We prove a new tableaux formula for the symmetric Macdonald polynomials
$P_{\lambda }(X;q,t)$ that has considerably fewer terms and simpler weights than previously existing formulas. Our formula is a sum over certain sorted non-attacking tableaux, weighted by the queue inversion statistic
$\operatorname {\mathrm {\texttt {quinv}}}$. The
$\operatorname {\mathrm {\texttt {quinv}}}$ statistic originates from a formula for the modified Macdonald polynomials
$\widetilde {H}_{\lambda }(X;q,t)$ due to Ayyer, Martin, and the author (2022), and is naturally related to the dynamics of the asymmetric simple exclusion process (ASEP) on a circle.
We prove our results by introducing probabilistic operators that act on non-attacking tableaux to generate a set of tableaux whose weighted sum equals
$P_{\lambda }(X;q,t)$. These operators are a modification of the inversion flip operators of Loehr and Niese (2012), which yield an involution on tableaux that preserves the major index statistic but fails to preserve the non-attacking condition. Our tableaux are in bijection with the multiline queues introduced by Martin (2020), allowing us to derive an alternative multiline queue formula for
$P_{\lambda }(X;q,t)$. Finally, our formula recovers an alternative formula for the Jack polynomials
$J_{\lambda }(X;\alpha )$ due to Knop and Sahi (1996) using the same queue inversion statistic.