College football expected points model

Seven-class next-score model. Column i of the softmax output is cfb_model_manifest$ep_class_contract$class_order[i], worth cfb_model_manifest$ep_class_contract$point_values[i] points; expected points is the probability-weighted sum.

Usage

cfb_ep_model

Format

A raw vector representation of an 'xgboost' model (multi:softprob, 7 classes). Features, in order: TimeSecsRem, yards_to_goal, distance, down_1, down_2, down_3, down_4, pos_score_diff_start.

Source

The cfb_model_artifacts release of https://github.com/sportsdataverse/sportsdataverse-data, built by the sportsdataverse-py model pipeline.

Examples

model <- xgboost::xgb.load.raw(cfb_ep_model)
xgboost::getinfo(model, "feature_name")
#> [1] "TimeSecsRem"          "yards_to_goal"        "distance"            
#> [4] "down_1"               "down_2"               "down_3"              
#> [7] "down_4"               "pos_score_diff_start"