# 04 — Recruiting: Tables, Commitment, Boards, Portal, Pipelines Documentation suite for **CoachCarouselRecruitTool**. Synthesized 2026-07-14. Source tags as in doc 01 (`[api] [fc] [pipe] [jersey] [carousel] [web] [schema] [dump] [diff]`). UNVERIFIED = not yet proven on a live save / write test. --- ## Table of contents 1. [Data-model overview](#1-data-model-overview) 2. [The Recruit table](#2-the-recruit-table) 3. [Recruit-linked Player fields & position classification](#3-recruit-linked-player-fields--position-classification) 4. [Top schools: ProspectTargetSchool](#4-top-schools-prospecttargetschool) 5. [How commitment is actually represented](#5-how-commitment-is-actually-represented) 6. [Recruiting boards: RecruitingBoard → RecruitTarget / UserRecruitTarget](#6-recruiting-boards) 7. [Scholarships, NIL, and hard limits](#7-scholarships-nil-and-hard-limits) 8. [Motivations, dealbreakers, pitches](#8-motivations-dealbreakers-pitches) 9. [MySchoolTrackingTable — the school grades recruits check](#9-myschooltrackingtable) 10. [Pipelines](#10-pipelines) 11. [The transfer portal](#11-the-transfer-portal) 12. [The force-commit mutation recipe (proven writes)](#12-the-force-commit-mutation-recipe) 13. [Write-survival semantics (community-tested)](#13-write-survival-semantics) 14. [Gotchas checklist](#14-gotchas-checklist) 15. [Sources](#sources) --- ## 1. Data-model overview Reference graph (all edges schema-verified; `[schema §12]`, live-verified where noted): ``` Franchise ─┬─ RecruitingClassPlayers : Recruit[] ← the HS/JUCO class └─ Transfers : Recruit[] ← THE TRANSFER PORTAL master list Recruit ─┬─ Player → Player row (recruit prospects share the Player table; TeamIndex = 255) └─ TopSchoolsList → ProspectTargetSchool[] row (10 slots) └─ each slot → ProspectTargetSchool { TeamId (=Team.TeamIndex), TeamInfluence } Team ─┬─ RecruitingBoard → RecruitingBoard row ─ Recruits → RecruitTarget[] row (35 slots) │ └─ slots → RecruitTarget rows (AI schools) or UserRecruitTarget rows (the human school) ├─ CommittedPlayers : Player[] ├─ MySchoolTrackingTable → grades (CoachPrestigeGrade, CoachStabilityGrade, …) └─ SchoolPipelineInfluenceList → SchoolPipelineInfluence[] → 10 pipeline rows RecruitTarget ─ Recruit → Recruit (board rows point AT recruits; not vice versa) ProspectTargetSchool.TeamId ──(int join)── Team.TeamIndex (NOT the Team row index) ``` ## 2. The Recruit table `Recruit`, uniqueId **1873209313**, tableId 4269, cap 7600, 4101 non-empty (row 0 = `Invalid` sentinel — skip it). Unique by name. `[dump §1,§2][diff §3]` ### 2.1 Fields (15; bits from live offset table) | Field | Type | Bits | Range | Notes | |---|---|---|---|---| | `TopSchoolsList` | ref → ProspectTargetSchool[] | 32 | | offset 0 (⇒ destroyed on freed rows) | | `Player` | ref → Player | 32 | | validate `ref.tableId === playerT.header.tableId` | | `RecruitStage` | enum RecruitStage | 4 | | §2.2 | | `RecruitStageAdvance` | enum RecruitStageAdvance | 3 | | §2.3 | | `Class` | enum RecruitingClass | 4 | | §2.4 | | `CommitScore` | int | 10 | 0–1023 | commitment threshold; same scale as TeamInfluence | | `NationalRank` | int | 13 | 0–4500 | 1 = best | | `PositionRank` / `StateRank` | int | 12 | 0–4000 | | | `ProductionGrade` | int | 7 | 0–127 | obs ~84–92 | | `TotalScholarshipOffers` | int | 6 | 0–63 | NOT a lifetime counter (Signed recruits can read 0) | | `QualityModifier` | enum GemBust | 3 | | NORMAL=0, GEM=1, BUST=2, HIDDEN=3, Invalid=4 | | `AlternatePosition1/2` | enum DraftPositionE | 5 | | def Invalid_ | | `SurnameAudioID` | int | 15 | 0–32767 | | `[dump §2][schema §3]` Real rows (QQ): row 512 = `{Top10, HighSchool, NationalRank 1821, CommitScore 519, QualityModifier BUST, TopSchoolsList → PTS[]#1820, Player → Player#8762}`. `[dumps/QQ/Recruit sample]` ### 2.2 RecruitStage enum (4 bits) — the commitment state machine Raw-value map (formatted names; aliases share raws): `[dump §2]` ``` First_/Top10 = 0 Top5 = 1 Top3 = 2 Battle = 3 SoftCommitted = 4 HardCommitted = 5 Last_/Signed = 6 Count_ = 7 Invalid = 8 ``` - **There is NO 'uncommitted/None' literal — `Top10` (raw 0) IS the uncommitted base state.** Every recruit in a fresh class reads `Top10`. A decommit writer sets the stage back to `'Top10'`/`'Top5'`/`'Top3'` (Top3 keeps the old school prominently re-recruitable). `[dump §2,§14]` - Live distribution at the portal stage (QQ2): Top10 1965, Top5 363, Top3 154, Battle 29, SoftCommitted 27, HardCommitted 5, Signed 1557, Invalid 1. `[dumps/QQ2/scan.json]` - Stage progression thresholds (leader TeamInfluence as % of CommitScore): Top5 at 50, Top3 at 75, Commit at 100 (`RecruitingTunables.PointsPercentageThresholdCommit/Top3/Top5` = 100/75/50) `[schema §3]`. (The brooksg357 docs quote Top5 at 35% `[web §5a, 1SRC]` — the schema default is 50; trust the save's tunables.) - Stage school-count caps: Top10=10, Top5=5, Top3=3, Battle=2, SoftCommitted=1, HardCommitted=0, Signed=1. `[schema §3]` - **[1SRC]** `RecruitStage` is a *ratchet* — the engine never recomputes it downward; whether a manual downward write sticks in-game is UNVERIFIED (top open question for our decommit feature). `[web §5a,§9]` ### 2.3 RecruitStageAdvance enum (3 bits) — one-shot engine commands ``` First_/None = 0 Advance = 1 Decommit = 2 InstantCommit = 3 (Last_) Count_ = 4 Invalid = 5 ``` - **`Decommit` (2) exists as a first-class engine action.** Live resting values: `Invalid` on ~99% of rows (default parked state), a few `Advance`/`None` linger — the field is a one-shot command consumed on week advance, then left at Invalid/None. Writing `'Decommit'` on a committed recruit is the top candidate for a native decommit (behavior UNVERIFIED — needs a scratch-save write test). `[dump §2,§14][schema §10]` - force-commit writes `'InstantCommit'` here as part of its proven commit recipe (§12). `[fc §4]` ### 2.4 RecruitingClass enum (4 bits) ``` HighSchool = 0 JuniorCollege_Sophomore = 1 JuniorCollege_Junior = 2 JuniorCollege_Senior = 3 Transfer_Freshman = 4 Transfer_Sophomore = 5 Transfer_Junior = 6 Transfer_Senior = 7 Count_ = 8 Invalid = 9 ``` Observed (QQ2): HighSchool 3740, JUCO 219/141/1, **zero Transfer_\*** at portal-stage week 1 — see §11. Schema default is `Transfer_Junior`, so a blank row is NOT a valid HS recruit (set Class + RecruitStage + StageAdvance when minting rows). `[dump §2][schema §14]` ## 3. Recruit-linked Player fields & position classification Recruits live in the shared `Player` table (uid 1612938518; 16,257/16,500 rows; 282 live fields). A recruit's Player row has `TeamIndex = 255`. `[dump §9]` ### 3.1 Recruiting-relevant Player fields | Field | Type / bits | Live notes | Source | |---|---|---|---| | `Position` | enum PositionE 6b | depth-chart names (ROLB etc.) DO appear on real rows | `[dump §9]` | | `ProspectStarRating` | enum ProspectQuality 3b {ONE_STAR=0 … FIVE_STAR=4, Invalid=6} | | `[schema §10]` | | `SchoolYear` | enum 3b {Freshman=0, Sophomore=1, Junior=2, Senior=3, Invalid_=5} | | `[jersey §1][schema §9]` | | `RedshirtStatus` | enum 2b {Eligible=0, Ineligible=1, Current=2, Previous=3} | | `[jersey §1]` | | `RecruitingDealbreaker` | enum RecruitingMotivationType 4b | live: ProximityToHome, CoachPrestige, PlayingStyle, BrandExposure, ChampionshipContender, ProPotential; `Invalid` on veterans | `[dump §9]` | | `IdealRecruitingPitch` | enum RecruitingPitchType 5b | live: Grassroots, Aspirational, HometownHero, TimeToGetToWork; `Invalid` on veterans | `[dump §9]` | | `Motivation1/2/3` | enum MotivationType 4b | **reads back as raw `'0000'` on every sampled row** (stored value matches no member) — handle raw-bits fallback | `[dump §9,§15]` | | `HomePipeline` | enum Pipeline 6b | state names + sub-state regions (Tidewater, SouthFlorida) + Invalid | `[dump §9]` | | `PLYR_HOME_STATE` | enum StateName 6b (Alabama=0…Wyoming=49 alphabetical, NonUS=50, INVALID=51) | | `[schema §9,§10]` | | `IsNIL` | bool | real-likeness flag; NIL rows can carry odd data (Age 0) | `[jersey §1][dump §9]` | | `BaseNILValue` | s_int 11b −255–1023 | can be negative (obs −20) | `[schema §9][dump §9]` | | `CurrentNILCompensation` | int 10b | | `[schema §9]` | | `AbsoluteTransferChance` | int 7b −1–100, def −1 | UNVERIFIED: transfer-out propensity lever | `[schema §9][web §6]` | | `PLYR_PREVTEAMID` | int 11b 0–2047 | | `[schema §9]` | | `TraitDevelopment` | enum 3b {Normal=0, College_Impact=1, College_Star=2, College_Elite=3, Hidden=4, Invalid_=5} | | `[schema §9]` | | `OverallRating` | int 7b 0–100 | **calculated, not authoritative** (stored 80 vs displayed 84 example) `[web §5b, 1SRC]` | | | 52 `*Rating` fields | int | exact list enumerated in RO27's settings artifact | `[web §3, CODE]` | ### 3.2 Offense/defense classification (for the OC-change/DC-change decommit rule) PositionE (6-bit) full domain `[jersey §5][schema §10]`: ``` OFFENSE (0–9): QB=0 HB=1 FB=2 WR=3 TE=4 LT=5 LG=6 C=7 RG=8 RT=9 DEFENSE (10–18): LE=10 RE=11 DT=12 LOLB=13 MLB=14 ROLB=15 CB=16 FS=17 SS=18 SPECIAL (19–21): K=19 P=20 LS=21 PSEUDO (22–34): KR PR KOS 3DRB GAD PWHB SLWR RLE RRE RDT NT SUBLB SLCB ← classify explicitly if seen COACH (35–38): HC_CFM OC_CFM DC_CFM Owner_CFM (CFM contexts only — NOT Coach.Position) Invalid_ = 63 ``` - Numeric rule: `0–9` offense, `10–18` defense, `19–21` special teams. Depth-chart-style names (ROLB) appear on real Player rows, so build the classifier over the full name set, not just "QB/RB/WR". `[dump §9]` - The jersey tool's name sets (verbatim): OFFENSE `QB HB FB WR TE LT LG C RG RT`, DEFENSE `LE RE DT LOLB MLB ROLB CB FS SS`. `[jersey §2]` - force-commit's position→team-rating-group map (`POSITION_TO_RATING_GROUP`) is another ready mapping. `[fc §10]` ## 4. Top schools: ProspectTargetSchool - Chain: `Recruit.TopSchoolsList` → `ProspectTargetSchool[]` row (uid 2332540366; fields `ProspectTargetSchool0..9`, 10 slots) → `ProspectTargetSchool` rows (pool A uid 3789266353), each exactly two fields: **`TeamId: int 11b 0–2047` (= Team.TeamIndex, NOT the Team row) and `TeamInfluence: int 16b 0–65535`** (schema widths; live values use the 0–1023 CommitScore scale). `[dump §3][schema §3]` - 41,000 of 41,010 pool-A rows used = 4,100 recruits × 10 — **every recruit always has 10 populated slots**. A second pool (uid 3843719174, cap 35,000) exists; no sampled slot pointed into it (purpose UNVERIFIED). Validate slot refs against the resolved table's tableId. `[dump §3][diff §3]` - **Slots are sorted by TeamInfluence descending; slot 0 = current leader / committed school.** `[dump §3]` - Real examples `[dumps/QQ2/commit-analysis.json]`: - Signed CB Jordan Jacobs (Recruit row 5): CommitScore 431; slot0 = TeamId 16 "California" influence **561**; slots 1–9 all influence 0 (zeroed once signed). - Uncommitted Top10 WR Uriah Rogers (row 7): CommitScore 459; influences 67, 61, 61, 56, 55, 54, 54, 54, 52, 51 (dense, sorted). - Signed WR Pete Sails IV: CommitScore 771, slot0 TeamId 4 (Arizona State) influence 1000. `[diff §3]` - **[1SRC]** UI reads physical slot order, not influence order — after any edit **re-permute the 10 slots to influence-descending** or the UI misrenders. `[web §5a]` ## 5. How commitment is actually represented Empirically verified binding (task-critical): `[dump §4][diff §3]` > A recruit is committed to school S iff `RecruitStage ∈ {SoftCommitted, HardCommitted, Signed}` **and** > `TopSchoolsList` slot 0's `TeamId == S.TeamIndex`. Supporting evidence and non-indicators: - Every sampled Signed/HardCommitted recruit's slot-0 school matches; Signed recruits show `slot0.TeamInfluence ≥ CommitScore` (561 ≥ 431; 1000 ≥ 771). `[dump §4][diff §3]` - **Board entries are NOT authoritative**: a SoftCommitted recruit had zero RecruitTarget rows anywhere; HardCommitted recruits still sat on 2 different schools' boards; the winning school's board entry can read `ScholarshipStatus='None'`; `CommittedWeekNumber` is 0 on most committed entries. `[dump §4]` - **Signed ≠ rostered**: all 400 checked Signed recruits' `Player.TeamIndex` = 255 — roster assignment happens at season rollover. `[dump §4]` - `TotalScholarshipOffers` is not a lifetime counter (Signed recruit with 0). `[dump §4]` - A second, denormalized commit signal exists: `Team.CommittedPlayers: Player[]` — keep it in sync when decommitting (UNVERIFIED whether the engine rebuilds it weekly). `[schema §3]` - By the portal stage most of the signed class's Recruit rows are already **freed** (nextRecordToUse 5064 → 488); the 106 remaining Signed rows still carry full data. `[diff §3]` ## 6. Recruiting boards Chain (live-verified): `Team.RecruitingBoard → RecruitingBoard row → .Recruits → RecruitTarget[] row → RecruitTarget0..34 → RecruitTarget rows`. `[dump §4]` - **RecruitingBoard** (uid 220276943, 138 rows): `Recruits: ref`, `RecruitingHoursTotal/Assigned/Processed` int 12b 0–4095 (350 total at preseason). `[dump §4][schema §3]` - **RecruitTarget[]** (uid 2412159097): 138 rows × **35 slots**. Row index empirically equals the team's TeamIndex, but that's an allocation coincidence — **follow `Team.RecruitingBoard`** (the schema-sanctioned route). At preseason every row has `arraySize=0` — **boards are wiped between cycles**. `[dump §4][schema §3]` - **RecruitTarget** (uid 59043175, cap 4870; 4795 live mid-cycle, 0 otherwise) — 20 fields: | Field | Type / bits | Notes | |---|---|---| | `Recruit` | ref Recruit | empty/dangling ⇒ slot reusable ("hijackable") `[fc §3.4]` | | `ProspectInfluenceTotal` | int 10b 0–1023 | **the team-side MASTER influence store** `[web §5a, 1SRC]`; matches slot-0 TeamInfluence for winners (561/561) `[dump commit-analysis]` | | `ProspectInfluenceTotalLastWeek` | int 10b | | | `ProspectInfluenceDelta` | s_int 11b −200–1023 | can be negative | | `ProspectHoursSpentCurrent` | int 7b | | | `ScholarshipStatus` | enum 3b | §7.1 | | `CommittedWeekNumber` | int 5b 0–31 | usually 0 even for committed entries | | `CurrentNILOffer` / `NILExpectation` | int 10b 0–1023 | | | `OriginalNILExpectation` | int 9b 0–510 | | | `CurrentScholarshipBonus` | s_int 8b **−200–50** | negative legal | | `SearchSocialMedia` / `ContactHighSchoolCoaches` / `ContactFriendsAndFamily` / `SendTheHouse` / `VisitRecruitsSchool` | bool | weekly action flags | | `ActivePitches` | ref ActiveRecruitingPitch[] | pool can be exhausted `[web §5a, 1SRC]` | | `SwayPitch` | enum RecruitingPitchType 5b | | | `ScheduledVisit` | ref ActiveVisitInfo | | | `UnlockedIntelBitfield` | int 14b | maps to the 14 motivations | `[dump §4][schema §3]` - **UserRecruitTarget** (uid 3987156317, cap 1120 = 32 users × 35) — subclass: all 20 fields **plus** `IsFavorite: bool`, `RecruitingFeedback` + `ImmediateRecruitingFeedback` (refs → RecruitingActionFeedbackEntry[]). Confirmed live: the user school's RecruitTarget[] row (row 43 = LSU) holds 31 refs into UserRecruitTarget. This is why force-commit flags boards with foreign-table refs as "user school". Structurally editable like RecruitTarget (write test UNVERIFIED). `[dump §4][schema §3][fc §3.4]` - Real linked rows `[dumps/QQ2/commit-analysis.json]`: Signed recruit row 5 ← RecruitTarget row 4146 on board 16 (California): `{ScholarshipStatus: None, CommittedWeekNumber: 0, ProspectInfluenceTotal: 561, CurrentNILOffer: 0, NILExpectation: 55}`; Signed row 11 ← RT row 854 on board 4 (Arizona State): `{Offered, CommittedWeekNumber: 13, ProspectInfluenceTotal: 1000, CurrentNILOffer: 200}`. ## 7. Scholarships, NIL, and hard limits ### 7.1 ScholarshipStatus enum (3 bits) ``` First_/None = 0 Revoked = 1 New = 2 Offered = 3 Committed = 4 (Last_) Count_ = 5 Invalid = 6 ``` Live: AI boards show only None (2031) / Offered (2764); **`Committed` was never observed**, even on Signed recruits' winning entries — commit truth lives on the Recruit side (§5). `Revoked`/`New` unobserved. `[dump §4]` ### 7.2 Limits & tunables (defaults; league-editable in save) | Limit | Value | Source | |---|---|---| | Board size (`MaxRecruitingBoardTargets`, RecruitTarget[] slots) | **35** | `[web §5a][dump §4]` | | `MaxTeamScholarshipOffers` | 35 | `[web §5a, 1SRC]` | | `Recruit.TotalScholarshipOffers` field cap | 63 (6 bits) | `[schema §3]` | | Recruits generated per class | ~4,100 + 1 sentinel (`RecruitsToGenerateMin/Max` 4100) | `[web §5a][dump §1]` | | `MaxFiveStarsInClass` | 32 | `[web §5a, 1SRC]` | | Recruiting hours | weekly flow, `RecruitingBoard.RecruitingHoursTotal` (obs 350); refill by prestige × phase splines, 50 h/recruit/week cap **[1SRC]** | `[dump §4][web §5a]` | | Program points | seasonal stock: `Team.ProgramPointBudget` 0–30000 (obs 1250), `RemainingProgramPoints` −6000–30000, spent buckets (`NILProgramPointsSpent` obs 575) | `[schema §6][dump Team sample]` | | Commit thresholds | `PointsPercentageThresholdCommit`=100 / Top3=75 / Top5=50 of CommitScore; `HardCommitAdditionalCommitScorePercentage`, `BattleAdditionalCommitScorePercentage` | `[schema §3]` | | Instant commit | `InstantCommitOddsPerStarLevel: int[]`, `InstantCommitBonusPrestige: int[]` | `[schema §3]` | | CPU AI | `AIAggressivenessSlider` 1.15, `AINILAggressivenessSlider` 2.25, `CPU_AI_MaxPitchCount` 2 **[1SRC]** | `[web §5a]` | | League toggle | `LeagueSetting.IsRecruitFlippingEnabled: bool` (affects flipping committed recruits — relevant to our decommit/re-recruit design) | `[dump §12]` | ## 8. Motivations, dealbreakers, pitches ### 8.1 RecruitingMotivationType (4 bits) — the 14 recruiting motivations / dealbreakers ``` AcademicPrestige=0 AthleticFacilities=1 BrandExposure=2 CampusLifestyle=3 ChampionshipContender=4 CoachPrestige=5 CoachStability=6 ConferencePrestige=7 PlayingStyle=8 PlayingTime=9 ProPotential=10 ProgramTradition=11 ProximityToHome=12 StadiumAtmosphere=13 (Last_) Invalid=15 ``` `[schema §10]` — **CoachPrestige(5) and CoachStability(6) are the natural priority targets for carousel-driven decommits.** `Player.RecruitingDealbreaker` uses this enum (live values confirmed `[dump §9]`). **[1SRC]** a dealbreaker "fails" when the school's grade in that motivation drops below B−; only 8 of the 14 can be dealbreakers (excluded: AcademicPrestige, AthleticFacilities, CoachStability, PlayingStyle, ProgramTradition, StadiumAtmosphere). `[web §5a]` (Note: live saves DO show CoachPrestige, ProximityToHome, PlayingStyle... as dealbreaker values `[dump §9]` — the PlayingStyle sighting contradicts the 1SRC exclusion list; trust live data.) ### 8.2 MotivationType (Player.Motivation1–3) — a DIFFERENT enum Pro/Madden-style members (None=1, NoIncomeTax=2, WarmWeatherState=3, BigMarket=4, ChampionshipContender=5, TeamPrestige=6, SchemeFit=7, ToptheDepthChart=8, TeamHasFranchiseQB=9, MentoratPosition=10, HeadCoachHistoricRecord=11, CloseToHome=12). Live reads return raw `'0000'` (no member for stored 0) — any motivation logic must handle the raw-bits fallback. Do not conflate with RecruitingMotivationType. `[schema §10][dump §9]` ### 8.3 RecruitingPitchType (5 bits) — 20 real pitches ``` CollegeExperience=0 TeamPlayer=1 CampusPersonality=2 ItsGameTime=3 Prestigious=4 StudentOfTheGame=5 HometownHero=6 ProveYourself=7 TheClutch=8 TVTime=9 CoachsFavorite=10 Aspirational=11 ToTheHouse=12 FootballInfluencer=13 TimeToGetToWork=14 Starter=15 Grassroots=16 ConferenceSpotlight=17 SundayBound=18 WorkHorse=19 (Last_) Count_=20 Invalid=22 ``` `[schema §10]` Pitch intensity: `RecruitingActionIntensity` {SoftSell=0, HardSell=1, Sway=2}. `[schema §10]` force-commit's `DEALBREAKER_TO_PITCH` map (10-pitch subset) pairs each dealbreaker with a pitch — e.g. CoachStability→TheClutch, ChampionshipContender→TVTime, ProximityToHome→ConferenceSpotlight. `[fc §10]` ### 8.4 Visits & scouting (context) `ActiveVisitInfo {Activity: VisitActivityType(14 values), WeekNumber, WeekType}`; `MaxRecruitVisitsPerWeek=4` **[1SRC]**; scouting = `ProspectInteraction` rows (max 5 scouts/recruit; 14-bit intel bitfield maps to the 14 motivations; reveal display driven by `TimesScouted` **[1SRC]**). `[schema §10][web §5a]` ## 9. MySchoolTrackingTable uid 349376083, 138 rows × 36 fields — per-school data behind the motivation grades recruits check. `[dump §8]` - Ten motivation grades (LetterGrade): `AcademicPrestigeGrade, AthleticFacilitiesGrade, BrandExposureGrade, CampusLifestyleGrade, ChampionshipContenderGrade, `**`CoachPrestigeGrade, CoachStabilityGrade`**`, ConferencePrestigeGrade, ProgramTraditionGrade, StadiumAtmosphereGrade` + positional `ProPotentialGradeQB/RB/WR/TE/OL/DL/LB/DB/K/P`. - Raw stats: `AthleticFacilitiesScore` 0–2000 (obs 510), `CampusLifestyleScore`, BrandExposure TV/streaming/game-of-week counters, `ChampionshipContenderCurrentYearRank` (+Yr+1..3), PlayingStyle grade/rank/stat ref tables. - **`CoachStabilityGrade` / `CoachPrestigeGrade` are the game's own post-carousel recruiting inputs** — example row 0: CoachStabilityGrade=F, CoachPrestigeGrade=D+. Our tool can read (or adjust) these to align its decommit odds with what the game itself signals. `[dump §8]` - Reached only via `Team.MySchoolTrackingTable` ref (row alignment with Team rows is NOT guaranteed — FIU team row 35 → MySchool row 30). `[dump §7]` - Grade→dealbreaker mapping keys (force-commit's `GRADE_TO_DEALBREAKER`): the ten `*Grade` columns map 1:1 onto RecruitingMotivationType members. `[fc §3.9]` ## 10. Pipelines - **Pipeline enum** (6 bits, 43 real regions + Invalid=44): Alabama=0, Arizona=1, Arkansas=2, BigApple=3, BigSky=4, CentralFlorida=5, Colorado=6, EastTexas=7, Hawaii=8, Illinois=9, Indiana=10, Iowa=11, Kansas=12, Kentucky=13, Louisiana=14, MetroAtlanta=15, Michigan=16, Minnesota=17, Mississippi=18, Missouri=19, Nebraska=20, Nevada=21, NewEngland=22, NewMexico=23, NorthCarolina=24, NorthFlorida=25, NorthTexas=26, NorthernCalifornia=27, Ohio=28, Oklahoma=29, PacificNorthwest=30, Pennsylvania=31, SouthCarolina=32, SouthFlorida=33, SouthGeorgia=34, SouthernCalifornia=35, SouthwestTexas=36, Tennessee=37, Tidewater=38, Utah=39, WestVirginia=40, Wisconsin=41, International=42. `[schema §7]` - School side: `Team.SchoolPipelineInfluenceList` → `SchoolPipelineInfluence[]` (10 refs) → `SchoolPipelineInfluence {Pipeline, InfluenceLevel, InfluenceValue 0–1000}`; tier enum `PipelineInfluenceLevel` {Unrecognized=0 < NicheInterest=1 < Respected=2 < Popular=3 < HouseholdName=4 < CulturalPillar=5, Invalid=7}; observed tier cutoffs on the 0–1000 scale ≈ [0, 40, 80, 150, 250]. `[schema §7][pipe §2]` - Player side: `Player.HomePipeline` (region enum, not a state code); `PLYR_HOME_STATE` is the finer state enum; state→region mapping table in the pipeline tool's `data/stateToPipeline.json` (e.g. Texas→EastTexas, NonUS→International). `[pipe §2]` - Coach side: `Coach.PrimaryPipeline` — a personal pipeline region that travels with the coach between jobs; weighted in hiring (`HalfPipelinePoints` 25 / `MaxPipelinePoints` 50) and in the pipeline tool's model (HC 0.6 / OC 0.2 / DC 0.2 with tenure ramp `min(1, SeasonsWithTeam/ramp)`). `[pipe §2,§6][schema §7]` - Tunables: `InfluenceRequiredPerPipelineLevel[]`, `RecruitInitialInfluencePerPipelineLevel[]`, `ChanceToSwayBoostFromPipelineInfluenceLevel[]`. `[schema §7]` - Writes to SchoolPipelineInfluence rows are game-honored (pipeline tool's 980-row write survived a season and showed in-game effects). `[pipe intro][diff §4]` ## 11. The transfer portal - **There is no `TransferPortal` table.** The portal is: `Franchise.Transfers: Recruit[]` (master list) + Recruit rows with `Class = Transfer_*` + SeasonInfo period flags + events. `[schema §3][web §6]` - Portal entrants are seeded by star rating via `TransferStartingRecruitData {MinCommitScore, MaxCommitScore, RecruitStage, StarRating}` and `RecruitingTunables.TransferDefaultCommitScore = 1000`. `[schema §3]` - Events: `AddToTransferPortalEvent{Recruit}`, `EncourageTransfersStart/EndEvent`, `SendPlayerToTransferPortalAction`, `ForceTransferAction`; reaction `RecruitingEvalTransferAddedToPortalReaction`. `[schema §3][web §6]` - Team counters: `Team.LastSeasonTransfersLost` (0–30) / `LastSeasonTransfersSigned` (0–35). `[schema §6]` - **Observed**: at portal-stage week 1 (QQ2: `IsTransferPortalNewlyAvailable` + `IsTransferSignPeriodActive` both true) there were **zero Transfer_\* Recruit rows** — UNVERIFIED where/when portal players materialize (possibly deeper into stage 2; needs a later-week save). `[dump §2][diff §3]` - Retention levers on the roster side: `Player.AbsoluteTransferChance` (UNVERIFIED), `CoachTalentEffects.PlayersLeaving_LessTransferChance[8]` / `PlayersLeaving_InterestBoost_Transfers[8]` / `PlayersLeaving_ThresholdDecrease_Dealbreakers[8]`, `ProgramPoints_DecreaseRiskOfTransfer`; `LeagueSetting.CPUPlayerTransferChance` / `UserPlayerTransferChance` sliders. `[web §6][dump §12]` ## 12. The force-commit mutation recipe The only **proven** commit write-set for CFB27 (from `force-commit-recruits`, run by real users). Verbatim core (`instantCommit`): `[fc §4]` ```js // 1. Board side — hijack an existing RecruitTarget row on the school's board (never allocates rows) W(targetEntry, 'Recruit', makeRef(recruitTableId, recruitRow)); W(targetEntry, 'CommittedWeekNumber', currentWeek); // SeasonInfo.CurrentWeek W(targetEntry, 'OriginalNILExpectation', 0); W(targetEntry, 'CurrentNILOffer', 0); W(targetEntry, 'NILExpectation', 0); W(targetEntry, 'ScholarshipStatus', 'Offered'); // 2. Recruit side — stage machine W(recruitRec, 'RecruitStage', 'HardCommitted'); W(recruitRec, 'RecruitStageAdvance', 'InstantCommit'); // 3. Player side — compatibility sweetener (UNVERIFIED intent: anti-decommit?) W(playerEntry, 'RecruitingDealbreaker', schoolBestGradeDealbreaker); W(playerEntry, 'IdealRecruitingPitch', DEALBREAKER_TO_PITCH[dealbreaker]); // 4. Top-school side — the BINDING: slot 0 of TopSchoolsList const firstCol = topSchoolsListT.offsetTable.map(o => o.name)[0]; // 'ProspectTargetSchool0' topSchoolRec.TeamId = boardRow; // school's TeamIndex topSchoolRec.TeamInfluence = commitScore; // = Recruit.CommitScore → satisfies the commit threshold ``` Notes: no rollback on partial failure (earlier writes stick); the tool runs at "week 4 of the transfer-portal period" per its README; post-run, the game AI may still decommit some forced commits on advance. `[fc §4,§8,§11]` Divergence worth testing: force-commit writes `ScholarshipStatus='Offered'` while the engine's own enum has a `Committed` value that live data never shows — 'Offered' matches observed engine behavior. `[schema §10][dump §4]` An alternative, community-tested commit route **[1SRC]**: make the target team the influence leader (`RecruitTarget.ProspectInfluenceTotal`), then set `Recruit.CommitScore ≤ leader influence` → HardCommitted on next weekly advance (rival influence zeroed by the engine). `[web §5a]` ## 13. Write-survival semantics Community-verified (brooksg357, game-tested 2026-07-07) — all **[1SRC]**, crucial for durable writes: `[web §5a]` - `RecruitTarget.ProspectInfluenceTotal` (team-side) is the **master store**; the weekly advance re-derives the recruit's TopSchoolsList entry from it. List entries with no backing target row persist untouched. - **Sticky (one-shot) writes**: `Recruit.RecruitStage` (ratchet — never recomputed downward), `Recruit.CommitScore`, Player identity/ratings. - **Transient (re-apply weekly)**: board membership, RecruitTarget rows, ProspectInfluenceTotal, ActivePitches, ScheduledVisit, ScholarshipStatus. **The CPU re-curates its board every weekly advance** and drops injected 0-influence targets ⇒ durable orchestration needs per-week re-application by an external tool (exactly our tool's job). - `ActiveRecruitingPitch[]` pool can be exhausted (4,830/4,830 observed in a year-2028 save) — recycle pitch arrays instead of allocating. ## 14. Gotchas checklist 1. Skip Recruit row 0 (Invalid sentinel); gate every row on `record.isEmpty`. `[dump §2]` 2. RecruitStage has NO uncommitted literal — Top10 is base; `Signed` shares raw 6 with alias `Last_`. `[dump §2]` 3. `RecruitStageAdvance` rests at `Invalid`; it is a one-shot command field. `[dump §2]` 4. Commit truth = Recruit.RecruitStage + TopSchools slot 0. Board entries, `CommittedWeekNumber`, `ScholarshipStatus`, `TotalScholarshipOffers` are all unreliable indicators. `[dump §4]` 5. TopSchools slots must stay influence-desc sorted after edits **[1SRC]**. `[web §5a]` 6. `TeamId`/`TeamIndex` ≠ Team-table row; FCS pseudo-teams share TeamIndex 255. Build a TeamIndex→row map and expect a collision on 255. `[fc §11][diff §5]` 7. User boards point into `UserRecruitTarget`, not `RecruitTarget` — resolve each slot ref's actual table. `[dump §4]` 8. Boards are wiped between cycles; a brand-new dynasty's preseason has RecruitTarget.nextRecordToUse = 0 — a "hijack existing row" strategy has nothing to hijack there. `[diff §3]` 9. RecruitTarget rows are a scarce fixed pool (cap 4870, 4795 used mid-cycle — only ~75 spare). `[fc §11][dump §1]` 10. `Motivation1/2/3` read as raw `'0000'` — handle non-name enum reads everywhere. `[dump §9]` 11. Negative-capable fields: `CurrentScholarshipBonus` (−200..50), `ProspectInfluenceDelta` (−200..1023), `BaseNILValue` (−255..1023) — don't clamp at 0. `[schema §14]` 12. Player.readRecords() full-width is slow (16.5k × 282) — always use field projections. `[dump §15]` ## Sources - `research/notes-force-commit.md` (proven commit recipe, board mechanics, savePicker) - `research/notes-save-dump.md` + `research/dumps/QQ|QQ2|QQ2b/*.sample.json`, `research/dumps/*/scan.json`, `research/dumps/QQ2/commit-analysis.json` (live enums, distributions, commit-binding evidence) - `research/notes-save-diff.md` + `research/dumps/save-diff.json` (cross-save recruit lifecycle) - `research/notes-schema.md` + `research/schema-relevant-tables.json` (field defs, enums, tunables) - `research/notes-web.md` (brooksg357 recruiting docs [1SRC], table-map.json schemas [CODE], RO27 artifacts) - `research/notes-jersey-tool.md` (PositionE dump, offense/defense sets) - `research/notes-pipeline-tool.md` (pipeline model, state→region map)