Calibration record

How the ranges have scored

Every range is published before the decision, frozen on the day it is made, and graded against the public register when the decision lands. Nothing is revised afterwards.

19 of 26graded files inside the 80% range

19 of 26 Decisions that landed inside the 80% range we published, out of the decisions graded since the record began on 14 August 2026.

The record: Ontario exploration permits

No percentage until 50 files

On a handful of files a percentage reads as a verdict, and a record this short is not one in either direction. Below 50 graded files the tally is published instead. The rule was fixed before any permit type came close to meeting it, and it applies whether the count flatters us or not.

If the ranges are honest, about eight in every ten graded files should land inside. On 26 graded files that expectation is 20.8, which tells you nothing yet. That is why the count is shown rather than a rate.

The narrower 50% range, which should contain about half of all decisions, contained 5 of 26, where 13 would be expected.

Nothing from before 14 August 2026 is pooled with these numbers.

Graded

Every graded file, with the range it was graded against

Each file is graded against the first range published for it, not the last one before the decision landed, and against the day the decision was posted to the register.

Permit	Applied	First forecast	80% range	Middle	Decided	Result	By how much
026-0408	14 Apr 2026	15 Aug 2026	18 Aug 2026 to 11 Dec 2026	16 Sep 2026	16 Aug 2026	Outside	3 days before the range opened
026-0627	11 Jun 2026	15 Aug 2026	5 Sep 2026 to 3 Jul 2033	9 Apr 2027	3 Sep 2026	Outside	3 days before the range opened
026-0721	8 Jul 2026	15 Aug 2026	22 Aug 2026 to 1 Jun 2032	18 Oct 2026	8 Sep 2026	Inside
026-0762	20 Jul 2026	15 Aug 2026	1 Sep 2026 to 13 Jan 2031	16 Oct 2026	9 Sep 2026	Inside
026-0763	20 Jul 2026	15 Aug 2026	31 Aug 2026 to 21 Oct 2028	10 Oct 2026	9 Sep 2026	Inside
026-0783	23 Jul 2026	15 Aug 2026	2 Sep 2026 to 30 Jul 2029	2 Oct 2026	9 Sep 2026	Inside
026-0824	7 Aug 2026	15 Aug 2026	12 Sep 2026 to 31 Dec 2026	29 Sep 2026	9 Sep 2026	Outside	3 days before the range opened
019-7204	9 Jun 2023	15 Aug 2026	1 Jan 2028 to 11 Apr 2029	11 Apr 2029	11 Sep 2026	Outside	478 days before the range opened
026-0842	11 Aug 2026	15 Aug 2026	17 Sep 2026 to 8 May 2027	6 Oct 2026	16 Sep 2026	Outside	1 day before the range opened
026-0722	8 Jul 2026	15 Aug 2026	23 Aug 2026 to 13 Oct 2029	3 Oct 2026	21 Sep 2026	Inside
026-0858	16 Aug 2026	22 Aug 2026	19 Sep 2026 to 18 Nov 2026	2 Oct 2026	21 Sep 2026	Inside
026-0859	16 Aug 2026	22 Aug 2026	18 Sep 2026 to 20 Nov 2026	1 Oct 2026	21 Sep 2026	Inside
019-8552	16 Apr 2024	15 Aug 2026	30 May 2027 to 4 Mar 2030	4 Mar 2030	25 Sep 2026	Outside	247 days before the range opened
026-0704	7 Jul 2026	15 Aug 2026	21 Aug 2026 to 13 Nov 2029	25 Sep 2026	25 Sep 2026	Inside
026-0814	31 Jul 2026	15 Aug 2026	14 Sep 2026 to 27 Jun 2032	18 Nov 2026	25 Sep 2026	Inside
026-0815	31 Jul 2026	15 Aug 2026	14 Sep 2026 to 13 Jul 2032	18 Nov 2026	25 Sep 2026	Inside
026-0816	31 Jul 2026	15 Aug 2026	15 Sep 2026 to 24 Sep 2032	6 Dec 2026	25 Sep 2026	Inside
026-0846	11 Aug 2026	15 Aug 2026	21 Sep 2026 to 30 Sep 2028	7 Nov 2026	28 Sep 2026	Inside
026-0857	16 Aug 2026	22 Aug 2026	25 Sep 2026 to 29 Aug 2028	20 Nov 2026	28 Sep 2026	Inside
026-0537	1 Jun 2026	15 Aug 2026	4 Sep 2026 to 22 Apr 2033	24 Mar 2027	29 Sep 2026	Inside
026-0866	20 Aug 2026	22 Aug 2026	5 Oct 2026 to 24 Jun 2033	21 Dec 2026	8 Oct 2026	Inside
026-0915	3 Sep 2026	3 Oct 2026	19 Oct 2026 to 1 Apr 2033	9 Feb 2027	15 Oct 2026	Outside	5 days before the range opened
026-0888	27 Aug 2026	2 Sep 2026	11 Oct 2026 to 10 Aug 2032	29 Dec 2026	16 Oct 2026	Inside
026-0778	22 Jul 2026	15 Aug 2026	4 Sep 2026 to 27 Aug 2032	29 Oct 2026	21 Oct 2026	Inside
026-0918	3 Sep 2026	3 Oct 2026	16 Oct 2026 to 9 Jan 2029	19 Nov 2026	22 Oct 2026	Inside
026-0929	9 Sep 2026	12 Sep 2026	15 Oct 2026 to 17 Jan 2027	29 Oct 2026	22 Oct 2026	Inside

What happened, one miss at a time

Written by hand, one sentence per miss. A miss with no note renders as “not yet discussed”, which is a true statement and is meant to look like an unpaid debt.

026-0408

We said 5 to 119 days and the decision appeared in 2. This file had been on hold since 2 June 2026 pending consultation, and the notice that began the hold named 23 June as the expected decision date. By the time we forecast, that date was 52 days past. The model reads the date named when a hold begins; it does not notice when that date has come and gone.

Missed

026-0627

Not yet discussed.

Missed

026-0824

Not yet discussed.

Missed

019-7204

Not yet discussed.

Missed

026-0842

Not yet discussed.

Missed

019-8552

Not yet discussed.

Missed

026-0915

Not yet discussed.

Missed

Excluded

Files that cannot be graded, and why

A file that cannot be graded is listed with its reason rather than dropped.

Permit	Recorded	Reason
019-1877	15 Aug 2026	The register recorded a refusal without a date to grade against.

Annulled

11 files whose questions were retired, each with its reason

A question is one claim made about one file: its timing range, or its chance of a decision within 13 or 26 weeks, or of being put on hold. A question is retired when grading it would measure something other than the model. Most of these are permits a data defect had forced to look pending when the register had already decided them, in some cases years earlier. These files account for 65 retired questions, 6 of them timing ranges of the kind the tally above counts.

Permit	Recorded	Questions	Reason
019-3040	2 Sep 2026	11	This permit had been decided years before the forecast was made, but an error in Kestrel's own database listed it as still awaiting a decision. Grading the forecast would measure nothing about the model, so the question was retired rather than graded.
019-4519	2 Sep 2026	8	The ministry had already issued this permit before Kestrel's first forecast for it was made, though the public registry only posted the decision later. Grading the forecast would measure nothing about the model, so the question was retired rather than graded.
019-7460	2 Sep 2026	9	The ministry had already issued this permit before Kestrel's first forecast for it was made, though the public registry only posted the decision later. Grading the forecast would measure nothing about the model, so the question was retired rather than graded.
025-0692	2 Sep 2026	9	The ministry had already issued this permit before Kestrel's first forecast for it was made, though the public registry only posted the decision later. Grading the forecast would measure nothing about the model, so the question was retired rather than graded.
026-0671	2 Sep 2026	7	The ministry had already issued this permit before Kestrel's first forecast for it was made, though the public registry only posted the decision later. Grading the forecast would measure nothing about the model, so the question was retired rather than graded.
026-0733	2 Sep 2026	5	The ministry had already issued this permit before Kestrel's first forecast for it was made, though the public registry only posted the decision later. Grading the forecast would measure nothing about the model, so the question was retired rather than graded.
025-1325	19 Sep 2026	4	The ministry had already issued this permit before Kestrel's first forecast for it was made, though the public registry only posted the decision later. Grading the forecast would measure nothing about the model, so the question was retired rather than graded.
026-0485	19 Sep 2026	2	The ministry had already issued this permit before Kestrel's first forecast for it was made, though the public registry only posted the decision later. Grading the forecast would measure nothing about the model, so the question was retired rather than graded.
026-0520	19 Sep 2026	2	The ministry had already issued this permit before Kestrel's first forecast for it was made, though the public registry only posted the decision later. Grading the forecast would measure nothing about the model, so the question was retired rather than graded.
026-0620	19 Sep 2026	4	The ministry had already issued this permit before Kestrel's first forecast for it was made, though the public registry only posted the decision later. Grading the forecast would measure nothing about the model, so the question was retired rather than graded.
026-0621	19 Sep 2026	4	The ministry had already issued this permit before Kestrel's first forecast for it was made, though the public registry only posted the decision later. Grading the forecast would measure nothing about the model, so the question was retired rather than graded.

Withdrawn

The withdrawal register

A single forecast is withdrawn only when it was produced with a defect we can demonstrate without reference to how it scored, and never because it scored badly. The withdrawal is recorded, dated and published here, and the withdrawn forecast is out of the record above. Separately, until the record is formally launched, any part of it Kestrel no longer stands behind, from a single forecast to a whole permit type, can be removed from the record entirely, and those removals are not listed here until the launch; the method says more.

Model version	Made for	Recorded	Reason
ON-ero-2026l	9 Jun 2026	12 Aug 2026	This set of forecasts was produced twice on the same day. The first run had a fault and was replaced by a corrected run later that day, so the first run is withdrawn and only the corrected one counts.

9 weekly batches of forecasts were refused admission for Ontario exploration permits, from 5 model versions: each failed a check that a batch must pass to enter the record, so none of them is graded here.

How a range is made, which date it forecasts, and every rule the record is graded by:

The method →