Four scores, four questions
A notice can show up to four scores. They do not agree because each one counts different facts. None of them uses AI, and none is a probability of winning.
What each one is
Quick fit (inbox, 0 to 100). Points earned over points possible across up to six checks, scaled to 100; a set-aside or vehicle roster you lack caps it at 35. It skips the bid gates. Use it to sort a long list.
Fit (Bid / No-Bid Cockpit, 0 to 100). Eight weighted criteria: a fit earns full weight, a check half, a risk none; a set-aside you lack or a closed response date forces Pass. You can override any row. Use it to decide whether to pursue.
Match (0 to 100). Five parts of your profile against the notice, weighted 30, 20, 20, 15, 15, shown only when at least 60 points of facts are known.
Feasibility (GO, REVIEW or NO-GO, 0 to 100). Five dimensions of up to 25 points each from your profile, summed; GO at 70, REVIEW at 40; a NAICS code not in your profile holds it at REVIEW. A set-aside you do not hold makes it NO-GO. GO also needs at least 3 of the five dimensions above their starting value. A label alone does not mean you should bid.
P(win) is separate
P(win) is an Estimate: the number you typed, or a default for the pipeline stage (10 percent for prospecting, 20 for qualified, 35 for proposal, 40 for submitted). It does not use any of the four scores.
Which to trust
Sort with Quick fit, then read the Fit for the decision. Open the reasons under each number. When two scores disagree, compare the reasons, not the numbers.