Podos v1 Baseline

Podos is a small baseline transformer model for soccer match prediction.

Model Details

Podos predicts soccer match outcomes based on 23 input parameters including sportsbook odds, recent team performance, win/loss streak, and more.

For direct use, download the source pytorch class, label_encoder (optional), and load the model.

PodosTransformer.from_pretrained("Bettensor/podos_soccer_model")

The label encoder contains the id mappings to all teams the model was trained on. Ensure you have Torch installed with:

pip install torch

scikit-learn version 1.4.2 if you want to use the label_encoder:

pip install scikit-learn==1.4.2

newer versions of sklearn may work but are untested.

You also need HuggingFace_hub and safetensors, install with:

pip install huggingface_hub

pip install safetensors

model expects 23 parameters for input, with team names mapped as ids:

The label_encoder currently contains mappings for 569 unique teams

Model is available to use with Bettensor at https://github.com/Bettensor/bettensor

podos v1 presents some home team bias, and may provide overconfident scores to its predicted outcome.

Model was trained on 100,000 games with 569 individual teams.

qucat | Nickel5