With persistent practice of intuitive prediction, one will eventually find oneself aware of certain future events.

Multi-center trial with approximately 1200 participants randomly assigned to 12 weeks of intuitive prediction practice or equal-time perceptual training; each participant completes at least 1000 two-alternative predictions; after answers are submitted and time-stamped, random targets are generated by an independent server, with locked data, stopping rules, and analysis code. The primary test examines accuracy change in the training group relative to the control and whether the final hold-out phase exceeds 50%, with participant and center as clustering units. Covering Supplementary Question 1: approximately 300 of these participants undergo multiple laboratory sessions with randomized high/low arousal, different valences, and varying personal relevance induction, with simultaneous skin conductance, heart rate, and respiration recording; pre-registered main effects of arousal, inverted-U relationships, and training interactions, controlling for previous trial outcomes and sequential expectancies. Total duration 6–9 months, with an independent confirmation sample.

Does Repeated Practice Improve Prediction of Future Random Outcomes? A Multicenter Pre-registered Trial

If, after excluding leakage and generator bias, stable accurate prediction of randomly generated targets after response is obtained, this would challenge the assumption that "given all prior information, the answer is conditionally independent of independent future random targets."

Zoltan Kekecs，罗兰大学：Transparent Psi Project; Stefan Schmidt，弗赖堡大学及 Institute for Frontier Areas of Psychology and Mental Health; Patrizio Tressoldi，帕多瓦大学; Fetzer Franklin Fund of the John E. Fetzer Memorial Trust：相关重复研究的既往资助方

It offers a directional prediction of improvement with practice, but without a minimum effect size, training dose, or clear deadline; "eventually" and "certain events" allow indefinite postponement or post hoc selection, requiring prior tightening.

"Contested" does not mean the evidence is evenly balanced: strict replication studies have not stably reproduced the effect; emotional arousal may alter confidence and accuracy separately, both must be modeled separately, and ordinary cue-based prediction cannot prove precognition.