Other systems (or finite table limits) "can" do it if you're lucky. But over time, the bankroll function always decreases and asymptotes to a pure loss of slope -EV.
With martingale, like I said, the bankroll is a diverging, oscillating function that grows with magnitude over time. If you keep playing, you'll always go positive and negative at various times. Pick a positive time, which, per QFIT's example, will always arrive, and you're a winner.
Your equation describes the expectation at any specific number of hands (or total wager), but because the function is so unstable, it's a poor predictor.
What's the sum of the series 1-2+3-4...? But what are the values you'll actually see? It all depends on when you decide to stop.