Skip to main content
def AgentsComparatorWorker.transform(self, data, agents=tuple(AGENTS)) [source]
Re-evaluate the full comparison on a new dataset. Runs the complete multi-agent scoring pipeline on any data sample. Use this to:
  • Monitor agent rankings on subsequent OOT windows.
  • Compare a subset of agents (e.g., the top 2 from the initial ranking).
  • Generate a fresh scoring report for a different segment or time period.
  • Perform “What-If” analysis: does the ranking hold on a different sample?

​
Parameters

data : pDF
The dataset for performance evaluation (must contain
target + features).
 agents : tuple
Subset of agents to include (e.g., (1, 2) for
a head-to-head comparison, or (1, 3, 4) for
multiple agents).

​
Returns

AgentsComparatorResult
Full scoring package (overview, scoring, lift,
calibration) for the requested agents.
No State Change: This method does not modify the fitted state. It produces a fresh analysis on the new data. The original fit results remain accessible via .result and .stats.Head-to-Head: Use agents=(i, j) for a focused comparison between two specific agents.