PNOR — PetroNor E&P ASA

PetroNor E&P ASA (PNOR.OL) — deep-value reanalysis after the fall of the roe falsifier

September 18, 2026 · price NOK … · market cap NOK 1,756.7 mil = … · reporting currency USD, listing currency NOK

A resumption of the analysis from 2026-09-03 (rapoarte/deep/2026-09-03-deep-PNOR.md), triggered mechanically on 2026-09-17 by the fall of one of the seven falsifiers of the thesis written then.

Executive summary (1 page: thesis, estimated value, verdict)

The task’s question was: does the assumption supporting the verdict still hold or not? The answer is that the assumption holds, and the falsifier was miswritten from birth. It isn’t structural deterioration and it isn’t even quarterly noise — it’s a test-specification error, of the same family as errors already documented on other tickers: the chosen metric could never measure the claim it was supposed to test.

Load-bearing claim #3 from the thesis written on 2026-09-03 said: “Reported earning power is a lifting-cycle peak, not a run-rate: the trailing 12 months contain A SINGLE oil lift, at …/bbl, and the company has explicitly said it won’t sell oil again in 2026. Normalized cycle profit is ~…-20 mil/year at …/bbl, so normalized ROE is 11…, not the reported 29.7%.” The figure cited in support of the claim was, verbatim, “reported ROE 29.7%.” The attached falsifier was roe > 0.25. In other words: the claim cited its own evidence as the rejection threshold. The test would trigger exactly when the claim was most true. It was born triggered, and on 2026-09-17 it was read as such, at a value of 0.2723 — 8.9% above the 0.25 threshold, not 40%.

The arithmetic shows why the metric could never work here. Profit attributable to shareholders over the last four quarters is … (… in Q2 2026 · −3.4 in Q1 2026 · +3.3 in Q4 2025 · −2.3 in Q3 2025). A single quarter — the one containing the 964,593-barrel cargo sold in April 2026 at …/bbl — brings in …, i.e. 107% of the total; the other three quarters summed give minus … Attributable equity at 2026-06-30 is …, so ROE = 36.4/133.7 = 27.2%. The same company, measured over a four-quarter window with no lift, would have an ROE of …. For a producer that sells once or twice a year, roe isn’t a business indicator, it’s an indicator of where a tanker’s schedule fell relative to the reporting date. It will keep triggering at every check until the April 2026 lift exits the TTM window — i.e. until the Q2 2027 report, in August 2027 — and will retrigger right after, if the May 2027 lift falls in the same calendar quarter.

The check also uncovered a second, independent error, not caught by the mechanical run: falsifier #7, fcf_yield > 0.30, is dead for a different reason. teza_deep.py computes the yield as FCF / marketCap with no currency conversion (line 408 onward: fcf from the cash flow statement, mcap from info). At PetroNor the flow is in USD and the market cap in NOK, so the ratio comes out … / NOK 1,758.7 mil = 4.76% — exactly the figure the check reported. The real yield, correctly calculated by the data pack with conversion (market cap NOK→USD @ 0.1061), is 44.9%. The error is off by a factor of 9.42, exactly the USD/NOK rate. The falsifier should have triggered (44.9% > 30%) and couldn’t. Where roe spoke for nothing, fcf_yield stayed silent for nothing.

A third observation, also from the check: the minimum-distance rule introduced into teza_deep.py on 2026-09-17 (DISTANTA_RELATIVA_MIN = 0.15) would today reject, at write time, two of the old thesis’s seven falsifiers: roe at a 0.25 threshold versus a current value of 0.2723 is at 8.2% distance, and actiuni > 150,000,000 versus 142,356,855 is at 5.4%. The 09-03 thesis was written a day or two before the guard existed. Those two falsifiers weren’t tests, they were alarms glued to the quote.

What actually changed at the company in those 15 days: nothing. There’s no new reporting (Q3 2026 comes in November), no new legal development, no new operational release. What changed is the price: … → … (…), against a Brent that rose from 94.52 to … (…). P/B rose from 1.33 to 1.41 — a new high for the five-year series (0.39-0.82× in 2022-2025). The simulation’s median margin of safety, on identical assumptions, fell from … to …. The “too expensive” thesis wasn’t disproven by anything; it became a little more true.

Estimated value: range … … …, median … (five models, detailed in the valuation chapter). At NOK …, the price embeds a normalized Brent of …/bbl over the license’s remaining life — below the trailing-12-month Brent average (…) and well below today’s spot (…). That’s an important nuance, and one in the bull’s favor: the market is not extrapolating the spot price. What you don’t get at NOK … is a margin of safety: you pay a reasonable valuation for a unique, non-operated, depleting asset (R/P 9.3 years, 2025 reserve-replacement rate 14%), with a “grand corruption” criminal proceeding on the very license generating 100% of revenue, starting in 66 days (2026-11-23) and for which there’s no provision.

Verdict: DON’T BUY NOW, WATCH — reaffirmed and marginally strengthened versus 2026-09-03. Re-evaluation threshold: NOK 10.0, where net asset value calculated at the company’s own reserve deck is reached (…/bbl → …). Real deterioration, if it appears, won’t show up in ROE — it will show up in gross margin per barrel, which today is stable at …… in quarters with and without a lift, and in the Q3 2026 report, where the overlift position must fall by ~150,000 barrels.