What does T = 0.7 mean? A betting reading of neutrosophic triplets built from several sources, with anytime error guarantees and its limits, and its refinement when T is part of a refined neutrosophic set

We ask what a neutrosophic triplet means when it summarises several sources. T and F are the winnings of two opposed bettors, I splits into conflict and ignorance, and verdicts keep their error rates under repeated analysis of 33 streptokinase trials. An added section refines T itself, as in the n-valued refined neutrosophic set. Read with the sources as possible worlds, the betting T is a relative truth, true in at least one source; an absolute truth, true in every source, is obtained from the smallest single-source e-value, carries its own error bound, and by construction never coexists with a positive falsity. Truth can also be refined into graded sub-hypotheses, whose degrees are merged into one T with the guarantee preserved by averaging the capitals, not by a t-conorm on the degrees. In simulation, absolute truth is rarely reached with many sources, which makes the existential reading of T measurable. Preprint, version of 15 October 2026. Submitted to the International Journal of Analysis and Applications; not yet peer reviewed. The archive supplementary_code.zip contains the Python code, tests, registrations and result files.