). Secondly, we have to stop playing Roberts & Kemp together as it simply does not work at all, across any metric. If I was to suggest a third fix, less backed up by the stats but by pure hunch, I would argue that it’s worth actually playing Kemp & Young together, to at least see if they can reproduce their form from last season.
I would love to be able to do these assessments with formations in mind (as I cannot fathom why anyone would ever play Kemp, List, or Young on the wing as we’ve often seen this year), but the data gets quite messy and this was just for a bit of fun!
TL;DR: Simpson isn’t the answer, Jamie Reid is average at best, Jake Young is our best striker statistically, and we should really reconsider the utility of playing Roberts & Kemp together.
If you reached here, thanks for reading (I appreciate it’s a bit of a slog!), and I’d love to know your thoughts!
*Some of the statistical relationships I found should be taken with a pinch of salt. In addition to the fact that these are just basic correlations, there frankly isn’t a lot of data, and we therefore can’t really draw any cast-iron conclusions. For example, as an Aaron Pressley critic, I’d love to be able to say that we lose more often when he gets more minutes (the stats do suggest that this is the case). However, he has only featured in 4 games, and his maximum playing time in any one game was 45 minutes, so the validity of any inferences we can make are shaky at best. I’ve tried to only highlight the strongest conclusions, but they aren’t watertight by any means.
**Hear me out. In the early season we were creating chances. Granted, we weren’t scoring many (and I remember List & Simpson in particular missing some absolute sitters), but the chances were coming. If we split the season in 2 (first 8 games vs. last 8 games), Simpson had played 50% of all possible minutes in the first 8 games, compared to only 18% in the last 8. I thought that his presence and work ethic made up for his lack of threat by creating space for others to exploit (as Beardsley did so famously in that team). Turns out it was (probably) List. This was my logic – I accept I was wrong, and you may now boo me accordingly.
reply to this article | return to the front page
1 person
Re: Our Attacking Issues (using Statistics!)
Views: 287
Without knowing the stats, this was pretty much my gut feeling.
Young shouldn't be pushed out on the left though.
reply to this article | return to the front page
Re: Our Attacking Issues (using Statistics!)
Views: 473
I admire the intent, though not sure how much you can take from it. Correlation doesn't necessarily mean causation and in football there are so many other variables it would be difficult to say with any confidence.
Also a relative small dataset (because we don't score enough goals lololol).
Having said that the data somewhat supports my views, which is nice.
I rate Young and Kemp if played correctly. I don't rate White and think that's our main problem; relying on one midfielder to create anything who isn't very good at creating anything.
I rate pace, and List is our quickest. I rate 2 up front, with one being quick to run off the other, and the other one being able to hold up the ball or win headers etc. Appéré would have been an asset in that respect, and is theoretically a miss.
I also rate width and pace on the wings but we don't even have the option of that let alone any data to interrogate.
Edited by Sev at 13:45:21 on 3rd December 2024
reply to this article | return to the front page
2 people
Re: Our Attacking Issues (using Statistics!)
Views: 272
All these posts are too long to read.
reply to this article | return to the front page
1 person
Re: Our Attacking Issues (using Statistics!)
Views: 371
Oh yes, I absolutely agree on the correlation/causation point. Of course you'll get better results if there is more data, but if we saw some very strong correlations even in this small sample then they would definitely still deserve to be studied, especially in low-scoring scenarios. If we had scored every time Young started it could absolutely be a complete coincidence, but on the balance of probabilities you can make some initial inferences which you can test more thoroughly later on.
The goal was to put player minutes & combinations to a rather basic test, and this can be really revelatory when you compare different players under the same (admittedly poor) assumptions. The fact that the data does point in the direction you'd think it does judging "by eye" gives me confidence it has at least some basis in reality.
The Harvey White stats really were the bit that shocked me, and it's pretty much exactly what you said. I think Forster-Caskey is a huge, huge miss, and his reduced minutes towards the latter end of last season could be quite meaningful form-wise, but then again, we were creating at the start of the season (even with White and without JFC), we were just missing absolute sitters, which is something that's only changed in the last month using the same players!
reply to this article | return to the front page
1 person
Re: Our Attacking Issues (using Statistics!)
Views: 308
JFC is a cracking player. He linked up very well with Reidy too, perhaps he's the missing link?
Stats aside, we just need to stop slowing the game and over playing it. It's got to the stage where a single chance is so rare the players snatch at them. JUST SHOOOOOOOT
reply to this article | return to the front page
Re: Our Attacking Issues (using Statistics!)
Views: 174
It may seem ridiculous to some, but the missing link is probably Kane Hemmings. Reidy was at his most prolific when playing alongside KH, who was largely underrated.
reply to this article | return to the front page
2 people
Re: Our Attacking Issues (using Statistics!)
Views: 359
The Harvey Shite stat doesnt surprise me. Chocolate teapot.
Will be interesting to see how we go without Robbo tonight.
reply to this article | return to the front page
Re: Our Attacking Issues (using Statistics!)
Views: 290
One less yellow card.
Edit: if you say that a few times it becomes a tongue twister or maybe it's just me.
Edited by Sev at 19:07:34 on 3rd December 2024
reply to this article | return to the front page
Re: Our Attacking Issues (using Statistics!)
Views: 258
*fewer
Saving Chuds a post
reply to this article | return to the front page
Re: Our Attacking Issues (using Statistics!)
Views: 257
That was half the fun of posting it.
reply to this article | return to the front page
Re: Our Attacking Issues (using Statistics!)
Views: 431
Maybe Revs & all coaches should have a read they might learn something.
reply to this article | return to the front page
Re: Our Attacking Issues (using Statistics!)
Views: 433
Be interesting to do this analysis again after another dozen or so games when you have more data.
Also be interested to see Roberts' data from last season.
reply to this article | return to the front page
Re: Our Attacking Issues (using Statistics!)
Views: 482
Just stick it in the back of the net
reply to this article | return to the front page
Previous thread: The table by RockyBottom3/12 23:41Tue Dec 3 23:41:01 2024view thread