FPL Quant

Quantitative FPL predictor. Treating FPL like the stock market.
Hand-built by a specialist in Computational Modelling.
Not a machine learning model!

[FPL Quant @QuantFPL]

Thinking of creating your own FPL model?
I spent a summer building one. Nine seasons of data, a component-wise expected-points model, and a portfolio optimiser that treats the crowd as a benchmark.
Here's everything I got wrong first, so you don't have to. 🧵

[FPL Quant @QuantFPL]

@OfficialFPL Best Captain Choices for #GW9:

[FPL Quant @QuantFPL]

GW9 Wildcard: Completely off my model's predictions
#WC #WC9 #Wildcard #GW9 #FPL

[FPL Quant @QuantFPL]

GW9 Wildcard:
Risk-adjusted against the top 10k
This is at 60% view weight, the optimal blend that represents the top 10k while taking calculated risk against it. #FPL #GW9 #WC #WC9

[FPL Quant @QuantFPL]

7. Metrics of Predictions so far:
The model is very accurate, with finding under-represented assests

[FPL Quant @QuantFPL]

Differentials Thread: 🧵🔽
#GW9 #Differential #FPL

[FPL Quant @QuantFPL]

6. Differential XI

[FPL Quant @QuantFPL]

There's a mountain of accounts with their own personal models.
I'm calling out to those creators to reply to this post with their own metrics. Reply with these:
RMSE and MAE
RMSE of a flat 2.7 for every player, same rows
Avg Spearman rank correlation per GW.
Please share around

[FPL Quant @QuantFPL]

8. Aerial Ability vs Set Piece Threat. #FPL #GW9
#Muharemovic #Ajer #Silva

[FPL Quant @QuantFPL]

GW9 Statpack: A Thread🧵👇

[FPL Quant @QuantFPL]

7. Goalkeeper Workload vs Shot Stopping.
#Raya or #Trafford are really the best two options to consider. Two other shouts are:
#Kinsky, who is living off a strong defence in front of him. high xCS
#Hornicek, the other end of the spectrum: leaky defence, but high numbers of saves