Wizard or NPC

  1. Press Space to start the 10 second counter.
  2. Think magically and press Space again to “mark” the time of your intent.
  3. Once the counter finishes, you'll get a Wizard Likelihood score. Click on subscores for details.
  4. ???
RNG: 
 Focus: 

Run status
Status
Idle
Elapsed
0.000 s
Samples
0
Rate
0 Hz
Mark time
—
Window
±100ms
Random stream
—
Wizard likelihood (this run)
0%
NPC
0% (NPC) 100% (definitely a wizard)
No scored runs yet.
Latest result

No runs yet.

What just happened

So, wizard or NPC? If you can nudge a random number generator (RNG) in a statistically suspicious way, you might at least qualify as a High-Functioning NPC.

The idea draws from Dean Radin’s work on micro-psychokinesis and projects like the Global Consciousness Project.

It’s straightforward: the app pulls a stream of random numbers for about 10 seconds. Pressing Space doesn’t touch the RNG—it simply tags the moment you focus your intent. Afterward, we check the numbers in a narrow window around that tag against the rest of the stream and score how unusual the difference is.

A single strong “wizard” result isn’t proof that you bent reality. But, rack up a few in a row?

...

Spooky.

FAQ

Sure, the keypress event listener isn't logically connected to the RNG... but it's not strictly independent of the RNG, right?
Let's just pretend that it is.
I've spotted some errors in your methodology.
I bet there are several you haven't yet spotted.
I just learned I'm a wizard, what do I do?
Powers for good only, please!
I'm definitely a wizard but I'm getting low scores. What gives?
Probably dark forces interfering. Be watchful.
This test is messed up! It says I'm an NPC?!
Right.
NPC meme
History (persisted in localStorage)
Ended Samples Mark (ms) nWindow Δmean p(mean) Δp(x≄0.5) p(prop) p(any) adj
No history yet.
Method & notes
  • Press Space to start a 10s high-frequency RNG capture. Press Space again during the run to record a timestamp. After 10s, we compare the ±100ms window around the timestamp to the rest of the run.
  • Important: the spacebar has no logical effect on RNG generation; it only records a timestamp for later analysis — i.e., we’re only “measuring” your focused psychokinetic powers at the moment of the spacebar press.
  • RNG values are sampled from crypto.getRandomValues when available (fallback: Math.random).
  • Mean test: a normal approximation using the standard error sqrt(var1/n1 + var2/n2) (Welch-style).
  • Proportion test: compares P(x ≄ 0.5) between window and rest using a two-proportion z-test.
  • Extra anomaly checks compare window vs rest on variability (Brown–Forsythe/Levene), distribution shape (KS), binned histograms (χÂČ / G), tail rates, and simple dependence metrics (lag-1 correlation, |Δ|, runs).
  • For several checks, we also compute an empirical p by sampling hundreds of random “fake marks” in the same run and asking: “how often is a random window at least this weird?”
  • Inspired by and loosely based on RNG/PK-style experiments described by Dean Radin (this tool is for fun and is not a scientific instrument).
  • This is a toy; browser scheduling, key event latency, and timer throttling can affect the effective sampling cadence. We analyze using each sample’s recorded timestamp.