FPL 2025/26 reconstructive replay: what the season shows
Over 38 gameweeks the forecast optimiser scored 2,010 net points versus 1,990 for the template. Same-state evidence copied the optimiser every week. An exploratory post-season fork that kept a different squad state from Gameweek 12 finished 65 points ahead; that fork is not a live freeze and is not a skill headline. No chip was used.
The published path
Each gameweek page is built from the lab’s reconstructive artifacts: a template that rolls transfers, a forecast optimiser (the selected plan), and a same-state evidence arm. The cutoff is first kickoff minus 90 minutes, rebuilt after the season. It is not a two-hour live freeze.
The optimiser finished 2,010 net points. The template finished 1,990. Same-state evidence scored 2,010 — identical week by week, because the evidence arm selected the same plan as the optimiser in every gameweek folder.
What “evidence” did and did not do
A separate exploratory fork, recovered after outcomes were known, kept a different squad from Gameweek 12 and finished 2,075 hybrid net versus 2,010 canonical (+65 from Gameweek 12 to 38). Paired same-state tests from Gameweek 13 to 38 attribute only +16 points directly to that week’s evidence, in four weeks.
Those fork numbers are on the season page as context. They are not the squads published on each gameweek page, and they are not a claim that unstructured news beat the forecast.
Chips and Gameweek 34
Wildcard, Free Hit, Triple Captain and Bench Boost were never played. In Gameweek 34, a blank, both the optimiser and the template took three transfers with one free transfer and paid eight points. Free Hit and Wildcard were still available. That is the main chip lesson on this path: the replay evaluated transfers, lineups and captains, not a complete FPL policy.
Is 2,010 a live mini-league score?
No. It is the net total of a reconstructive optimiser path with rebuilt inputs. Private manager state, exact pre-deadline prices, and live news were not archived.
Why is the evidence column the same as the optimiser?
In the per-gameweek artifacts, the evidence arm produced the same validated plan as the forecast optimiser all 38 weeks. The +65 figure comes from a later forked squad trajectory, not from those same-state plans.