According to the standard account of time reversal, namely the account found in physics books, a time-reversal transformation involves a temporal operator 𝑇 that, when acting on a sequence of states, inverts the order with which states happen, and suitably changes the properties of the entities in the state so as to make the theory time-reversal invariant. This ‘symmetry first’ approach imposes symmetries on the theory: the changes in the states are a consequence of requiring the theory to be time-reversal invariant. Some (Albert, Callender) find this view unjustified: we discover a theory has a given symmetry, on the basis of the theory’s ontology, not the other way around. So, they propose a ‘metaphysics first’ approach, sometimes dubbed ‘pancake account’ of time reversal: 𝑇 inverts the order of the states but does nothing else. Consequently, since there are no obvious independent reasons for the state to change as 𝑇 prescribes to preserve time-reversal symmetry, then the theory is not time-reversal invariant. In this chapter I wish to further motivate the pancake account of time reversal by arguing the standard account is far more problematical than has been suggested. Moreover, I defend the pancake account from recent objections raised by Roberts. Finally, since I value symmetries, I propose an alternative account, which aims at retaining the best of both approaches: the 𝑇 operator changes the order of the states, it leaves the state unaffected (like the pancake account), but also makes the theory time-reversal invariant (like the standard account).