Various techniques such as spacing, retrieval practice, and the use of mnemonic strategies have each been shown to be effective at optimising learning under laboratory conditions, typically when employed on their own. How well do these techniques work in the real world, and what happens when several techniques are combined? Learners acquiring the vocabulary of a foreign language have a limited time budget and hence effective learning methods must adopt an optimal trade-off between the value of a technique and the time taken to implement it. We conducted a two-stage competition in which international research teams developed methods to maximize the number of correct translations that learners could acquire in 1 h and successfully recall 1 week later. In Stage 1, the teams tested their method for learning Lithuanian-English words pairs against a standardized control method. Five shortlisted methods and the control condition were then compared on a common online platform in Stage 2, using Lakota-English pairs, with retention data collected from over 6000 users of an online learning tool. After exclusions, the final sample was 3803. Retention declined with age and was higher in females, and participants showed some metacognitive insight into the durability of their learning. The winning entry, which combined a visual mnemonic technique with retrieval practice and an adaptive algorithm for introducing new words, achieved an average of 27.23, 95% CI [26.08, 28.38] word pairs recalled. This work highlights the importance of retrieval practice in real-world learning.