Skinwalker Ranch by the numbers: five short tests of five claims

The question

Skinwalker Ranch comes with big claims: a basin with a busier sky, a legend centuries deep, a run of experiments, a time warp that upsets GPS, a "wormhole" in a laser scan. This page takes five, one at a time, and asks a narrow question of each. None settles the subject.

The tests were run by an AI model (Claude) on public data, at Paul L. G. Harris's direction. Claude cannot watch television, so what is said about the programme comes from written recaps, network pages, forum posts and, for two seasons, caption text. The team has released no raw data.

The test

Each study had a plan saved before the data were collected or the code was run, fixing the question, method and reading rule. Where possible there was a control: matched counties, Hessdalen in Norway, or known answers the code had to reproduce. The lidar plan was written knowing the sceptics' answer, and the ledger plan declared prior knowledge of the 1996 article.

What came out

Do people near the ranch report more sightings?

Uintah and Duchesne counties produced 25 reports to the National UFO Reporting Center in 1995 to 2019, about one a year: 0.211 per 10,000 residents per year, against 0.258 for thirty rural counties chosen by a fixed rule before their figures were seen. The rate ratio is 0.83 (95% interval 0.34 to 2.02), and the counties rank 13th and 18th of 31. No clear rise followed the 1996 article, the 2005 book or the 2020 series. None of 32 summaries mentions the ranch, nor do the 18 full reports that could be read.

How much of the legend was in the 1996 press?

Fifteen story elements were traced in date order through 116 ledger rows. One is fully there in 1996, the cattle. Three appear in plainer form: lights, mid-air "doorways" and a line on earlier owners. Eleven are absent: seven first appear in George Knapp's articles of November 2002 (including the wolf and the skinwalker) and four in 2020 or later. Cattle rose from seven in June 1996 to thirteen by October (at second hand) and fourteen in 2002. At Hessdalen in Norway, where instruments logged lights, 22% of later rows grew, against 49% for the ranch. A late first appearance does not prove invention, and the rule cut both ways: the earlier owners' family denial reached print sixteen years after the sale.

How checkable are the show's experiments?

Twenty-four episodes gave 75 experiments, scored from recaps. A prediction was stated in 6 (8%), a control run in 14 (19%), the instrument named in 8 (11%), and an ordinary cause raised in 25 (33%) and tested in 12 (16%). No raw data release was found. These are floors, because most answers were "not stated". Scored from captions, Seasons 3 and 4 reach 38% for predictions and 71% for ordinary causes raised. The Curse of Oak Island scores about the same.

Credit is due. The Doppler on Wheels team (Karen Kosiba and Josh Wurman) fired a rocket as a known target first, and Travis Taylor switched a Tesla coil off to test it as a signal source. No paper or data release by the team was found. Brandon Fugal wrote on 10 August 2025 that his team had not yet submitted papers for peer review.

Could a time warp explain the GPS faults?

The code first reproduced eight known answers, including the GPS clock offset (38.58 microseconds a day). A receiver solves for its own clock error at every fix, so a slow local clock moves no position. Delaying signals enough to shift a fix by 10 metres needs about 3.4 × 10^27 kg, or 1.77 Jupiters. The mildest of nine cases adds about 900 million g. A distortion small enough to escape a gravimeter is about 10^17 times too weak.

Does a hovering drone leave a hole in a lidar scan?

Season 4, episode 11 (10 July 2023) showed a round black gap in a drone scan. Forum posters reported it was called a wormhole and identified a DJI Zenmuse L1 scanner at 35.4 feet (10.8 metres), which the team has not confirmed. Rebuilt from the maker's figures, a drone that hovers and turns with the scanner tilted below level leaves a round unscanned patch beneath it. At 10.8 metres it is 27.1 metres wide with the scanner level, 15.4 at 16 degrees and 2.5 at 45, and there is none at 60 degrees or straight down. Straight mapping passes left no hole. The red ring was not reproduced.

What this does not show

• Reports: self-reported to one organisation, with small counts, so a large excess is ruled out, not a modest one. Events on the ranch itself went elsewhere.
• Ledger: absence from a short newspaper article shows only that something was not published then. Four quotations could not be reached, and the 2021 book is known second-hand.
• Audit: recaps are not the programme, and editing may cut controls that were run.
• Time warp: it tests gravity as general relativity describes it, using 10 metre to 1 kilometre errors chosen by the plan, because no GPS logs are public.
• Lidar: scanner, height and hovering rest on forum readings, and the hole's width on screen is unpublished.

Nothing here shows the ranch is ordinary. It shows that the public evidence does not yet need anything extraordinary, and that without raw data nobody can test more. The page is marked provisional because some checks are still open.

Where it goes in the books

Book 7, The Ranch at the Edge of Time. All five tests, with the 1.6 GHz signal test, are in the Interlude, "The Machine Checks the Signal". The report counts and 1996 baseline are in Chapter 1, the audit in Chapter 7 and the lidar in Chapter 8. Check it yourself: the six audit questions fit on one sheet to fill in while watching, and the lidar claim needs a drone, a car park and ten minutes.