I have been thinking about how to go about doing this review and have settled on doing it in segments. My goal is to lay out an analytical framework for measuring historical performance for Celtic in order to offer context for making performance assessments. This first segment
will lay out the framework and why I am doing this. I decided to look into this as an extension of the analysis I started doing on the recent Celtic seasons about 18 months ago. Like when I began 18 months ago, I didn't have a good sense of what an analysis would yield.
Let me also state that I do not presume to "know" anything. I am putting forth my analysis using the tools I have found available. I've tried my best to source accurate data but there will be some estimations.
I am sharing this because I was going to do it for my own curiousity
anyway and hope that my conclusions may be of interest to someone in a position of power to make some sort of positive difference. Obviously, my hope on that front failed miserably as it related to this season.
A simple way to communicate this performance model is to think of it
in 3 parts:
Wage Bill
Player Trading
Player Utilization
I have a strong sense that the people running Celtic do not actually understand what drives the performance of football clubs. That probably sounds arrogant, but I base this assessment on the confluence of decisions
made along with statements from leaderships at the club and those around them.
Wage Bill- this is the big one. There is a robust body of research on this topic dating back at least 30 years, with the 2007-2020 period including significant research contributions by people like
Bernd Frick. I am cheap so haven't paid to read all of the papers, but read some & done my own simple regression analysis across sample of leagues using wage bills, expected goal differential, expected points, etc, & there does seem to be a relatively stable contribution for wage
bills with a range between 75%-85%. There is normal variance season to season but also relative to the wage distribution in leagues. For example-the SPFL has a vastly different distribution from the EPL.
I start with this concept because it is foundational but also common sense.
For perspective, the 5,000-1 Leceister City team had a wage budget relative to Manchester City which was comparable to Aberdeen relative to Celtic. Outside of Rangers, there is no viable sporting challenge domestically for Celtic. Of course these factors are longer term- meaning
that anything can happen in single games or ties, but over a 38 game season, wage bill translates into a competitive advantage that is incredibly difficult to overcome when the disparity is 6+ to one.
This provides a simple first step in assessing the relative wage bills
of Celtic vs domestic competition historically. With the gap being so large vs other clubs, the obvious variable is the wage bill relative to Rangers.
I've cobbled together data dating back to 2002 for both- sourcing Rangers' data prior to liquidation required some research.
Price Waterhouse conducted a survey/report over many of those years I was able to find online. Celtic obviously have historical PLC public reports. The Global Salary Survey offered additional context starting in 2013-2014. I made educated guesses in my data for the 2 seasons
where I could not find specific amounts for Rangers, which were fiscal years 2005 and 2006.
After wage bills comes the next two, which including the quality of recruitment/trading, sports science, academy and player development, and 1st team coaching. Combined, they are
dwarfed by the importance of wage bill. A well run team the size of Aberdeen has little chance against a poorly run team of Celtic's size. I contend that this simple concept is at the root of how and why Celtic's leadership have misattributed the accomplishments of
the club, CEO, and current manager. As the data and analytics revolution has unfolded, player trading and utilization have made the "old" model uncompetitive. The idea that a single person can compete at player trading and utilization vs modern tools and expertise is
like saying a thoroughbred horse can compete with an Edsel. Even the best of past eras could not compete
today.
The next segment will begin to look at the historical record dating back to the O'Neill era and immediately preceding the current CEO's tenure. End.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Time for the good, the bad, & the ugly from yesterday's Celtic at St Johnstone game. Had a long morning of driving today so a bit late posting. Non-penalty xG was 1.94 vs 0.34 on 10 vs 5 shots. That differential of 1.60 was very close to the avg I shared ahead of the game- 1.51
So I would characterize the overall output as fine. Similar to the St Mirren game, the dispersion of performance levels in the game was skewed late, as had 0.94 xG before Edouard's 1st goal, with 0.40 of that coming on the Duffy chance off the Turnbull free kick delivery.
Celtic also won just 42.32% of duels overall- poorly placed/sloppy passing was widespread. Christie offered 2 moments of excellence which enabled the same from Edouard- other than Edouard's 1st half cross, that was pretty much the entirety of Celtic's attack.
Performance benchmarking for today's Celtic at St Johnstone game. Non-penalty xG had avg 2.23 vs 0.72 on 17.70 shots in Perth. Celticnhave outperformed that avg xG by averaging 3 goals over the 9 league games since the start of the 2015-2016 season.
St. Johnstone are probably
the oddest statistically team in the league this season, as they are currently 3rd in xG differential overall after Rangers and Celtic, and have conceded about the same amountof non-penalty xG per 90 minutes as Celtic.
However, their finishing & keeper play have been very poor.
They've scored just 24 non-penalty goals vs over 34 in npxG and conceded 34 non-penalty goals vs just 24 in npxG. That is a massive 20 goal swing overall and explains why they are 8th in the real table instead of 3rd.
Clark has been particularly poor this season overall, avg
Time for the good, the bad, & the ugly from yesterday's Celtic vs Rangers game. Non-penalty xG was 0.14 vs 1.26. To address one of the hot topics coming out of the game, next tweet shows other games with low number of shots on target since Deila's last season-only gladbach @ home
A reminder that event-based xG models are just one of many analytical tools. For example, Wyscout's assigned very low probability to Moi's chance. As has been the case this season generally, we lacked attacking output from midfield and the two "strikers."
Understanding why xG was low on Moi's shot is important- he struck the ball with his right foot- from that position and relative to McGregor and the last defender is probably not a shot that goes in...almost ever. Had he used his left foot the xG would likely have been higher.
Performance benchmarking in advance of tomorrow's big game: first some ranking models. 538's SPI currently has Celtic at 52 and Rangers at 99, which places them just behind Sporting and Braga for some context. The Elo system I monitor has Celtic at 37 and Rangers up to 51.
The derby games the past two seasons have been very tight both with regards to goals and underlying performance metrics. Back to Wyscout data today to go back to 16/17 season, and we can see the shift in relative derby performances pre and post Gerrard/Beale arrival.
The relative wage budget went from about 15% in 16/17 of Celtic to over 70%, which has assuredly been a huge factor. Notable development in last 3 derbies was the extreme lopsidedness of Celtic's attack in 4-2-3-1 in the three games. That had not been an issue under Rodgers.
Warning- this is a nerdy thread on long range shots and why I worry about Celtic's shooting. I hear repeatedly that Christie is a good shooter so I thought I would examine his record from shots outside the box and the penalty arc. I reviewed 17-18 at Aberdeen & 18-19/19-20.
Christie had 274 non-penalty shots over that time with 147 outside the area. He had about 56% of his shots outside the area under McInnes and last season under Lennon, while 48% in the hybrid season under Rodgers/Lennon in 18-19. He scored 8 goals on aggregate of 4.62 xG...
on those 147 shots for avg xG per shot of about 0.03. One way to interpret that is that Christie is a "good shooter" because he has outperformed his xG. Of the 147 shots, 20% were on target, 43% were wide, 32% blocked, and 5% scored.
Time for the good, the bad, and the ugly from yesterday's Celtic vs Ferencvaros game. Wyscout had the game at 1.72-0.31 xG with 25 vs 7 shots, 8 vs 3 on target, and pretty much 70-30% possession. I would say that the main good is that despite Lennon's post-game comments, players
appear to have played hard overall. I think players generally seemed to try to make the best of flawed selection and tactics. I am not going to slaughter players who are not good enough at no fault of their own, or playing out of position for the umpteenth time.
Lots of bad, so I will just highlight a couple and they will be familiar. Been harping on midfield structure, lack of creativity, and transitions being problems, with stepping up to better quality opponents likely to exacerbate. Here is typical picture of midfield issue.