A stronger CTR model can beat a fresher user intent signal when the fresh signal is noisy, sparse, or unstable. This happens because long-term click patterns smooth out randomness, while recency alone may overreact to short-term behavior spikes. When auction bids are equal, the system trusts calibrated CTR predictions more if historical data is rich and well-trained. However, fresher intent wins when behavior shifts quickly or context changes suddenly, making old patterns less reliable.