A guessing exercise is more useful when you know what is recorded and when the answer becomes available. PrecogTrainer follows a simple sequence: choose, confirm, reveal.
Before a round is assigned, the server generates its target card and correct position and stores them in the database. The player sees the target to look for and three possible positions. The correct position is withheld from the browser.
You can change your selection before confirming. Once you confirm, the server commits your choice, a timestamp, and the resulting statistics before returning the answer. The recorded result lets you review what you chose without relying on memory.
Random positions and the one-third baseline
The current system selects positions using the platform's cryptographic random-number generator. The underlying .NET method uses a discard-and-retry approach to avoid bias when mapping random values into a range.
This is not a claim that we currently use a separately audited physical or quantum randomness service. We are considering those sources for a future research protocol, where the source, handling, and verification of randomness would need to be documented.
Each position is intended to have the same probability. That does not mean every batch contains exactly equal numbers of Left, Middle, and Right answers. Random samples naturally fluctuate. At 100 rounds, the expected number of correct guesses under independent chance is about 33.3; an actual person will have a whole-number result that can fall above or below that expectation.
We use an internal distribution diagnostic to check position counts and flag invalid values. It does not change the pool to make it look balanced. Equal totals alone would not prove that answers are independent or concealed.
What a percentage tells you
Your percentage correct is the number of correct guesses divided by completed rounds, multiplied by 100. For example, 40 correct guesses in 100 completed rounds gives 40% correct.
That is a useful description of what happened. It is not, by itself, evidence that you receive information from the future. Interpreting a result also requires knowing how many people were compared, how often their scores were checked, and whether the analysis was decided in advance.
The launch leaderboard requires at least 100 completed rounds and shows opted-in players with approved display names. Its design puts rounds played alongside percentage correct because the denominator matters. The threshold is a participation rule, not a scientific significance test.
If many people play, the highest scores will tend to look impressive even when everyone is guessing. A leaderboard answers a competition question: who currently has the highest recorded percentage correct among qualifying players? A study needs its own question and analysis.
Where trust still enters
Answer concealment protects the normal player experience. The server and trusted database operators can access pre-generated answers. These controls do not establish that nobody could know an answer before a choice, or that a timestamp has been independently witnessed.
A future research setup may require stronger answer custody, independently verifiable commitments, and external auditing. Those would be additional safeguards with their own verification requirements.
Our aim is to make the exercise understandable and the record useful. Enjoy a good streak, keep the misses in view, and interpret your score with the same curiosity you brought to the first round.