readingsignal

2026-09-26

The markets board was published from an out-of-date copy of the model code between 11 September and 16 September. All 136 updates in that window used the scenario weights adopted at the second evaluation rather than the third.

No published figure has been altered; each affected update is flagged in the provenance record. The publisher now checks that its code and weights match what is adopted before it runs, and refuses to publish if they do not.

Weights: Unchanged (final): Labour recovers 87.4 · Reform ahead 1.2 · Reform replaces Con 0.6 · Reform slips 0.4 · Right splits 0.2 · Reform collapses 0.1 · held back 10 (of 100)

2026-09-16

This scenario set is archived. Its evaluation window closed on 11 September; a re-registered successor framework opened today and no weight here will move again.

The scenario that fired and was confirmed at all three evaluations — “Labour recovers under a new leader” — is no longer a possible future: it is the baseline the polls now describe. Under the rules committed in advance, absorbing a confirmed scenario into the baseline requires a fresh pre-registered framework rather than editing this one, so a successor set of six scenarios around the new baseline was registered today with its own weights and triggers. These weights stand as this set's final record and will still be scored at the general election under this set's own rules. This section moves to a resolved-and-retired record when the successor's page lands.

Weights: Unchanged (final): Labour recovers 87.4 · Reform ahead 1.2 · Reform replaces Con 0.6 · Reform slips 0.4 · Right splits 0.2 · Reform collapses 0.1 · held back 10 (of 100)

2026-09-15

Every date on this page is now read from this update log rather than written into the copy. No weights moved.

Copy only: two sentences still said the last evaluation happened on 31 July — two evaluations stale — and one scenario card carried its trigger date as fixed text. Dated claims (when weights were last evaluated, whether the window is closed, when a trigger fired) are now derived from this log or from the scenario’s own dated fields, so they update themselves at any future evaluation instead of quietly going stale. Definitions, weights and triggers are unchanged.

Weights: Unchanged: Labour recovers 87.4 · Reform ahead 1.2 · Reform replaces Con 0.6 · Reform slips 0.4 · Right splits 0.2 · Reform collapses 0.1 · held back 10 (of 100)

2026-09-11

The “Labour recovers under a new leader” card now states when its weight was set, read from this update log. No weights moved.

Copy only: the card still said its weight was set on 31 July — two evaluations stale. The provenance line (“This weight was set on 11 September 2026 at E3. The window is closed.”) is now derived from the latest logged evaluation rather than written by hand, so it updates itself if a future evaluation point reopens the weights. Definitions, weights and triggers are unchanged.

Weights: Unchanged: Labour recovers 87.4 · Reform ahead 1.2 · Reform replaces Con 0.6 · Reform slips 0.4 · Right splits 0.2 · Reform collapses 0.1 · held back 10 (of 100)

2026-09-11

Third and final scheduled evaluation: “Labour recovers under a new leader” strengthens again (75.2 → 87.4). The evaluation window is now closed.

Eight weeks after the leadership change, the last pre-committed evaluation scored forty-nine fully post-transition polls against the same pre-transition baseline. Labour is up 5.3 points, and the composition test passes a third time: the rise is funded from Green and Liberal Democrat switchers rather than from the right. New this time, the leader’s published approval ratings crossed the threshold set in advance — at the first two evaluations they sat just below it and counted for nothing, so this is the first evaluation at which they add anything. Two honest limits on that: the published recall tables have not been updated since the first evaluation, so the composition evidence is the same evidence scored again rather than new confirmation; and the approval reading is a mean across seven polling houses spanning +5 to +26. An independent precision audit re-derived every step before adoption, and confirmed the result is unchanged whether or not a duplicated poll is included. This was the last scheduled evaluation of this window: the weights now stand until a named event reopens them.

Weights: Labour recovers 87.4 · Reform ahead 1.2 · Reform replaces Con 0.6 · Reform slips 0.4 · Right splits 0.2 · Reform collapses 0.1 · held back 10 (of 100)

2026-08-14

Second scheduled evaluation: “Labour recovers under a new leader” strengthens further (47.6 → 75.2).

Four weeks after the leadership change, the pre-committed evaluation scored twenty-five fully post-transition polls against the same pre-transition baseline. Labour is up close to five points, and the decisive test again: the rise is funded from Green and Liberal Democrat switchers, not from the right — this scenario's specific signature, confirmed by the published recall tables. The update uses the same arithmetic committed before any of these polls existed, and an independent precision audit re-derived every step before adoption — including checking that the result is unchanged when an ineligible large-sample poll and one double-ingested poll are removed from the set. Every other scenario shrinks proportionally; the held-back share is unchanged.

Weights: Labour recovers 75.2 · Reform ahead 7 · Reform replaces Con 3.3 · Reform slips 2.5 · Right splits 1.2 · Reform collapses 0.8 · held back 10 (of 100)

2026-07-31

Scenario descriptions refreshed to reflect the 31 July evaluation, and “Things stay roughly as they are” renamed “Reform remains clearly ahead”. No weights moved.

Copy only: the descriptions of “Things stay roughly as they are” and “Labour recovers under a new leader” were written before the evaluation and still read as if the pre-change picture were the most likely world. The renamed scenario is the same world, definition and trigger — earlier entries' “Stay as-is” refers to it. Definitions and weights are unchanged by this edit — the weights were adopted at the first scheduled evaluation, recorded in the entry below. The “most likely scenario” tag now follows the weights automatically instead of being written by hand.

Weights: Unchanged: Labour recovers 47.6 · Reform ahead 20 · Reform replaces Con 9.4 · Reform slips 7.1 · Right splits 3.5 · Reform collapses 2.4 · held back 10 (of 100)

2026-07-31

First scheduled evaluation after Labour's leadership change: “Labour recovers under a new leader” becomes the most likely single scenario (18 → 47.6).

Andy Burnham was confirmed as Labour leader on 17 July, which fired this scenario's trigger. Under the evaluation rules committed before his confirmation, we scored the eleven fully post-transition polls against the pre-transition baseline: Labour is up about four points, and — the decisive test — that rise is funded almost entirely by voters returning from the Greens and Liberal Democrats, not from the right. That is this scenario's specific signature, so its weight roughly triples and every other scenario shrinks proportionally. The held-back share is unchanged.

Weights: Labour recovers 47.6 · Stay as-is 20 · Reform replaces Con 9.4 · Reform slips 7.1 · Right splits 3.5 · Reform collapses 2.4 · held back 10 (of 100)

2026-07-06

First publication: the six scenarios, their weights, and their triggers.

Initial weights set under the pre-registered scenario framework — committed before any scenario was run through the seat model. One trigger is already armed: Labour's leadership contest is under way, so “Labour recovers under a new leader” gets a dated review when a successor is elected.

Weights: Stay as-is 34 · Labour recovers 18 · Reform replaces Con 16 · Reform slips 12 · Right splits 6 · Reform collapses 4 · held back 10 (of 100)