File size: 2,765 Bytes
c2a0c37 1c076e7 c2a0c37 5df422c 1c076e7 8381201 8391b86 1c076e7 bebdb5a 1c076e7 cfd7fb9 1c076e7 174c253 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 |
---
# For reference on model card metadata, see the spec: https://github.com/huggingface/hub-docs/blob/main/modelcard.md?plain=1
# Doc / guide: https://huggingface.co/docs/hub/model-cards
{}
---
# Podos v1 Baseline
Podos is a small baseline transformer model for soccer match prediction.
## Model Details
### Model Description
- **Developed by:** Bettensor | Nickel5
- **Model type:** PyTorch Transformer
- **Parameters** 276K parameters
## Uses
Podos predicts soccer match outcomes based on 23 input parameters including sportsbook odds, recent team performance, win/loss streak, and more.
### Direct Use
For direct use, download the source pytorch class, label_encoder (optional), and load the model. <p><code>PodosTransformer.from_pretrained("Bettensor/podos_soccer_model")</code></p>
The label encoder contains the id mappings to all teams the model was trained on.
Ensure you have Torch installed with:
<p><code>pip install torch</code></p>
scikit-learn version 1.4.2 if you want to use the label_encoder:
<p><code>pip install scikit-learn==1.4.2</code></p>
newer versions of sklearn may work but are untested.
You also need HuggingFace_hub and safetensors, install with:
<p><code>pip install huggingface_hub</code></p>
<p><code>pip install safetensors</code></p>
model expects 23 parameters for input, with team names mapped as ids:
- HS - Home shots
- AS - Away shots
- HST - Home shots on target
- AST - Away shots on target
- HC - Home corners
- AC - Away corners
- HO - Home offsides
- AO - Away offsides
- HY - Home yellow card
- AY - Away yellow cards
- HR - Home red cards
- AR - Away red cards
- oddsH - Home win odds
- oddsD - Draw odds
- oddsA - Away win odds
- home_encoded - Home team id
- away_encoded - Away team id
- WinStreakHome - Home win streak
- LossStreakHome - home loss streak
- WinStreakAway - Away win streak
- LossStreakAway - Away loss streak
- HomeTeamForm - Home team recent performance
- AwayTeamForm - Away team recent performance
The label_encoder currently contains mappings for 569 unique teams
### Downstream Use
Model is available to use with Bettensor at https://github.com/Bettensor/bettensor
## Bias, Risks, and Limitations
podos v1 presents some home team bias, and may provide overconfident scores to its predicted outcome.
### Recommendations/Future work
- reduce bias by encoding home field advantage
- more teams and leagues, especially with more rigorous performance metrics
- Additional layers for larger input size
- team embedding layers
- individual player performance
### Training Data
Model was trained on 100,000 games with 569 individual teams.
- data source: https://www.football-data.co.uk/downloadm.php
## Model Card Authors
qucat | Nickel5
## Model Card Contact
www.nickel5.com |