From Model Training Pack to Predictions: How to Use Your Model

Since launching the Model Training Pack, one of the top questions I've heard is:

“I’ve got the trained model… now what?”

If that’s you, this guide is for you.
We’ll walk through:

• What kind of data your model needs to make predictions.
• Where to get that data from the CollegeFootballData API.
• How to load different types of models from the pack and run predictions.
• How to skip the data prep entirely with Tier 3 weekly CSV drops.

1. What Your Model Needs

The models in the training pack were all built on feature-ready CSV files.

That means:

• The CSV has the exact same columns as the training data.
• The columns are in the same order.
• The numbers are calculated the same way (e.g., using stats from games before the game you’re trying to predict).
• If your CSV doesn’t match, your model will throw errors or give bad predictions.

2. Two Ways to Get the Data

Option 1: Build it yourself

You can pull comparable data from the CollegeFootballData API.
Here are the endpoints you’d use, at a high level:

Note: If you build your own CSV, you’ll need to join these datasets together and make sure your stats only include games before the prediction week.

Option 2: Skip the work

Starting Week 5, Tier 3 patrons will get a weekly CSV that already:

• Has all the right columns.
• Is in the correct order.
• Uses stats from games before that week.

With that file, you can go straight to loading your model and running predictions.

3. Loading and Predicting

Once you have your CSV, here’s how to use it with each type of model from the pack.
Replace "week5_features.csv" with your file and "path/to/model" with your model file.

Random Forest / Regression (scikit-learn)

import pandas as pd, joblib

# Load your features
X_live = pd.read_csv("week5_features.csv")

# Load your model
model = joblib.load("models/sklearn_rf.pkl")

# Make predictions
preds = model.predict(X_live)
X_live['prediction'] = preds

XGBoost

import pandas as pd, xgboost as xgb

# Load your features
X_live = pd.read_csv("week5_features.csv")

# Load your model
model = joblib.load("models/xgb_model.pkl")

# Make predictions
preds = model.predict_proba(X_live)[:, 1]
X_live['prediction'] = preds

fastai (tabular)

import pandas as pd
from fastai.tabular.all import load_learner

# Load your features
cat_features = [...] # list out categorical features
cont_features = [...] # list out continuous features

X_live = pd.read_csv("week5_features.csv")
X_live = X_live[cat_features + cont_features]


# Load your model
learn = load_learner("models/fastai_model.pkl")
dls = learn.dls.test_dl(X_live)

# Make predictions
batch_preds = learn.get_preds(dl=dls)[0].numpy()
X_live['prediction'] = batch_preds

4. Common Gotchas

Wrong column order → reorder to match your training data before predicting.

Missing columns → make sure your CSV includes everything from training.

Wrong data types → convert strings to numbers where needed.

fastai category mismatch → your categories must match what the model was trained on.

5. The Fast Lane

If you want to:

• Avoid merging multiple datasets,
• Skip figuring out lag logic, and
• Be sure your columns match perfectly…

…join Tier 3 on Patreon.
Every week starting in Week 5, you’ll get a CSV that’s ready to feed directly into your model.

Join Tier 3 here →

6. Your Next Steps

1. Pick one of your models from the pack.
2. Grab a CSV, either your own or from the pack, and run the code above to test.
3. Get your hands on current-season features (DIY or Tier 3) and start making real predictions.

Bottom line:
If you can build the CSV yourself, great. You now know exactly what your model needs.
If you want to skip the grunt work and start predicting in minutes, Tier 3’s weekly CSV drops are your fastest path.