Does intention shift a quantum random stream?
Every signed-in game is committed before play begins. These totals combine every committed stream and compare them with the 50/50 outcome expected by chance.
How the game worksWhat the data shows
Each bit and ball landing is compared with the direction selected before it appeared. Abandoned and interrupted runs stay in these totals because the stream was committed before play. Chance predicts a 50/50 split.
Bits aligned with intention
49.98%3,914,635 of 7,832,600 · chance expects 3,916,300
Balls landing on target
49.83%11,612 of 23,303 · chance expects 11,651.5
Committed runs
6,111
Bits analyzed
7,832,600
Effect per 1,000 bits
-0.21
Bit-level odds
1 in 1
Ball landing odds
1 in 1
The combined bit stream is within normal random variation.
Combined bit-level z-score: -1.19 (one-tailed p = 0.8829). Ball landings give z = -0.52 (1 in 1).
Experiment coverage
Participants
204
All committed runs
6,111
Explicitly abandoned
128
Started, not completed
57
A run is committed before play starts, including the full bit stream and targets. Outcome statistics use every committed run — abandoned and interrupted streams stay in the totals so quitting cannot remove a bad result.
Average image unfuzzed
51.4%
Personal details stay private
This page only uses anonymous totals. Result notes and tags are never included, and only their owner can view or change them.
How this fits wider PK research
Laboratory micro-PK research has tested questions much like this game: can intention shift a physical random number generator away from its 50/50 expectation? The published effects are generally tiny, and the evidence is disputed. To make the scale concrete, the studies below show the reported rate, the lift in percentage points (pp) above 50%, and the amount of data collected.
Early quantum RNG experiments
Helmut Schmidt (1971)
- Observed rate
- 50.9% / 52.4%
- Above 50% chance
- +0.9 / +2.4 pp
- Scale
- 32,768 / 12,800 binary outcomes
Two formal experiments reported above-chance results after screening for promising participants. The first used 15 selected participants; the second used two especially promising performers, so these rates should not be treated as population averages.
The PEAR benchmark experiment
Jahn et al. (1997), Journal of Scientific Exploration
- Observed rate
- ≈50.013% in the high condition
- Above 50% chance
- ≈+0.013 pp
- Scale
- 2,497,200 trials overall
Across 12 years and 91 operators, PEAR reported a significant separation between high- and low-intention conditions. Each trial contained 200 bits; the displayed rate is the high-intention mean from 839,800 trials, while the overall count also includes low and baseline trials.
RNG research meta-analysis
Radin & Nelson (1989), Foundations of Physics
- Observed rate
- 50.018%
- Above 50% chance
- +0.018 pp
- Scale
- 597 experimental studies
The quality-weighted result combined many different RNG protocols and reported a very small but statistically significant shift. The source reports studies, not one comparable total of binary trials.
Large quantum RNG test
Maier, Dechamps & Pflitsch (2018), Frontiers in Psychology
- Observed rate
- 50.02%
- Above 50% chance
- +0.02 pp
- Scale
- 1,257,100 binary selections
Each of 12,571 participants observed 100 quantum-selected outcomes after an optimism exercise. Despite the slight numerical lift, the Bayesian analysis favored chance (BF₀₁ = 10.07). This tested passive observation rather than the game’s explicit intention task.
Important context: a three-laboratory replication in 2000 collected 750,000 experimental trials but did not reproduce PEAR's pre-specified high-versus-low result (Z = 0.59, p = 0.278). A 2006 meta-analysis of 380 studies found a significant but very small and highly variable effect, and concluded that publication bias could explain it. Other researchers dispute that interpretation. Trial counts are not directly interchangeable: some papers count individual bits, some count 200-bit blocks, and meta-analyses count studies.