This study evaluates whether low-complexity algorithms can meaningfully enhance returns for retail investors who trade with balances below $5,000. Using split-adjusted daily prices for three broad U.S. ETFs and seven liquid equities, we back-test three allocation rules over the 2024 calendar year: (i) a naïve 1/N portfolio rebalanced monthly, (ii) an 80/20 momentum rule that rotates one-fifth of capital into six-month winners, and (iii) a continuous-action Soft Actor–Critic (SAC) reinforcement-learning agent trained on 2020–2023 data. All trades incur a realistic 5-bp round-trip cost and pass a 10-bp stress test. Relative to the naïve benchmark, the SAC strategy raises cumulative return by ∼45 percentage points, more than doubles the Sharpe ratio, and maintains its edge when costs quintuple. The momentum rule delivers intermediate gains but proves more cost-sensitive. Bootstrap resampling and comparison with 1,000 random buy-and-hold portfolios confirm statistical significance at the 1% level. These results demonstrate that transparent heuristics and lightweight reinforcement learning can significantly narrow the performance gap faced by very small accounts, thereby strengthening the case for fintech platforms that combine full fee disclosure with basic algorithmic guidance to promote broader financial inclusion.