How to Read Group Analytics

A row is one completed group_target_selections record inside the selected window. For psychokinesis, the success flag is the challenge’s recorded match. For telesthesia, the flag is the consensus result when one was stored, and otherwise the challenge’s recorded match. Dice trials are checked again by comparing the manifested faces with the target: one die must match that face, and two dice match if the two faces are the same pair in either order. The same chance baselines as personal analytics are used when a p-value is shown: Zener 20%, one die 16.67%, two dice 2.78%, coin 50%, 13-card set 7.69%, full deck 1.92%.

Totals sum challenge counts across the groups in your membership. The rate is 100 × successful challenges / challenges. A group with more trials weighs more than a group with a few. The chart beside the percentage is that same split of hits and misses.

A dice target that contains a comma is counted as two dice. A single face is one die. Each bar’s p-value, when present, is a one-sided upper binomial probability against that type’s chance rate. It asks whether the group’s hit count is higher than chance. It is computed from the number of challenges, not the number of members.

Psychokinesis rows come from members who submitted a selection on a psychokinesis trial. Telesthesia rows come from validated guesses, the ones kept after consensus. Average participants is the mean number of distinct members recorded on the challenges in that mode.

The difference between the modes is the task, not a second scoring trick. Influence asks the group to shift what is generated. Sensing asks the group to name what was already set and hidden. Personal practice of the same split is described in Challenges Help.

Headcount for a challenge is the number of distinct members in the result rows for that challenge. The success and miss averages are means of those headcounts, split by the group match flag. A higher headcount on hits than on misses is a pattern in this window. It is not a test that adding people caused the hits. The page used to state that larger groups succeed more often. That sentence overreached the count, and the dashboard now points here instead.

Each telesthesia challenge in the window is one attempt. A success is the same group match flag described above: the stored consensus match when it exists. The breakdown table repeats that count by challenge type. Psychokinesis trials are left out of this panel. Inside a single challenge-type row elsewhere, psychokinesis “consensus” fields reuse the manifestation match, because there is no separate hidden-target vote. The panel titled Consensus Accuracy does not include those rows.

Daily points use the group match flag, grouped by date, challenge type, and mode. Hourly points extract the hour of created_at after converting from UTC to your saved timezone, and they count distinct challenges. They are not per-member guesses.

Individual correctness compares that member’s submitted value with the target. Two-dice comparisons ignore order. The member p-value is a one-sided binomial test against a baseline weighted by the challenge types they actually played. “Significant” means that p-value is below 0.05. The expanded row shows the same comparison per challenge and mode, each with its own expected rate. Admins and members see this breakdown on the analytics page. During a live trial, admins see fuller participation detail and other members see a simpler selections view, as described in Groups Help.

There is no pooled test that “the group is above chance” across mixed challenge types, because the baselines differ. Per-type p-values use that type’s baseline and the challenge count. They are exploratory descriptions of the recorded trials, not a demonstration of a cause. For the personal chance model, use the baseline list and Methodology & Evidence.