The divergence flag identifies markets where the model’s assessment and the market price disagree by an exceptional margin. It is the final checkpoint before the expected value calculation: once the model has produced its probabilities and blended them with historical rates, it compares each market’s view against what the odds imply and flags any gap that stands out.
A big disagreement can mean one of two things. The model may have found something the market is slow to price, which is the kind of opportunity the platform exists to surface. Or the market may know something the model cannot see in the data: late team news, a motivational situation, money moving for reasons outside the numbers. The flag exists because these two cases look identical in the output, and only context tells them apart.
Treat a flagged market as an instruction to look closer rather than a stronger recommendation. Before acting on one, spend thirty seconds checking for an external explanation: confirmed line-ups, key absences, cup rotation risk, a dead-rubber fixture. If nothing external explains the gap, the flag is doing its job and pointing you at a price the model believes is meaningfully wrong. If something external does explain it, you have dodged a bet the numbers alone could not warn you about.
Very large positive EV figures tend to arrive with a divergence flag attached, for this reason. The bigger the claimed edge, the more those thirty seconds are worth.