brackt/app/services/simulations/nfl-simulator.ts
Claude 4142b21c55
Honor engine knobs across simulators, de-dupe odds, unify config UI
Two problems addressed:

1. Favorites' P(1st) was too sharp (e.g. NHL top teams ~20% vs ~12% implied).
   - The NHL simulator hardcoded its parity factor (1000) and ignored the
     season config's parityFactor, so the knob meant to flatten the
     distribution did nothing. It also re-blended raw futures odds into every
     game on top of the odds->Elo conversion, double-counting the same signal.
   - NHL now reads parityFactor/iterations/seasonGames/overtimeRate from config
     and no longer re-blends odds per game (odds enter once, via the central
     odds->Elo resolver). Honoring parity 2500 flattens a top team from ~29% to
     ~13% title odds.

2. "Season Config" and "Input Policy" were two forms over the same stored
   object that didn't reflect each other, and the engine-knob half was inert
   for many simulators.
   - Every simulator now reads its engine knobs (iterations everywhere;
     parityFactor for all Elo-based sims) from the merged config, passed in by
     the runner via the Simulator interface. Defaults equal the former
     hardcoded constants, so behavior is unchanged unless a season overrides.
   - The admin simulator page is now a single "Simulator Configuration" card
     with structured Engine and Input-derivation sections (profile-driven, so
     each sport shows only the knobs it honors) plus an Advanced raw-JSON
     escape hatch — all writing the same config.

Also: centralized the duplicated configNumber helpers into config-access.ts;
the central odds->Elo resolver now maps onto the configured Elo floor/ceiling
so those bounds set the odds-derived spread (a real flattening dial).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PAFMogMkFJf52YpHyCDvuf
2026-06-30 22:00:33 +00:00

476 lines
19 KiB
TypeScript
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

/**
* NFL Season + Playoffs Simulator
*
* Monte Carlo simulation of the NFL regular season and playoffs.
*
* Algorithm:
* 1. Load all participants for the sports season from DB
* 2. Load Elo ratings from participantExpectedValues.sourceElo (user-maintained)
* Falls back to sourceOdds → convertFuturesToElo() if no sourceElo set.
* 3. Validate that all 8 NFL divisions have at least one Elo-rated team.
* 4. Load current regular season standings (wins, gamesPlayed, conference, division)
* If no standings in DB → pre-season mode (all teams start 0-0).
* 5. For each simulation:
* a. Simulate remaining games (17 total) for each team using Elo vs average (1500)
* → projectedWins = currentWins + Binomial(remainingGames, winProb)
* b. Per conference (AFC, NFC):
* - Division winners: best projectedWins in each of 4 divisions (random tiebreak)
* - Wildcards: top 3 non-division-winners by projectedWins
* - Seed 1: best division winner (gets bye)
* - Seeds 2-4: remaining division winners, sorted by projectedWins desc
* - Seeds 5-7: wildcards, sorted by projectedWins desc
* c. Simulate Wild Card (seed 1 has bye): 2v7, 3v6, 4v5 per conference
* Higher seed has home-field advantage (+48 Elo ≈ 57% baseline win rate).
* d. Simulate Divisional: seed 1 vs lowest remaining, seed 2 vs next
* (higher seed is home)
* e. Simulate Conference Championships (higher remaining seed is home)
* f. Simulate Super Bowl at a neutral site (no home-field adjustment)
* 6. Track placements:
* - probFirst / probSecond: Super Bowl winner / loser
* - probThird / probFourth: Conference Championship losers (1 per conf, split evenly)
* - probFifthprobEighth: Divisional losers (4 total, split evenly)
* - Wild Card losers and missed playoffs → 0
*
* Elo ratings are maintained by the admin via the sourceElo field on
* participantExpectedValues. Update before each simulation run.
* Source: nfelo.app (nfelo ratings system)
*
* Home-field advantage:
* NFL playoff home field is worth ~+48 Elo points (≈57% baseline win rate for
* the higher seed). This applies to Wild Card, Divisional, and Conference
* Championship rounds. The Super Bowl is played at a neutral site.
*/
import { database } from "~/database/context";
import { eq } from "drizzle-orm";
import * as schema from "~/database/schema";
import type { Simulator, SimulationResult } from "./types";
import { normalizeTeamName } from "~/lib/normalize-team-name";
import { eloWinProbabilityWithParity, convertFuturesToElo } from "~/services/probability-engine";
import { getRegularSeasonStandings } from "~/models/regular-season-standings";
import { positiveConfigNumber } from "./config-access";
// ─── Simulation parameters (defaults; overridable via season config) ───────────
const DEFAULT_NUM_SIMULATIONS = 50_000;
/** Elo parity factor. NFL defaults to 400 (standard formula). */
const DEFAULT_PARITY_FACTOR = 400;
/** NFL regular season games per team. */
const DEFAULT_REGULAR_SEASON_GAMES = 17;
/** Average opponent Elo for regular season game projection. */
const AVG_OPPONENT_ELO = 1500;
/**
* Home-field Elo advantage for NFL playoff games.
* Adds to the higher seed's Elo before computing win probability.
* +48 Elo ≈ 57% baseline win rate for the home team, consistent with
* NFL regular-season home-field data (~2.5 point spread advantage).
*/
const NFL_HOME_FIELD_ELO_ADVANTAGE = 48;
type Conference = "AFC" | "NFC";
type Division = "East" | "North" | "South" | "West";
const CONFERENCES: Conference[] = ["AFC", "NFC"];
const DIVISIONS: Division[] = ["East", "North", "South", "West"];
// ─── Team data ────────────────────────────────────────────────────────────────
interface NflTeamData {
names: string[];
conference: Conference;
division: Division;
}
const TEAMS_DATA: NflTeamData[] = [
// AFC East
{ names: ["Buffalo Bills", "Bills"], conference: "AFC", division: "East" },
{ names: ["Miami Dolphins", "Dolphins"], conference: "AFC", division: "East" },
{ names: ["New England Patriots", "Patriots"], conference: "AFC", division: "East" },
{ names: ["New York Jets", "Jets"], conference: "AFC", division: "East" },
// AFC North
{ names: ["Baltimore Ravens", "Ravens"], conference: "AFC", division: "North" },
{ names: ["Cincinnati Bengals", "Bengals"], conference: "AFC", division: "North" },
{ names: ["Cleveland Browns", "Browns"], conference: "AFC", division: "North" },
{ names: ["Pittsburgh Steelers", "Steelers"], conference: "AFC", division: "North" },
// AFC South
{ names: ["Houston Texans", "Texans"], conference: "AFC", division: "South" },
{ names: ["Indianapolis Colts", "Colts"], conference: "AFC", division: "South" },
{ names: ["Jacksonville Jaguars", "Jaguars"], conference: "AFC", division: "South" },
{ names: ["Tennessee Titans", "Titans"], conference: "AFC", division: "South" },
// AFC West
{ names: ["Denver Broncos", "Broncos"], conference: "AFC", division: "West" },
{ names: ["Kansas City Chiefs", "Chiefs"], conference: "AFC", division: "West" },
{ names: ["Las Vegas Raiders", "Raiders"], conference: "AFC", division: "West" },
{ names: ["Los Angeles Chargers", "Chargers"], conference: "AFC", division: "West" },
// NFC East
{ names: ["Dallas Cowboys", "Cowboys"], conference: "NFC", division: "East" },
{ names: ["New York Giants", "Giants"], conference: "NFC", division: "East" },
{ names: ["Philadelphia Eagles", "Eagles"], conference: "NFC", division: "East" },
{ names: ["Washington Commanders", "Commanders"], conference: "NFC", division: "East" },
// NFC North
{ names: ["Chicago Bears", "Bears"], conference: "NFC", division: "North" },
{ names: ["Detroit Lions", "Lions"], conference: "NFC", division: "North" },
{ names: ["Green Bay Packers", "Packers"], conference: "NFC", division: "North" },
{ names: ["Minnesota Vikings", "Vikings"], conference: "NFC", division: "North" },
// NFC South
{ names: ["Atlanta Falcons", "Falcons"], conference: "NFC", division: "South" },
{ names: ["Carolina Panthers", "Panthers"], conference: "NFC", division: "South" },
{ names: ["New Orleans Saints", "Saints"], conference: "NFC", division: "South" },
{ names: ["Tampa Bay Buccaneers", "Buccaneers"], conference: "NFC", division: "South" },
// NFC West
{ names: ["Arizona Cardinals", "Cardinals"], conference: "NFC", division: "West" },
{ names: ["Los Angeles Rams", "Rams"], conference: "NFC", division: "West" },
{ names: ["San Francisco 49ers", "49ers"], conference: "NFC", division: "West" },
{ names: ["Seattle Seahawks", "Seahawks"], conference: "NFC", division: "West" },
];
// ─── Internal types ───────────────────────────────────────────────────────────
interface TeamState {
participantId: string;
conference: Conference;
division: Division;
elo: number;
currentWins: number;
gamesPlayed: number;
}
interface SeededTeam {
participantId: string;
elo: number;
seed: number;
}
interface ConferenceBracketResult {
finalist: SeededTeam;
confChampLoser: SeededTeam;
divisionalLosers: [SeededTeam, SeededTeam];
}
// ─── Helpers ──────────────────────────────────────────────────────────────────
/**
* Look up team data by participant name.
*/
export function getTeamData(name: string): NflTeamData | undefined {
const normalized = normalizeTeamName(name);
return TEAMS_DATA.find((t) =>
t.names.some((n) => normalizeTeamName(n) === normalized)
);
}
/**
* Simulate projected wins for a team over its remaining games.
* Each game is an independent Bernoulli trial.
*/
function simulateProjectedWins(
currentWins: number,
gamesPlayed: number,
winProb: number,
seasonGames = DEFAULT_REGULAR_SEASON_GAMES
): number {
const remaining = Math.max(0, seasonGames - gamesPlayed);
let additional = 0;
for (let g = 0; g < remaining; g++) {
if (Math.random() < winProb) additional++;
}
return currentWins + additional;
}
/**
* Determine the 7-team playoff seeding for one conference.
*
* Seeds 1-4 are division winners (sorted by projectedWins desc).
* Seeds 5-7 are wildcards (top 3 non-division-winners).
* Seed 1 receives the bye.
*
* Ties in projectedWins are broken by a pre-computed random jitter, which
* avoids the instability of using Math.random() directly inside a sort comparator.
*/
export function seedConference(
teams: Array<{ participantId: string; elo: number; division: Division; projectedWins: number }>
): SeededTeam[] {
// Pre-compute per-team jitter so each sort comparison is deterministic
const jitter = new Map(teams.map((t) => [t.participantId, Math.random()]));
const cmp = (a: typeof teams[0], b: typeof teams[0]) =>
b.projectedWins - a.projectedWins ||
(jitter.get(b.participantId) ?? 0) - (jitter.get(a.participantId) ?? 0);
const divisionWinners: typeof teams[0][] = [];
const nonWinners: typeof teams[0][] = [];
for (const div of DIVISIONS) {
const divTeams = teams.filter((t) => t.division === div).toSorted(cmp);
if (divTeams.length === 0) continue;
divisionWinners.push(divTeams[0]);
nonWinners.push(...divTeams.slice(1));
}
divisionWinners.sort(cmp);
nonWinners.sort(cmp);
const wildcards = nonWinners.slice(0, 3);
const seeded: SeededTeam[] = [];
divisionWinners.forEach((t, i) =>
seeded.push({ participantId: t.participantId, elo: t.elo, seed: i + 1 })
);
wildcards.forEach((t, i) =>
seeded.push({ participantId: t.participantId, elo: t.elo, seed: i + 5 })
);
return seeded;
}
/**
* Simulate a playoff game with home-field advantage for the higher seed.
* The team with the lower seed number (better seed) is the home team.
*/
function playGame(a: SeededTeam, b: SeededTeam, parityFactor = DEFAULT_PARITY_FACTOR): { winner: SeededTeam; loser: SeededTeam } {
const [home, away] = a.seed < b.seed ? [a, b] : [b, a];
const winProbHome = eloWinProbabilityWithParity(
home.elo + NFL_HOME_FIELD_ELO_ADVANTAGE,
away.elo,
parityFactor
);
if (Math.random() < winProbHome) return { winner: home, loser: away };
return { winner: away, loser: home };
}
/**
* Simulate a game at a neutral site (no home-field adjustment).
* Used for the Super Bowl.
*/
function playNeutralGame(a: SeededTeam, b: SeededTeam, parityFactor = DEFAULT_PARITY_FACTOR): { winner: SeededTeam; loser: SeededTeam } {
const winProbA = eloWinProbabilityWithParity(a.elo, b.elo, parityFactor);
if (Math.random() < winProbA) return { winner: a, loser: b };
return { winner: b, loser: a };
}
/**
* Simulate one conference's playoff bracket (seeds 1-7, seed 1 has bye).
*
* Wild Card: 2v7, 3v6, 4v5 (seed 1 bye; higher seed is home)
* Divisional: 1 vs lowest remaining, 2 vs next (higher seed is home)
* Conference Champ: divisional winners (higher seed is home)
*
* Returns the finalist, the conference championship loser, and the two
* divisional-round losers (for 5th-8th place EV).
*/
export function simulateConferenceBracket(seeds: SeededTeam[], parityFactor = DEFAULT_PARITY_FACTOR): ConferenceBracketResult {
const byIndex = new Map(seeds.map((t) => [t.seed, t]));
const s = (seed: number) => byIndex.get(seed) as SeededTeam;
const pg = (a: SeededTeam, b: SeededTeam) => playGame(a, b, parityFactor);
// Wild Card (seed 1 bye; lower seed = home)
const wc27 = pg(s(2), s(7));
const wc36 = pg(s(3), s(6));
const wc45 = pg(s(4), s(5));
// Divisional: sort WC winners by original seed (ascending = better seed = home)
// Seed 1 hosts the lowest-ranked (highest seed number) WC winner
const wcWinners = [wc27.winner, wc36.winner, wc45.winner].toSorted(
(a, b) => a.seed - b.seed
);
const div1 = pg(s(1), wcWinners[wcWinners.length - 1]);
const div2 = pg(wcWinners[0], wcWinners[1]);
// Conference Championship (higher remaining seed is home)
const conf = pg(div1.winner, div2.winner);
return {
finalist: conf.winner,
confChampLoser: conf.loser,
divisionalLosers: [div1.loser, div2.loser],
};
}
// ─── Simulator class ──────────────────────────────────────────────────────────
export class NFLSimulator implements Simulator {
async simulate(sportsSeasonId: string, config: Record<string, unknown> = {}): Promise<SimulationResult[]> {
const db = database();
const parityFactor = positiveConfigNumber(config, "parityFactor", DEFAULT_PARITY_FACTOR);
const numSimulations = Math.round(positiveConfigNumber(config, "iterations", DEFAULT_NUM_SIMULATIONS));
const seasonGames = Math.round(positiveConfigNumber(config, "seasonGames", DEFAULT_REGULAR_SEASON_GAMES));
// Load participants
const participants = await db.query.seasonParticipants.findMany({
where: eq(schema.seasonParticipants.sportsSeasonId, sportsSeasonId),
});
if (participants.length === 0) {
throw new Error(`No participants found for sports season ${sportsSeasonId}.`);
}
// Load EV rows (sourceElo + sourceOdds)
const evRows = await db
.select({
participantId: schema.seasonParticipantExpectedValues.participantId,
sourceElo: schema.seasonParticipantExpectedValues.sourceElo,
sourceOdds: schema.seasonParticipantExpectedValues.sourceOdds,
})
.from(schema.seasonParticipantExpectedValues)
.where(eq(schema.seasonParticipantExpectedValues.sportsSeasonId, sportsSeasonId));
// Build Elo map: prefer sourceElo, fall back to converting sourceOdds
const eloMap = new Map<string, number>();
const hasSourceElo = evRows.some((r) => r.sourceElo !== null);
if (hasSourceElo) {
for (const r of evRows) {
if (r.sourceElo !== null) eloMap.set(r.participantId, r.sourceElo);
}
} else {
const hasOdds = evRows.some((r) => r.sourceOdds !== null);
if (!hasOdds) {
throw new Error(
`No Elo ratings or futures odds found for sports season ${sportsSeasonId}. ` +
`Enter sourceElo ratings via Admin → Expected Values before simulating.`
);
}
const oddsInput = evRows
.filter((r) => r.sourceOdds !== null)
.map((r) => ({ participantId: r.participantId, odds: r.sourceOdds as number }));
const converted = convertFuturesToElo(oddsInput);
for (const [id, elo] of converted) eloMap.set(id, elo);
}
// Load standings (empty pre-season)
const standingsRows = await getRegularSeasonStandings(sportsSeasonId, db);
const standingsMap = new Map(standingsRows.map((r) => [r.participantId, r]));
// Build team state list — skip participants with no Elo or unknown team name
const teams: TeamState[] = [];
for (const p of participants) {
const teamData = getTeamData(p.name);
if (!teamData) continue;
const elo = eloMap.get(p.id);
if (elo === undefined) continue;
const standing = standingsMap.get(p.id);
teams.push({
participantId: p.id,
conference: teamData.conference,
division: teamData.division,
elo,
currentWins: standing?.wins ?? 0,
gamesPlayed: standing?.gamesPlayed ?? 0,
});
}
if (teams.length === 0) {
throw new Error(
`Could not match any participants to NFL team data for season ${sportsSeasonId}. ` +
`Ensure participant names match (e.g. "Kansas City Chiefs").`
);
}
// Validate that every division in every conference has at least one Elo-rated team.
// A missing division means seeding is broken for the whole simulation.
const covered = new Set(teams.map((t) => `${t.conference} ${t.division}`));
const missing: string[] = [];
for (const conf of CONFERENCES) {
for (const div of DIVISIONS) {
if (!covered.has(`${conf} ${div}`)) missing.push(`${conf} ${div}`);
}
}
if (missing.length > 0) {
throw new Error(
`Missing Elo-rated teams in: ${missing.join(", ")}. ` +
`All 32 NFL teams must have Elo ratings set before simulating.`
);
}
// Pre-compute per-team win probability vs average opponent (constant across sims)
const winProbMap = new Map<string, number>();
for (const t of teams) {
winProbMap.set(t.participantId, eloWinProbabilityWithParity(t.elo, AVG_OPPONENT_ELO, parityFactor));
}
// Placement counters
const counts: Record<string, Record<1 | 2 | 3 | 4 | 5 | 6 | 7 | 8, number>> = {};
for (const t of teams) {
counts[t.participantId] = { 1: 0, 2: 0, 3: 0, 4: 0, 5: 0, 6: 0, 7: 0, 8: 0 };
}
// ─── Monte Carlo ──────────────────────────────────────────────────────────
for (let sim = 0; sim < numSimulations; sim++) {
// Project wins for each team this simulation
const projected = teams.map((t) => ({
participantId: t.participantId,
elo: t.elo,
conference: t.conference,
division: t.division,
projectedWins: simulateProjectedWins(
t.currentWins,
t.gamesPlayed,
winProbMap.get(t.participantId) as number,
seasonGames
),
}));
const afcTeams = projected.filter((t) => t.conference === "AFC");
const nfcTeams = projected.filter((t) => t.conference === "NFC");
const afcSeeds = seedConference(afcTeams);
const nfcSeeds = seedConference(nfcTeams);
// Simulate conference brackets
const afcResult = simulateConferenceBracket(afcSeeds, parityFactor);
const nfcResult = simulateConferenceBracket(nfcSeeds, parityFactor);
// Super Bowl (neutral site)
const sb = playNeutralGame(afcResult.finalist, nfcResult.finalist, parityFactor);
// 1st / 2nd
counts[sb.winner.participantId][1] += 1;
counts[sb.loser.participantId][2] += 1;
// 3rd / 4th — Conference Championship losers (1 per conference, split 50/50)
counts[afcResult.confChampLoser.participantId][3] += 0.5;
counts[afcResult.confChampLoser.participantId][4] += 0.5;
counts[nfcResult.confChampLoser.participantId][3] += 0.5;
counts[nfcResult.confChampLoser.participantId][4] += 0.5;
// 5th8th — Divisional losers (4 total: 2 per conference, split evenly)
for (const loser of [...afcResult.divisionalLosers, ...nfcResult.divisionalLosers]) {
counts[loser.participantId][5] += 0.25;
counts[loser.participantId][6] += 0.25;
counts[loser.participantId][7] += 0.25;
counts[loser.participantId][8] += 0.25;
}
// Wild Card losers and missed playoffs remain at 0 (already initialized)
}
// Convert counts to probabilities
return teams.map((t) => {
const c = counts[t.participantId];
const N = numSimulations;
return {
participantId: t.participantId,
probabilities: {
probFirst: c[1] / N,
probSecond: c[2] / N,
probThird: c[3] / N,
probFourth: c[4] / N,
probFifth: c[5] / N,
probSixth: c[6] / N,
probSeventh: c[7] / N,
probEighth: c[8] / N,
},
source: "nfl_elo_simulation",
};
});
}
}