The perfect example is MN from 2016. Trump didn't fight there late and lost the state by 45k. He had rallies in WI and only won by 23k. His biggest buy was in VA, which he lost by 5 points (Clinton had already pulled ads because she had it locked up). He
completely pulled ads from Rust Belt states in September, and only returned to specific targeted ones later; good write-up of the strategy
here from late September 2016.
For all the memes about 4D underwater chess, Trump
does get things wrong. He's used to that in business, where he writes off losses, and in politics, where he pivots to a new focus or replaces personnel. As long as future successes outweigh the failures he comes out ahead. But he can't afford to get too many things wrong during an election; there's nothing to salvage if he loses the one big event.
You can spin 2016 as a perfectly calibrated, cost-efficient campaign that did the minimum he needed to win, and praise its efficiency. It was a legitimate accomplishment. But along with the good strategy came a series of small mistakes, which were offset by his opponent's larger mistakes. That's fine,
if you win, but relying on an opponent's screw-ups is still a risky strategy.
