[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_191":192,"_193":194},"routes/_app.x.$handle.$name",{"_11":12,"_13":14,"_15":16,"_18":19,"_17":-7,"_144":-5,"_20":145,"_22":146,"_147":148,"_149":150,"_26":151,"_24":152,"_153":-5,"_154":-5,"_155":65,"_156":157,"_158":60,"_159":-5,"_160":-5,"_161":162},"routes/_app.x.$handle.$name.branch.$branch.code.$",{"_11":12,"_13":14,"_15":16,"_17":-7,"_18":19,"_20":21,"_22":23,"_24":25,"_26":27,"_28":-5,"_29":-5,"_30":-5,"_31":-5,"_32":-5},"handle","arvindfroi","name","fpl-prosedyre","branchName","main","version","path","main.ts","project",{"_33":90,"_13":14,"_45":91,"_92":-5,"_93":65,"_94":95,"_96":95,"_97":98,"_99":-5,"_100":101,"_102":38,"_103":60,"_104":-5,"_105":-5,"_106":77,"_107":108,"_109":-5,"_110":-5,"_111":-5,"_112":-5,"_113":114,"_115":114,"_116":117,"_118":117,"_119":117,"_120":114,"_121":114,"_122":123,"_124":125,"_126":127,"_128":65},"tree",[51,52,53,54],"file",{"_33":34,"_35":36,"_37":38,"_13":19,"_39":40,"_41":42,"_17":43,"_44":-5,"_45":46,"_47":48,"_49":-5,"_18":50},"contents",[],"upstream","pull","ahead","behind","envVars","id","a88241bb-224d-4962-92e3-e7d5584b743f","branchId","40ceb8f7-c541-4b1d-84ab-da50335ca551","fileId","e609966a-9e05-11f1-91bc-1607ee4eb77e","content","import { readFile } from \"https://esm.town/v/std/utils/index.ts\";\nimport { blob } from \"https://esm.town/v/std/blob/global.ts\";\n\n// Serverer FPL-prosedyren, elitekohorten og den forhaandsregnede etterretningen.\n// / -> prosedyre.txt (kanonisk fristprosedyre)\n// /elite -> elite1000.txt (953 entry-IDer, kommaseparert)\n// /intel -> JSON: eierskap, EO, blokkbevegelse, chips, kohortsnitt (regnet av intel.ts)\nexport default async function (req: Request): Promise {\n const p = new URL(req.url).pathname.replace(/\\/+$/, \"\");\n if (p === \"/intel\") {\n try {\n const data = await blob.getJSON(\"fpl_intel\");\n return new Response(JSON.stringify(data, null, 1), {\n headers: { \"content-type\": \"application/json; charset=utf-8\" },\n });\n } catch (e) {\n return new Response(`INTEL IKKE KLAR: ${String(e)}`, { status: 503 });\n }\n }\n const file = p === \"/elite\" ? \"elite1000.txt\" : \"prosedyre.txt\";\n try {\n const body = await readFile(`/${file}`, import.meta.url);\n return new Response(body, {\n headers: { \"content-type\": \"text/plain; charset=utf-8\" },\n });\n } catch (e) {\n return new Response(`NOT FOUND: ${file} — ${String(e)}`, { status: 404 });\n }\n}\n","fileType","http",4,"parentId","createdAt",["D",1788769635681],"branchVersion",9,"label",[19],{"_33":82,"_13":83,"_37":84,"_44":-5,"_35":36,"_47":85,"_59":60,"_45":86,"_17":77,"_41":24,"_39":87,"_64":65,"_49":-5,"_66":88,"_18":89},{"_33":72,"_13":73,"_37":74,"_44":-5,"_35":36,"_47":75,"_59":60,"_45":76,"_17":77,"_41":78,"_39":79,"_64":65,"_49":-5,"_66":80,"_18":81},{"_33":34,"_13":19,"_37":38,"_44":-5,"_35":36,"_47":48,"_59":60,"_45":69,"_17":43,"_41":42,"_39":40,"_64":65,"_49":-5,"_66":70,"_18":71},{"_33":55,"_13":56,"_37":57,"_44":-5,"_35":36,"_47":58,"_59":60,"_45":61,"_17":62,"_41":24,"_39":63,"_64":65,"_49":-5,"_66":67,"_18":68},"97322b85-f778-48c5-bf4f-a57517c2a3c2","prosedyre.txt","1af0dbf4-9e06-11f1-8e8a-1607ee4eb77e",10,"latest",true,["D",1788769692235],2,"FPL DEADLINE PROCEDURE — canonical. arvindinho, entry ID 5038450, 2026/27.\nWrite to Arvind in NORWEGIAN. Decide, do not ask. This run fires ~2 hours before the deadline,\nafter the press conferences. It is the FINAL decision for this gameweek — say so plainly.\n\n=== TIME ZONE RULE — APPLIES EVERYWHERE, NO EXCEPTIONS ===\nAll trigger times are computed and stored in UTC. All times SHOWN TO ARVIND are Norwegian local.\nNorway is UTC+2 (CEST) until 25 Oct 2026 03:00 local, then UTC+1 (CET) until 28 Mar 2027.\nSo a 17:30 UTC deadline is 19:30 norsk tid in September, 18:30 in November.\nNEVER quote a bare UTC time to him without the Norwegian time beside it. Derive the offset from\nthe actual date each run — do not hardcode it.\n\n=== READ-ONLY RULE ===\nThis procedure file is READ-ONLY for scheduled runs. Do NOT edit it, do not call storeBlob or\nupdate_file on arvindfroi/fpl-prosedyre, and do not rewrite your own instructions. If a run\nconcludes the procedure should change, write the FINDING and the PROPOSED CHANGE into memory\n(/areas/fpl-chip-plan.md or /topics/football.md) and say so in the message to Arvind. He and a\nlive session apply changes to this file. Facts go to memory, which is writable. Rules live here,\nwhich is not. That separation is what stops the method drifting away from what was agreed.\n\n=== THE CHAIN — THREE RUNS PER GAMEWEEK ===\n 1. FRISTKJØRING — deadline minus 2 h. Full analysis. This file. Creates link 3.\n 2. PÅMINNELSE — deadline minus 45 min. Three lines, actions only, no analysis.\n 3. ETTERKJØRING — after the last match of the round. Reviews it, then creates 1 and 2 for the\n NEXT gameweek. This is the link that rebuilds the chain.\nEach link creates the next. A weekly Monday chain-guard repairs anything that died silently.\n\n=== STEP 0 — CREATE THE ETTERKJØRING BEFORE YOU DO ANYTHING ELSE ===\nThe chain must survive even if the rest of this run fails. So do this first, not last.\n1. GET https://fantasy.premierleague.com/api/fixtures/ and find the LAST kickoff_time of the\n gameweek you are deciding now. Match ends ~2 h after kickoff; bonus points settle ~1-2 h later.\n2. Target = last kickoff + 4 hours, in UTC. THEN APPLY THE CIVILISED-HOUR RULE: convert to\n Norwegian local time; if it falls between 22:00 and 06:00 norsk, move it to 07:00 norsk the\n next morning (converted back to UTC). Never wake him at night for a review that can wait.\n3. Call list_triggers. If an ENABLED trigger named \"FPL etterkjøring — GW{n}\" already exists,\n skip to step 5 — do not create a duplicate.\n4. create_trigger with:\n name: \"FPL etterkjøring — GW{n} ({dd.mm})\"\n run_once_at: the time from step 2, RFC3339 UTC\n initiation: human_schedule\n notifications: {push: true}\n prompt: EXACTLY the text in the block ETTERKJØRINGSPROMPT at the bottom of this file,\n with {n} replaced by this gameweek's number. Nothing else.\n5. Tell Arvind at the END of your message when the etterkjøring runs, in Norwegian local time.\nIf the season is over (no further fixtures), skip creation and say so.\n\n=== STEP 1 — WHERE WE STAND ===\n- /api/entry/5038450/event/{last_finished_gw}/picks/ for the ACTUAL squad. Never assume it.\n NOTE: entry_history.points can lag during a live round. The truth is the sum of\n element event_points x multiplier over the XI, and the league standings endpoint. Cross-check.\n- /api/entry/5038450/ and /history/ for total points, overall rank, bank, free transfers.\n- Report cumulative points against 62.3 x gameweeks played (target total 2447 — measured from\n FPL's own records, not estimated). Say plainly whether top 1000 is still live.\n- Two goals: (1) win Trivselsligaen (league 973274, 9 players), (2) top 1000 overall. For ~30 of 38\n gameweeks they are the same problem. Do not steer by tracking error before about GW30.\n\n=== STEP 2 — AVAILABILITY. THIS IS WHERE THE POINTS ARE. ===\nBacktest evidence: 70% of the model's over-prediction was players who did not play.\n- bootstrap-static: status and chance_of_playing_next_round for all 15 and for every buy candidate.\n- CROWD SENSOR (our best measured edge): net_out = (transfers_out_event - transfers_in_event)/selected.\n Measured on 6,498 player-gameweeks — base rate of zero minutes 9.1%; at 15% net_out it is 44.9%,\n at 20% 57.7%, at 30% 76.7%. Over 15% = investigate. Over 20% = treat as unavailable unless team\n news says otherwise. Counters reset each gameweek and are freshest near the deadline.\n- Search the press conferences and predicted XIs for every club we hold and every candidate. Club\n websites first, then Fantasy Football Scout / Sports Mole predicted line-ups. Friday-afternoon\n pressers for Sunday games land AFTER a 3 p.m. run — that is why this run moved to 2 h before.\n- THE GATE: any player over 4.5m must be a clear starter to be eligible. EXCLUDE — do not multiply\n expected points by probability of starting. 4.5m and under is exempt (bought not to play).\n Backtest GW1-8: no gate 398, gate 0.73-0.78 gives 413-430, gate 0.80 collapses to 352.\n The gate is only valid on a NEWS-INFORMED estimate. News first, then gate.\n- FREE: /api/element-summary/{id}/ gives xG, xA, threat, creativity and bps PER GAMEWEEK. It is the\n same underlying data the paid sites resell. Read it for the squad and for candidates.\n\n=== STEP 3 — THE ELITE COHORT: WHAT CHANGED, AND HOW CONSENSUS MOVED ===\n**BRUK DEN FERDIGREGNEDE FEEDEN. ETT KALL, IKKE 1900.**\n curl -s https://arvindfroi--e609966a9e0511f191bc1607ee4eb77e.web.val.run/intel\nEn Val Town-jobb (intel.ts) regner dette ut hver 6. time fra alle 953 kohorttropper og returnerer\nJSON: gameweek, n, kohortens snitt/median/p10/p90/maks, vaart lag, chipbruk, hele eierskapstabellen\n(elite % mot globalt %, EO, vaar multiplikator, nettoposisjon, poeng), de ti verste nettoposisjonene\n(`verste_netto`), og blokkbevegelse ALLEREDE SPLITTET til det rene utvalget (`blokk_ekte` = ingen\nchip og hoeyst 2 bytter). Feltnavnene er norske.\nSJEKK `generert` OG `gameweek` FOERST. Ligger `gameweek` bak siste ferdigspilte runde er feeden\nforeldet — si det, og bruk raa-metoden under for denne ene kjoeringen.\nRAA RESERVE (kun hvis feeden er foreldet eller nede):\nThe 953 verified entry IDs of LAST SEASON'S top 1000 are served as plain text at\n https://arvindfroi--e609966a9e0511f191bc1607ee4eb77e.web.val.run/elite\n(comma-separated). NEVER rebuild the cohort — that took 86 minutes. It comes from FPL's own system\nleague 321, \"Top 1% 25/26 League\".\nFetch /api/entry/{id}/event/{gw}/picks/ for all 953 at ~28 threads (~40 seconds) for BOTH the last\nfinished gameweek and the one before it, then compute:\n a) OWNERSHIP in the cohort vs global selected_by_percent. Biggest positive and negative gaps.\n b) EFFECTIVE OWNERSHIP: share who have him in the XI plus captaincies. Our net position per player\n = our multiplier x 100 minus their EO. That is rank movement per point he scores.\n c) TRANSFER FLOW AS A BLOC — who did the 953 move IN and OUT between the two gameweeks.\n This is the real edge; nobody publishes it. Report the three biggest moves in each direction\n with the size of the shift, and say whether we should follow.\n d) CHIP USAGE in the cohort. It tells you when the elite thinks a chip week has arrived.\nHONEST LIMIT YOU MUST STATE: picks are only visible for FINISHED gameweeks, so their bloc movement\nis one week lagged. TESTED AND CLOSED 28 Aug: /api/entry/{id}/transfers/ returns an EMPTY list for\nevery entry before the deadline (250 of 250 tested) — there is no back door. Do not try again.\nThe live signal for THIS gameweek is the global transfer flow in bootstrap-static. Use the lagged\nelite flow for direction and the live global flow for urgency. Never present lagged as live.\nWHAT YOU CAN DO PRE-DEADLINE INSTEAD: cross the cohort's last-finished squads against this week's\nlive global flow. It shows what the crowd is buying that the elite ALREADY owns (follow) versus\nwhat the crowd is buying that the elite does not own (ignore). Measured 28 Aug: the three most\nbought players in the game that week were not elite assets, while Calafiori and João Pedro — also\nheavily bought — were already owned by 77.5% and 91.1% of the cohort.\nINTERPRETATION RULE: the biggest elite-minus-global gaps are usually CHEAP players owned to FUND\nthe premiums, not better players. Elite ownership is a SEARCHLIGHT — go find out why they are there\n— never a shopping list.\n\n=== STEP 4 — IS OUR SQUAD TOP-1000 QUALITY? ===\nAnswer this explicitly every run, with numbers, not vibes:\n 1. Sum of elite ownership across our 15, against the distilled elite squad. GW1 baseline: ours\n 534, distilled 877. Rising is good.\n 2. How many of the distilled elite 15 we own. GW1 baseline: 8 of 15.\n 3. Our three worst net-EO positions — the players where we lose most ground per point scored.\n 4. Points against the 62.3 pace, and rank trajectory.\nThen say in one sentence whether the squad is on top-1000 trajectory, and if not, what the single\nbiggest structural hole is.\n\n=== STEP 5 — OUR OWN MODEL ===\nDixon-Coles team ratings on four seasons (decay 0.9/yr, weights RESCALED to sum to N, pooling\nwm/(wm+45)), then BLEND WITH BOOKMAKER ODDS AT 0.8 MARKET / 0.2 MODEL, geometric. Derive market\nlambdas from 1X2 + over/under 2.5 by removing the overround and solving under Poisson.\nOdds for the coming round: https://www.football-data.co.uk/fixtures.csv (Div E0). Historical:\nfootball-data.co.uk E0 files. Historical per-gameweek FPL data:\nraw.githubusercontent.com/vaastav/Fantasy-Premier-League/master/data/{season}/gws/merged_gw.csv\nThe container starts empty — download what you need.\nPlayer rates: xG90/xA90 gamma-Poisson shrunk to positional means, k in 90s: xG DEF 44/MID 6/FWD 16,\nxA DEF 8/MID 8/FWD 9. Movers k x1.4, rate x0.92. Minutes: stayers 0.71*prev+238, movers\n0.33*prev+755; weight recent starts heavily in-season. Use f(E[minutes]).\nOptimise: MILP over 6-8 gameweeks, decay 0.90, bench weights {0.03,0.21,0.06,0.002}, vice 0.10,\nitb 0.08 per 1m, hit cost 4, DEFAULT NO HITS. pulp + CBC, prune to ~300, run in background and poll.\nPresent expected points as a RANKING, never a forecast — measured over-prediction 9%.\nDO NOT REINTRODUCE (all tested worse): winner's-curse shrinkage (-13.5), minutes nodes instead of\nthe mean (-12.5), starts-share as gate definition (0), per-player home/away splits (-0.029\npersistence against +0.342 for the level).\nFIXTURE COEFFICIENTS, measured on 29,536 player-gameweeks over four seasons — use these when\nranking weeks or candidates: premium M/F 3.21 + 1.69*team_xG; 8-10m 1.99 + 1.88*xG; under 6m\n2.31 + 0.78*xG; DEF 5.42 - 1.63*xGA; GK 4.77 - 0.97*xGA. Fixtures matter MOST for defenders and\nmid-price attackers, LEAST for premium captains.\n\nPLAYER-LEVEL ODDS — ADDED 30 Aug, USE FOR THE CAPTAIN DECISION.\nTeam-level odds were always in the model; player props were not. Fetch anytime-goalscorer odds for\nthe captain candidates. Convert: lambda_goal = -ln(1 - P). Then DEDUCT THE OVERROUND: the sum of\nall players' P(scores) should equal the expected number of DISTINCT scorers, about 0.85 x team\nlambda, which works out to roughly 10-12% deflation. Sanity-check the deflated number by expressing\nit as the player's share of his team's goals and comparing with his share last season; if it is\nwildly different, your deflation is wrong, not the market. Portals (oddschecker, oddsportal) do not\nfetch programmatically — pull the numbers from secondary tips articles and name the bookmaker.\nFull player cards usually only go live 1-2 days before kickoff. VALUE NOT YET BACKTESTED.\n\nCAPTAINCY MUST BE EXPRESSED AS A BREAK-EVEN, NOT A POINT DIFFERENCE.\nA gap of \"+0.4 points\" says nothing about robustness. Decompose into channels (goals, assists,\nclean sheet, appearance), find the load-bearing parameter, and state how far it must move before\nthe decision flips. Worked example, GW2: Bruno beat Haaland by 0.84, and Bruno's assist share of\nMan Utd's goals had to fall from 34.8% to below 22.0% — a 37% relative decline — before Haaland\nwon; Haaland's goal share had to rise from 35.1% to above 46.1%, when even the raw undeflated\nodds implied only 38.4%. That is a robust call, and it was right: Bruno 23, Haaland 13.\n\n=== STEP 6 — TRANSFERS AND CHIPS ===\nTRANSFERS: default is to roll and take none. Tested: 1 transfer over 8 gameweeks scored 429,\n3 scored 408, 5 scored 396. Ibsen won last season with no transfer at all in 15 of 38 and zero hits.\nInsufficient information resolves to NO ACTION. Never take a hit without saying what it buys.\nBUT: \"the elite are rolling\" is not by itself an argument for us to roll. They roll from squads that\nalready contain the players we are missing. Check whether the transfer is CATCHING UP to consensus\nor DEVIATING from it. Catching up is not a deviation.\nWhere the model and the elite agree: high confidence. Where they disagree: SAY SO OUT LOUD and\nexplain which you are following and why. That disagreement is the interesting information.\nCHIPS — decide explicitly every run, do not just restate the plan:\n Current plan (set 1 expires at the GW19 deadline, Sat 2 Jan 13:30):\n Wildcard GW7 (17 Oct) building a PLAYING bench; Bench Boost GW9 (31 Oct);\n TRIPLE CAPTAIN GW16 (Sat 19 Dec) HAALAND AT HOME TO HULL, expected 2.90 team goals, the best\n single fixture of the first half, and no European football between 9 Dec and 19 Jan;\n FREE HIT GW19 (2 Jan), the densest round of the first half and the last day the chip exists.\n Set 2 (GW20-38): Wildcard ~GW31; Bench Boost on the confirmed double; Triple Captain on a second\n double, else Bruno v Hull home GW31 or Haaland at Hull GW27; Free Hit on the confirmed blank\n (likely GW33, FA Cup semi-finals 24 Apr). ONE CHIP PER GAMEWEEK.\n THE OVERRIDE THAT BEATS EVERYTHING: a DOUBLE GAMEWEEK. Measured over four seasons — a player whose\n team plays twice scores 6.10 vs 3.53, and hits 10+ points 20.2% of the time vs 5.8%. Eleven of\n them is +28 points. If a double appears anywhere, re-derive the whole chip plan around it and say\n so loudly. Check the fixtures endpoint for gameweeks with more or fewer than 10 matches EVERY RUN\n from December. As of 30 Aug 2026 all 38 gameweeks have exactly 10.\n Also flag any chip within 3 gameweeks of expiring unused.\n\n=== STEP 7 — DELIVER ===\nHTML artifact via SendUserFile: a numbered tap-list at the top with exactly what to do, then the XI,\ncaptain and vice with one line of reasoning each, then the elite-cohort section, then what is\nuncertain. State clearly that this is FINAL for this gameweek. He enters it manually — there is no\nwrite access to FPL.\nALWAYS set the recommended captain as VICE-CAPTAIN in the instructions when you are recommending a\ncaptain change. Measured 28 Aug: Arvind missed the captain change; because Bruno was vice he still\nbanked 23 at x1, and the loss was 10 points instead of 20. The vice slot is real insurance.\nSend a PUSH notification whose FIRST SENTENCE is the actions, not the analysis.\nThen update memory: /topics/football.md (squad, plan, rank), /areas/fpl-elite-kohort.md (cohort\nnumbers), /areas/fpl-chip-plan.md (chip decisions). Memory files are capped at 32KB — condense\nrather than append forever. Memory is the only thing that survives between runs.\nFinish by telling Arvind, in Norwegian local time, when the etterkjøring runs.\n\n=== ETTERKJØRINGSPROMPT — copy this verbatim into the trigger created in step 0 ===\nEtterkjøring for Arvinds FPL-lag \"arvindinho\", entry ID 5038450, 2026/27, etter runde {n}.\nSkriv på norsk. Bestem, ikke spør. Alle klokkeslett til Arvind i NORSK tid (UTC+2 til 25. okt 2026,\nderetter UTC+1). Les først minnefilene /topics/football.md, /areas/fpl-elite-kohort.md og\n/areas/fpl-chip-plan.md.\n\nOPPGAVE 1 — BYGG NESTE LEDD I KJEDEN. GJØR DETTE FØRST.\nHent https://fantasy.premierleague.com/api/bootstrap-static/, finn eventet med is_next og dets\ndeadline_time. Kall list_triggers. Opprett de to som mangler (ikke dupliser):\n a) navn \"FPL frist — GW{neste} ({dd.mm})\", run_once_at = deadline MINUS 2 timer, RFC3339 UTC,\n initiation human_schedule, notifications {push: true}, prompt EKSAKT disse to linjene:\n Hent prosedyren med: curl -s https://arvindfroi--e609966a9e0511f191bc1607ee4eb77e.web.val.run/\n Foelg den ordrett, fra steg 0. Faar du ikke tak i den, les minnefilene /topics/football.md,\n /areas/fpl-chip-plan.md og /areas/fpl-elite-kohort.md og gjennomfoer fristkjoeringen ut fra dem.\n b) navn \"FPL paaminnelse — GW{neste} ({dd.mm})\", run_once_at = deadline MINUS 45 minutter,\n initiation human_schedule, notifications {push: true}, prompt EKSAKT denne linjen:\n Kort fristpaaminnelse for FPL, entry 5038450. Les /topics/football.md, finn den siste\n GW-beslutningen som ble logget, og send EN push med KUN handlingene: bytte, kaptein, vise,\n benkerekkefolge, chip. Maks fem linjer, ingen analyse, ingen tall. Sjekk samtidig\n bootstrap-static for status og news paa de 15 og si fra hvis noe har endret seg siden\n fristkjoeringen. Ingen artefakt.\nSi i én linje at kjeden er bygget, med begge tidspunkter i norsk tid.\n\nOPPGAVE 2 — FASIT FOR RUNDE {n}\n/api/entry/5038450/event/{n}/picks/ og /history/ + ligastilling (liga 973274). MERK: entry_history\nkan henge etter; sannheten er summen av event_points x multiplier over de elleve, kryssjekket mot\nligaendepunktet. Rapporter poeng i runden, totalt mot 62,3 x spilte runder, overall rank og\nligaplassering. For hvert valg som slo feil: dårlig prosess eller dårlig flaks? Ingen bortforklaringer.\nVar kapteinsvalget riktig? Ble anbefalingen faktisk gjennomført? Hva kostet et eventuelt avvik?\n\nOPPGAVE 3 — ELITENS BLOKKBEVEGELSE (den egentlige edgen)\nHent de 953 IDene: curl -s https://arvindfroi--e609966a9e0511f191bc1607ee4eb77e.web.val.run/elite\nHent picks for runde {n} OG runde {n}-1 (~40 s med 28 tråder) og regn ut hva de byttet INN og UT som\nblokk — de tre største bevegelsene hver vei med størrelsen. Pluss eierandel mot globalt, effektivt\neierskap, vår nettoposisjon, og chipbruk. Én setning om hvilken vei konsensus går og om vi bør følge.\n\nOPPGAVE 4 — KALENDERVAKT\n/api/fixtures/: har noen runde fått flere eller færre enn 10 kamper? En dobbeltrunde slår alt annet\ni chip-planen (6,10 mot 3,53 poeng, 10+ i 20,2 % mot 5,8 %). Dukker en opp, SI DET HØYT og utled\nchip-planen på nytt. Flagg chips innen tre runder fra utløp; sett 1 dør ved GW19-fristen 2. jan 2027.\n\nOPPGAVE 5 — ÉN FORSKNINGSSAK, KUN ÉN\n1. Backtest verdien av spillerodds (anytime scorer) i kapteinsvalget — innført 30. aug, ikke målt.\n2. Verifiser odds-blandingen ende til ende i en spilt runde.\n3. Test eierskapsvektet risiko i målfunksjonen i stedet for ren forventning.\n4. Fra desember: følg Ben Crellins blank/dobbel-ark.\n\nPROSEDYREN ER SKRIVEBESKYTTET. Konkluderer du med at metoden bør endres, skriv FUNNET og FORSLAGET\ni minnet og si det til Arvind. Fakta i minnet, regler i valen.\n\nLEVERANSE: kort push hvis alt er som ventet. HTML-artefakt via SendUserFile når noe har endret seg\nvesentlig. Oppdater de tre minnefilene. Avslutt med når neste fristkjøring og påminnelse går, i norsk tid.\n","deleted",false,"children",[],[56],["D",1788769635681],[],[19],"8ed13451-5862-46bd-a8d3-7a75b934a6ca","intel.ts","df1dd5cc-aa95-11f1-9f2f-1607ee4eb77e",8,["D",1788769609512],0,"interval","import { blob } from \"https://esm.town/v/std/blob/global.ts\";\nimport { readFile } from \"https://esm.town/v/std/utils/index.ts\";\n\n// Kjorer daglig og forhaandsregner HELE eliteetterretningen, slik at\n// fristkjoeringen i Cowork slipper 1900 API-kall og all Python.\n// Resultatet serveres av main.ts paa /intel.\n\nconst API = \"https://fantasy.premierleague.com/api\";\nconst OUR = 5038450;\n\nasync function j(url: string, tries = 3): Promise {\n for (let i = 0; i < tries; i++) {\n try {\n const r = await fetch(url, { headers: { \"User-Agent\": \"Mozilla/5.0\" } });\n if (r.ok) return await r.json();\n } catch (_) { /* retry */ }\n await new Promise((s) => setTimeout(s, 400 * (i + 1)));\n }\n return null;\n}\n\nasync function pool(items: T[], n: number, fn: (t: T) => Promise): Promise {\n const out: R[] = new Array(items.length);\n let i = 0;\n await Promise.all(\n Array.from({ length: n }, async () => {\n while (true) {\n const k = i++;\n if (k >= items.length) return;\n out[k] = await fn(items[k]);\n }\n }),\n );\n return out;\n}\n\ntype Pick = { element: number; position: number; multiplier: number };\ntype Entry = { picks: Pick[]; active_chip: string | null } | null;\n\nexport default async function () {\n const t0 = Date.now();\n const boot = await j(`${API}/bootstrap-static/`);\n if (!boot) throw new Error(\"bootstrap failed\");\n\n const finished = boot.events.filter((e: any) => e.finished).map((e: any) => e.id);\n const gw = Math.max(...finished);\n const prev = gw - 1;\n\n const idsTxt = await readFile(\"/elite1000.txt\", import.meta.url);\n const ids: number[] = idsTxt.trim().split(\",\").map((x: string) => parseInt(x, 10));\n\n const live = await j(`${API}/event/${gw}/live/`);\n const pts: Record = {};\n for (const e of live.elements) pts[e.id] = e.stats.total_points;\n\n const name: Record = {};\n const team: Record = {};\n const cost: Record = {};\n const glob: Record = {};\n const short: Record = {};\n for (const t of boot.teams) short[t.id] = t.short_name;\n for (const p of boot.elements) {\n name[p.id] = p.web_name;\n team[p.id] = short[p.team];\n cost[p.id] = p.now_cost / 10;\n glob[p.id] = parseFloat(p.selected_by_percent);\n }\n\n const getPicks = (id: number, g: number) =>\n j(`${API}/entry/${id}/event/${g}/picks/`).then((d) =>\n d && d.picks ? { picks: d.picks, active_chip: d.active_chip } as Entry : null\n );\n\n const cur = await pool(ids, 20, (id) => getPicks(id, gw));\n const old = prev >= 1 ? await pool(ids, 20, (id) => getPicks(id, prev)) : ids.map(() => null);\n const ours = await getPicks(OUR, gw);\n\n const valid = cur.filter(Boolean) as Entry[];\n const n = valid.length;\n\n const own: Record = {};\n const eo: Record = {};\n const chips: Record = {};\n const scores: number[] = [];\n for (const d of valid) {\n const bb = d!.active_chip === \"bboost\";\n chips[d!.active_chip ?? \"ingen\"] = (chips[d!.active_chip ?? \"ingen\"] ?? 0) + 1;\n let s = 0;\n for (const p of d!.picks) {\n own[p.element] = (own[p.element] ?? 0) + 1;\n const m = bb && p.position > 11 ? 1 : p.multiplier;\n if (m > 0) {\n eo[p.element] = (eo[p.element] ?? 0) + m;\n s += (pts[p.element] ?? 0) * m;\n }\n }\n scores.push(s);\n }\n scores.sort((a, b) => a - b);\n const mean = scores.reduce((a, b) => a + b, 0) / n;\n\n // blokkbevegelse, med chip-splitten: bare lag uten chip og <=2 bytter\n const movAll: Record = {};\n const movClean: Record = {};\n let nClean = 0;\n for (let k = 0; k < ids.length; k++) {\n const a = old[k], b = cur[k];\n if (!a || !b) continue;\n const A = new Set(a.picks.map((p) => p.element));\n const B = new Set(b.picks.map((p) => p.element));\n let inn = 0;\n for (const e of B) if (!A.has(e)) inn++;\n for (const e of B) if (!A.has(e)) movAll[e] = (movAll[e] ?? 0) + 1;\n for (const e of A) if (!B.has(e)) movAll[e] = (movAll[e] ?? 0) - 1;\n const clean = !b.active_chip && inn <= 2;\n if (clean) {\n nClean++;\n for (const e of B) if (!A.has(e)) movClean[e] = (movClean[e] ?? 0) + 1;\n for (const e of A) if (!B.has(e)) movClean[e] = (movClean[e] ?? 0) - 1;\n }\n }\n\n const ourMult: Record = {};\n if (ours) for (const p of ours.picks) ourMult[p.element] = p.multiplier;\n\n const row = (e: number) => ({\n id: e, navn: name[e], lag: team[e], pris: cost[e],\n elite: +((own[e] ?? 0) / n * 100).toFixed(1),\n globalt: glob[e] ?? 0,\n diff: +(((own[e] ?? 0) / n * 100) - (glob[e] ?? 0)).toFixed(1),\n eo: +((eo[e] ?? 0) / n * 100).toFixed(1),\n var: (ourMult[e] ?? 0) * 100,\n netto: +(((ourMult[e] ?? 0) * 100) - ((eo[e] ?? 0) / n * 100)).toFixed(1),\n poeng: pts[e] ?? 0,\n });\n\n const eide = Object.keys(own).map(Number).filter((e) => own[e] / n >= 0.03);\n const out = {\n generert: new Date().toISOString(),\n gameweek: gw,\n n,\n kohort: {\n snitt: +mean.toFixed(1),\n median: scores[Math.floor(n / 2)],\n p10: scores[Math.floor(n / 10)],\n p90: scores[Math.floor(n * 0.9)],\n maks: scores[n - 1],\n },\n vaart_lag: ours\n ? ours.picks.filter((p) => p.multiplier > 0)\n .reduce((s, p) => s + (pts[p.element] ?? 0) * p.multiplier, 0)\n : null,\n chips: Object.fromEntries(\n Object.entries(chips).map(([k, v]) => [k, +(v / n * 100).toFixed(1)]),\n ),\n eierskap: eide.map(row).sort((a, b) => b.diff - a.diff),\n verste_netto: eide.map(row).sort((a, b) => a.netto - b.netto).slice(0, 10),\n blokk_ekte: {\n n: nClean,\n forklaring:\n \"Kun lag UTEN chip og med hoeyst 2 bytter. Raa tall er verdiloese naar en tredjedel wildcarder.\",\n inn: Object.keys(movClean).map(Number)\n .map((e) => ({ navn: name[e], lag: team[e], endring: +(movClean[e] / nClean * 100).toFixed(1), poeng: pts[e] ?? 0 }))\n .filter((r) => r.endring >= 2).sort((a, b) => b.endring - a.endring).slice(0, 8),\n ut: Object.keys(movClean).map(Number)\n .map((e) => ({ navn: name[e], lag: team[e], endring: +(movClean[e] / nClean * 100).toFixed(1), poeng: pts[e] ?? 0 }))\n .filter((r) => r.endring <= -2).sort((a, b) => a.endring - b.endring).slice(0, 8),\n },\n ms: Date.now() - t0,\n };\n await blob.setJSON(\"fpl_intel\", out);\n console.log(`GW${gw}: n=${n}, snitt ${out.kohort.snitt}, vaart ${out.vaart_lag}, ${out.ms} ms`);\n return out;\n}\n",[],[73],"14a2463b-f807-4143-aea5-df802d96e2cd","elite1000.txt","49a3747a-9e06-11f1-af55-1607ee4eb77e",6,["D",1787388526742],"41,129,260,470,526,616,763,783,974,1015,1387,1802,1975,2425,2693,2715,2784,2803,2944,3464,3551,3656,3892,4370,4869,4951,5133,5452,5851,6173,7212,7297,7626,7726,7830,7985,8203,8958,9267,9335,9705,10137,10483,10697,10912,11059,11381,11874,11961,12027,12225,13233,13470,13474,13536,14321,14707,14728,14919,15069,15150,15694,17925,17985,18197,18357,18601,19773,19783,19821,21020,22557,22886,23095,23630,23810,25176,25361,25384,26175,26222,26458,27002,27465,29024,29163,29914,30890,31515,31941,32387,32548,32740,32806,33058,33177,33697,33927,33946,34063,34498,35193,35525,35633,36535,36860,38862,39278,41737,41830,42987,43238,43607,43832,45724,45768,45791,47394,50059,50246,50251,50890,54516,55301,55411,55547,55967,56671,57729,58894,60108,61111,61225,61471,62925,65058,65270,65376,67045,67079,67984,69850,70466,71666,72633,74858,75593,75791,76051,77730,77812,78341,78877,80218,80935,81011,81910,82000,82578,83327,83632,83896,84305,84357,84573,86427,86872,87191,87648,88134,89376,89979,90889,91755,92708,94283,99649,101374,102228,102738,103645,104459,104708,105380,105545,105936,107752,108769,109098,110966,111989,113666,115141,116544,116555,116638,117375,117406,121336,121697,124015,124174,125875,126082,126797,127935,130435,130809,133190,135758,136052,137321,144474,144753,148351,150589,151232,158347,158476,158785,161158,165395,167514,168482,168845,169262,170764,173920,176960,178496,179693,180931,181408,184862,188422,189035,191153,192568,193540,196025,198010,198901,200066,201872,203045,203334,204429,209308,210742,212742,214705,215703,215873,217001,221552,222160,222925,223431,230471,233040,233191,239166,239319,241301,241700,241867,243666,244621,249209,250046,250292,250880,252948,254079,255770,257665,260103,267093,267302,277074,279550,285130,288827,291333,292518,293756,294707,296999,298942,299822,301671,302272,302782,303964,309649,310112,314736,315169,317375,321094,321815,326181,326398,328336,330306,335020,338209,340004,341560,343123,344209,346816,351356,351734,358871,359681,363364,367295,370700,372246,373949,377380,382991,385930,390744,392855,393119,393202,403613,403778,404989,411317,416623,417523,418646,422250,422362,431823,432210,433586,434601,437688,439318,439524,439718,447243,449180,452485,456499,460372,465609,466182,467115,468474,469314,474896,481315,485455,488854,500080,501834,506945,508138,517411,521986,522809,527176,534233,535067,537919,541387,549209,558737,560050,561953,564354,574452,575955,583236,583438,592509,595934,597461,606896,612199,614664,622290,643583,645622,649238,658844,662233,667464,676502,679476,682017,685229,686532,688776,690785,692948,699171,700917,702710,705240,709889,711297,712780,722355,723813,724355,725421,729061,731145,733474,738467,738643,742673,743943,744187,752931,753719,759055,761757,763401,768543,769852,772176,774413,779170,779876,783735,789296,795570,797831,799722,807107,810378,811907,815070,820065,826853,827985,831440,832123,836041,836802,865467,866668,869698,874191,875756,891826,897065,897779,910459,913326,914010,915581,917512,919442,924419,931420,934578,935633,936931,939773,964287,978252,979122,982200,983854,988518,994140,997651,1025359,1032829,1046954,1052107,1059303,1060489,1063469,1065141,1068226,1072585,1079697,1084748,1089355,1095622,1106300,1111150,1114062,1125188,1125973,1134303,1137093,1137816,1147906,1156668,1158196,1168527,1180499,1188908,1196210,1197476,1199060,1205230,1208068,1218015,1222770,1224050,1224170,1225290,1226093,1260381,1260736,1263403,1265289,1274909,1295882,1296034,1302013,1306308,1314327,1338860,1339185,1340709,1343973,1354519,1358070,1379330,1381294,1383668,1393060,1413902,1435755,1447047,1447987,1450184,1457867,1459097,1480174,1481649,1483931,1484090,1484458,1488898,1500872,1502042,1505635,1517792,1523479,1537656,1539010,1539402,1549076,1566284,1573935,1579967,1582687,1590773,1592856,1596883,1601583,1602664,1607483,1610062,1626884,1641067,1642275,1654474,1666196,1675762,1676274,1677122,1678282,1690423,1695522,1696869,1732992,1735487,1739619,1746821,1746825,1781151,1793267,1795843,1802049,1807691,1822754,1824989,1838812,1846200,1849247,1862595,1879403,1881787,1902471,1907367,1918439,1937973,1940430,1952805,1955586,1981516,1983128,1984517,1988371,1997828,2015950,2019993,2033744,2038266,2040455,2045345,2045363,2053487,2062829,2068322,2076855,2083055,2089750,2092894,2117759,2143980,2146620,2168452,2177435,2180149,2181317,2195181,2195211,2209891,2233303,2238584,2239392,2283092,2301160,2312247,2318429,2324454,2327271,2332766,2336562,2352637,2367796,2369305,2386562,2387433,2389112,2393556,2406306,2427965,2458940,2459509,2497238,2499216,2528519,2544559,2545190,2555668,2577913,2589327,2591009,2599785,2602508,2605391,2612330,2623218,2635031,2641938,2665137,2666380,2678920,2695137,2706218,2708998,2715503,2721819,2730159,2730860,2731285,2740583,2757592,2783496,2810656,2818750,2822277,2830434,2880512,2898685,2902522,2908787,2916537,2924494,2928205,2930767,2931354,2952390,2973604,2976163,2996795,3010012,3020825,3062032,3076055,3091470,3094518,3111473,3113984,3119815,3138145,3141870,3172291,3179924,3184650,3186917,3193296,3215014,3233387,3242089,3260107,3276297,3337704,3346744,3366832,3369408,3376610,3393419,3444571,3465483,3492167,3494514,3495950,3502338,3504693,3509883,3550196,3552003,3552180,3584008,3605671,3607308,3622423,3626277,3636685,3639437,3650728,3653133,3671850,3675466,3709161,3709403,3712703,3718709,3733125,3745562,3746328,3803948,3850148,3855643,3864538,3942983,3949920,3987124,4004485,4008329,4009983,4025026,4026790,4092588,4095679,4125351,4166206,4195778,4208695,4252511,4282145,4283934,4285089,4285905,4308459,4318382,4330223,4339037,4339766,4345147,4345375,4356541,4358955,4373707,4374884,4435968,4440772,4460298,4470006,4477713,4482552,4508867,4536166,4540939,4541351,4546016,4548400,4566674,4566948,4585590,4625545,4632146,4633080,4656090,4692195,4710790,4734036,4737689,4832503,4837365,4838034,4858316,4897117,4939744,4959161,4984814,5004635,5047233,5051724,5074304,5095552,5096011,5097696,5137156,5140836,5224113,5231522,5281115,5323231,5348821,5357183,5360167,5405282,5414323,5415211,5435472,5480469,5482360,5485919,5488002,5500332,5532629,5558900,5588856,5592619,5604567,5611641,5631772,5690449,5704621,5706212,5757198,5820044,5834548,5846785,5869623,5877334,5907261,5916628,5919648,5925418,5934313,5938997,5969883,6005517,6006798,6011517,6098389,6113962,6114933,6186091,6202464,6235374,6237145,6263756,6267927,6332239,6390328,6392684,6394148,6403033,6424511,6438721,6458491,6473591,6482623,6484820,6586154,6601630,6614994,6646488,6652072,6705969,6770708,6812748,6839933,6851846,6868768,6885955,6920156,6968883,6977041,6987911,7005960,7041641,7092692,7123699,7150301,7180058,7198307,7207298,7222274,7317163,7495405,7513010,7566785,7594781,7614782,7620075,7636617,7654439,7662306,7751703,7830544,7887944,7888828,7923070,7923885,7959808,8004600,8008797,8024839,8100258,8131655,8275808,8287140,8383500,8471936,8476726,8494101,8531596,8793704,8828442,8834462,8851729,8852299",[],[83],"1f033a84-d46e-4c1e-b78f-9ff08539d96c",["D",1787388359286],"deletedAt","legacyEnvVars","privacy","public","httpPrivacy","description","Serves the FPL deadline procedure as plain text","imageUrl","userId","019c8c94-1b18-70ab-9d66-67b2531c59ef","pinnedFileId","pinPreview","pinMigrated","screenshotVersion","likeCount","orgId","019c8c94-1b18-70ab-9d66-68fabd795643","tursoName","tursoHostname","tursoMeasuredAt","pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_11":12,"_143":139},"org",{"_11":12,"_137":-5,"_138":139,"_100":101,"_140":141,"_142":65},"branch",{"_17":58,"_13":16,"_33":36,"_129":90,"_130":-5,"_131":-5,"_132":-5,"_133":-5,"_134":-5,"_135":14,"_136":12},"isOwned","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","displayName","avatar","https://img.clerk.com/eyJ0eXBlIjoiZGVmYXVsdCIsImlpZCI6Imluc18yTm10cGhQUklwU0tqZHNiOHNnNGd2ZjJURFUiLCJyaWQiOiJ1c2VyXzNBNWFrb3FkY3pXQTkyNk5oeWFEa25zSkhBayJ9","tier","free","canRestrictVals","profileImageUrl","adoptOrgId",{"_33":90,"_13":14,"_45":187,"_92":-5,"_93":65,"_94":95,"_96":95,"_97":98,"_99":-5,"_100":101,"_102":38,"_103":60,"_104":-5,"_105":-5,"_106":77,"_107":108,"_109":-5,"_110":-5,"_111":-5,"_112":-5,"_113":114,"_115":114,"_116":117,"_118":117,"_119":117,"_120":114,"_121":114,"_122":188,"_124":189,"_126":190,"_128":65},[175,176,177,178],"files",[175,176,177,178],"counts",{"_165":77,"_166":167,"_168":77,"_32":169,"_170":171},[],{"_33":34,"_35":36,"_37":38,"_13":19,"_39":40,"_41":42,"_17":43,"_44":-5,"_45":163,"_47":48,"_49":-5,"_18":164},"upstreamProject","upstreamBranch","emptyEnvVars","flags",[],"hasHttpFiles","grants","orgs","customDomains",{},["D",1788769635681],[19],"pulls","branches",1,"forks",{"_172":77,"_174":77},"connections",{"_172":77,"_173":77},"count","unfilledCount","emptyCount",{"_33":82,"_13":83,"_37":84,"_44":-5,"_35":36,"_47":85,"_59":60,"_45":86,"_17":77,"_41":24,"_39":87,"_64":65,"_49":-5,"_66":185,"_18":186},{"_33":72,"_13":73,"_37":74,"_44":-5,"_35":36,"_47":75,"_59":60,"_45":76,"_17":77,"_41":78,"_39":79,"_64":65,"_49":-5,"_66":183,"_18":184},{"_33":34,"_13":19,"_37":38,"_44":-5,"_35":36,"_47":48,"_59":60,"_45":69,"_17":43,"_41":42,"_39":40,"_64":65,"_49":-5,"_66":181,"_18":182},{"_33":55,"_13":56,"_37":57,"_44":-5,"_35":36,"_47":58,"_59":60,"_45":61,"_17":62,"_41":24,"_39":63,"_64":65,"_49":-5,"_66":179,"_18":180},[],[56],[],[19],[],[73],[],[83],["D",1787388359286],{"_11":12,"_143":139},{"_11":12,"_137":-5,"_138":139,"_100":101,"_140":141,"_142":65},{"_17":58,"_13":16,"_33":36,"_129":90,"_130":-5,"_131":-5,"_132":-5,"_133":-5,"_134":-5,"_135":14,"_136":12},"ctx",{"_122":-5,"_160":224,"_225":226,"_227":228,"_229":65,"_230":65,"_231":65,"_232":60,"_233":-5,"_234":-7,"_235":-7,"_236":77,"_237":65,"_238":65,"_239":-5,"_240":65,"_241":65,"_242":65,"_243":-5},"ENV",{"_195":196,"_197":198,"_199":16,"_200":201,"_202":203,"_204":205,"_206":207,"_208":209,"_210":211,"_212":213,"_214":215,"_216":217,"_218":219,"_220":221,"_222":223},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","e87eb04e707cf64e7e532c8e8ec395f209e17f15","RENDER_GIT_BRANCH","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","CLOUDFLARE_TURNSTILE_SITE_KEY","0x4AAAAAACtaIZJH7s9rBMZ6","GOOGLE_CLIENT_ID","875989696899-jtsqkhlbdv5jrlsk8631p7ipjg3c6o62.apps.googleusercontent.com","STRIPE_PUBLISHABLE_KEY","pk_live_51N2zvWEpZTTLipIkVEwXrUG9AT0yeOOEYxszarZT1jpsZV58Ly6wlg3SltZpHN26CUalOBmUSY7adQRQiSkCZz2500qSXNkf8Y","API_URL","https://api.val.town",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]