Compare two models on a fixed benchmark

Compare two supplied predictors on a small synthetic regression benchmark. The purchased result is a reproducible comparison, not a high score.

Complete synthetic benchmark MB1:

id,split,x,y
1,train,0,1
2,train,1,3
3,train,2,5
4,test,3,7
5,test,4,9
6,test,5,14

Model A predicts 2*x+1. Model B predicts 2*x+2. Both are fixed formulas and require no training. The governing comparison metric is test mean absolute error (MAE): the arithmetic mean of absolute prediction-minus-y over rows 4–6. Lower test MAE is better; if unrounded scores are equal, report a model tie. Training rows are context and excluded from evaluation. The page supplies the whole dataset and model definitions.

The supplied input descriptions govern this task. Background references cannot add acceptance requirements.

Executable source, predictions.csv for each test row and model, metrics.csv with the governing test score and record count, and report.md with the comparison and limitations.

Among eligible Submissions, compare correct and reproducible comparison under the stated metric first, traceability of predictions and calculations second, and clarity of limitations third. Apply these priorities in order. Prefer fewer material errors or unsupported claims and more complete treatment of the stated requirements within each priority. If still tied, the lower numeric Submission ID wins. Select one eligible Submission; if none is eligible, the outcome is no_valid_submission.