Planning beats probability — Part 2
I recently published an overview of Everton’s 2023 relegation battle, and I made the argument that our innate attempt to predict the future is less useful than simply planning for different potential outcomes.
Unfortunately, because Everton were also in a relegation battle in 2024, this gives me an opportunity to explore this issue even further.
The main argument against assigning probability estimates to potential outcomes is epistemic — there is a fundamental difference between applying the concept of chance to sets of known distributions, versus dealing with unique events that take place under conditions of uncertainty.
It also adds an element of false precision. Consider the Opta Analyst. They have a model that works in the following way:
As of February 26th, it provided the following probabilities:
What does it mean that there’s a 16.4% chance that Everton will get relegated? I don’t mind saying that Sheffield United and Burnley look almost certain to go down, that Luton are very likely to join them, and Nottingham Forest are slightly more at risk than Everton. But a probability estimate implies greater precision.
What is particularly relevant for 2024 (and which was absent in 2023) is the assumption that relegation is purely a function of obtaining fewer points than your rivals across 38 fixtures.
Notice how, in the middle of November, Everton’s probability of relegation jumps from less than 5% to 50%(!) That type of sudden movement should cause major doubt about the validity of the underlying model. The reason was because Everton received a 10 point deduction due to breaching Profit and Sustainability Rules (PSR). Everyone knew that Everton had been charged, and there was an expectation that a points deduction may occur. But the magnitude was a surprise, and hadn’t been factored into these probabilities.
Perhaps the deduction could be considered a “black swan” event, which is typically used to excuse us from missing something important. We might say that the relegation tracker is accurate provided nothing unexpected happens. But while a model that is conditional on no external shocks is not a strong criticism against any specific model, it does cast doubt on whether we should rely too much on such models as a whole. Failure to prepare for the unexpected is still a failure of planning.
As an Everton fan I do not care why we might get relegated. A model that tells me the probability conditional on “nothing unexpected” happening is no use. I want to know whether I need to resurrect my coping mechanisms and start planning for the worst.
Everton appealed the severity of the points deduction, and on February 26th received 4 points back. But Everton and Nottingham Forest were both subsequently charged with a new PSR breach — Forest had 4 points deducted and Everton another 2. These were subject to appeal, with the final outcome potentially not being determined until after the season had ended. We are therefore in the ludicrous situation where Everton have up to 4 adjustments to their points tally at the discretion of a panel making a somewhat arbitrary decision.
Therefore, when the BBC report Nielson Griecote’s model that gives Everton a 3% change of relegation, that means little to me.
Probabilistic language is alluring because it gives the illusion of precision. But I contend it would be a big mistake to plan in accordance with most “likely” outcomes. The Premier League’s shambolic imposition of PSR hearings mid-season has introduced significant uncertainty, which should prompt attention to decision making tools that are fit for purpose. Scenario planning is one such example.
On April 6th Everton were at home to Burnley. I was in Chicago for a conference but found a pub showing the game.
Surely we’d win? I didn’t want to think about the odds. I just planned.
- If we won, I’d go back to the conference
- If we didn’t win, I’d go back to the conference
Through a fortuitous GK error we scored and held on to a 1–0 win.
Then, on April 21st Everton played Nottingham Forest in a crucial relegation 6 pointer. Sky Sports used the following graphic:
It suggests Everton had just a 5% chance of relegation, but that is no help to me. If a doctor told me there was a 5% chance I would die from a particular disease, the odds won’t settle my nerves. I need to plan:
- If we beat Nottingham Forest, I will have a double whisky to celebrate.
- If we draw with Nottingham Forest, I will have a single whisky to reflect.
- If we lose to Nottingham Forest, I will go home, pour myself some vodka, and cut up some meat and cheese.
We won!
https://youtu.be/eeztAaWh_tk?si=VQKFCkFfuQL4TVcr
Our next game was the Merseyside Derby. We were at home to our arch rivals Liverpool, who were in contention for the title. We hadn’t won this fixture since 2010 and although we often draw it was fanciful to even hope for anything more. Most Everton fans had written it off.
I spent the day driving from Serbia to Croatia, with a quick stop in Bosnia for coffee. I arrived at my hotel with just enough time to set up at the bar and watch the game on my iPad.
An emphatic and majestic performance of grit and power saw Jarred Branthwaite bundle home a first half opener, and then no.9 Dominic Calvert-Lewin made it 2–0 with a classic Everton goal — a towering header, direct froma corner, into the Gwladys Street net. A famous victory, made all the sweeter for ending the title hopes of our neighbours.
“You lost the league at Goodison Park…”
https://youtu.be/J8pwU2yZZuc?si=fwBjFDz4-hK7ibtO
That win meant that we only needed another 3 points from our remaining 4 fixtures, and next up was Brentford at home. We had confidence and resilience and won 1–0 to guarantee survival.
https://youtu.be/juiFYVkhMRU?si=H9uKD-kczk7wYHHU
Our Friday night away game vs. Luton was no longer a relegation 6 pointer, it was just an opportunity for a perfectly acceptable 1–1 draw. I was teaching in Paris and enjoyed the serenity. I wore the same t-shirt I wore for the Bournemouth match last season, just because. But there was little of importance riding on it.
I had may coping mechanisms, but this year, hurrah, I didn’t need to use them.
This reflection considers techniques to deal with uncertainty, and the general use of scenarios. The essence of scenario planning is to consider an array of potential outcomes, and plan accordingly. But the art and science of scenario construction goes far beyond this. I did not perform the necessary analysis and activities to construct a set of proper scenarios. I did not employ the scenario method. So there is much more to say about scenarios than the example I’m using here.
I hope to have convinced you that the scenario approach is useful, and that the starting position for proper scenario planning is a recognition that trying to guess what’s going to happen is neither necessary nor always helpful. We don’t need to predict the future, we just need to have frameworks that help us cope.
