Nice. What about publishing all the results, not just the 10 where the null hypothesis was (correctly or not) rejected? If we publish all 50 experiments, then about 4-5 of them would be erroneous due to statistical chance, which is the 5% error rate we were willing to accept.