Burnley’s Conditional Relegation Probabilities, May 8th

Ok, so this is imperfect, but in an attempt to preemptively quantify what today’s results mean for Burnley, I…reverse engineered FiveThirtyEight’s SPI? Not quite that, actually. First, Burnley’s estimated relegation probabilities with each pair of results today from Everton and Leeds. After that, I’ll explain how we got there.

Everton result todayLeeds result todayBurnley Relegation Probability
WinWin75%
DrawWin64%
WinDraw61%
LossWin52%
DrawDraw46%
WinLoss42%
LossDraw36%
DrawLoss34%
LossLoss29%

What I did here is take the SPI game-by-game forecasts from the EPL season so far and fit a second-order polynomial equation to them, then adjust home-field advantage to minimize error. The equation ended up being Win Probability = .00009x2 + .0128x +.3907, where x is the difference between the SPI of each team, pulled by six towards the home side. After using that, the probability of a draw was just whatever’s left. It’s imperfect—it’s off by an average of 5% on each probability, compared to the actual FiveThirtyEight model—but I lack the mental resources to full-on reverse engineer a model of this complexity or create one myself.

After doing this, I simulated Burnley, Everton, and Leeds’s remaining games 10,000 times. I did not run the simulations “hot”—this is another difference from FiveThirtyEight, their model accounts for the way each team’s SPI will change depending on future results—but even with this and the 5% error, I came out with relegation probabilities of 34% for Burnley, 27% for Everton, and 39% for Leeds, which is comparable to what FiveThirtyEight’s process turns out. To adjust for them having Burnley at 38%, relative to my 34%, I gave a blanket multiplier to each eventual conditional relegation probability, getting us to what’s listed above.

NIT fan. Joe Kelly expert. Milk drinker. Can be found on Twitter (@nit_stu) and Instagram (@nitstu32).
Posts created 3821

Leave a Reply

Your email address will not be published. Required fields are marked *

Related Posts

Begin typing your search term above and press enter to search. Press ESC to cancel.