Skip to content

Instantly share code, notes, and snippets.

@reggiechan74
Last active October 12, 2025 14:46
Show Gist options
  • Select an option

  • Save reggiechan74/e8fdf534710a8ea631d75c19d783fca3 to your computer and use it in GitHub Desktop.

Select an option

Save reggiechan74/e8fdf534710a8ea631d75c19d783fca3 to your computer and use it in GitHub Desktop.
Sheriff of Nottingham Strategy Guide

Sheriff of Nottingham — Complete Strategy Guide (now with round timing)

This is the integrated, tactical playbook. It merges the Merchant and Sheriff strategies, the bribe math, bag-loading rules, psychological play, and explicit corrections for game length by player count so you can schedule trust-building, deception windows, and inspection cadence precisely.


1. Core mechanics & notation (compact)

  • $P$ = penalty total on declared legal cards (coins).
  • $M$ = penalty total on misdeclared/illegal cards (coins).
  • $p_T$ = sheriff’s subjective probability the merchant is truthful (0–1).
  • $B$ = bribe offered (coins).

Sheriff:

$$ EV_{\text{open}} = p_T(-P) + (1 - p_T)M $$

Accept bribe if $B \ge EV_{\text{open}}$. Inspect if $EV_{\text{open}} > B$.

Merchant indifference bribe:

$$ B^* = \max(0,, p_T(-P) + (1 - p_T)M) $$

Merchant cap (never overpay):

$$ B_{\max} = (V_{\text{pass}} - V_{\text{caught}}) + M + \Delta_{\text{bonus}} $$

All decisions are EV-driven; psychology shifts perceived $p_T$.


2. Round-count corrections (how many rounds, and tactical timing)

Use these exact round budgets to plan reputation and deception windows.

  • 3 players → 9 rounds (each player Sheriff 3 times).

    • Trust-building window: Rounds 1–2.
    • Mid-game deception: Rounds 3–6 (one clear push).
    • Endgame control: Rounds 7–9 (bonus denial critical).
  • 4 players → 8 rounds (each player Sheriff 2 times).

    • Trust-building: Round 1.
    • Mid-game deception window: Rounds 2–5 (one concentrated push).
    • Endgame: Rounds 6–8 (two rounds to force bonus decisions).
  • 5 players → 10 rounds (standard) — if your table uses the standard 2× Sheriff rotation.

    • Trust-building: Rounds 1–2.
    • Two deception windows: Rounds 3–5 and 6–8.
    • Endgame: Rounds 9–10 (bonus denial & final pushes).
  • 5 players → 5 rounds (compressed variant) — some groups compress to one Sheriff rotation.

    • No long trust-building. Play accelerated: early deception, mid-game scramble, immediate endgame. Increase inspection frequency.
  • 6 players (Deputies module) — mechanics change (two inspectors). Treat every merchant bag as having higher inspection probability; bribe must sway two minds or be structured to make one deputy act alone. Round count and end condition depend on table rules—treat as high inspection risk environment.

Tactical implication: plan reputation investments proportional to number of rounds. More rounds → invest in reputation early; fewer rounds → front-load aggressive plays.


3. Merchant playbook — exact, per-phase

3.1 Phase timing (use above round counts)

  • Early phase (build $p_T$): be honest.

    • 3-player: Rounds 1–2 honest.
    • 4-player: Round 1 honest.
    • 5-player standard: Rounds 1–2 honest.
    • Compressed: 0–1 honest rounds (if any).
  • Mid phase (harvest trust with calibrated lies): insert small contraband mixes (one illegal + 3–4 legal). Use one or two deception windows depending on total rounds.

  • Late phase (positioning for bonuses or comeback): shift according to score — if leading, revert to honesty; if trailing, increase contraband density and accept higher bribes subject to $B_{\max}$.

3.2 Bag composition rules (operational)

  • Prefer 1 illegal + 3–4 matching legal. Reason: $M$ small, $P$ large → low $B^*$.
  • Never load a bag where $B^* > B_{\max}$. If computed $B^* > B_{\max}$, repack.
  • Avoid concentrating contraband in a single bag unless you have strong information that $p_T$ is high (≥0.7) or Sheriff is broke/out of motivation.

3.3 Bribe sizing (simple heuristics)

Given computed $p_T$, $P$, $M$:

  • If $p_T \ge \dfrac{M}{M+P}$ → offer 0 (sheriff loses EV opening).
  • Else compute $B^* = p_T(-P) + (1-p_T)M$ and offer $B = \lceil B^* \rceil$, but cap at $B_{\max}$.

Practical quick table (rounded):

  • $p_T \ge 0.5$ → 0–1 coin.
  • $0.35 \le p_T < 0.5$ → 1–2 coins.
  • $0.20 \le p_T < 0.35$ → 2–3 coins.
  • $p_T < 0.20$ → 3+ coins or repack.

3.4 Example (5-card bag, explicit)

Bag: declare 5 Apples (Apple penalty 2 each → $P=10$). Actual: 3 Apples + 1 Cheese (penalty 2) + 1 Silk (penalty 4) → $M=6$.

  • Break-even trust: $p_T^c = M/(M+P) = 6/16 = 0.375$.
  • If $p_T \ge 0.40$ → offer 0.
  • If $p_T = 0.30$$B^* = 1.2$ → offer 2 (round up).
  • $B_{\max}$ computed with values: pass goods value 13, caught value 6 → $\Delta_{\text{goods}}=7$, $B_{\max}=7+6=13$. So offering 2 is well under cap.

3.5 Table-size adjustments (explicit)

  • 3 players: be stingy with lies; sheriffs inspect often. Use one reliable deception window only.
  • 4 players: follow main guide. One or two deception pushes.
  • 5 players standard: you can execute two deception windows; reputation decay slower.
  • 5 compressed / 6 players: push earlier, expect more inspections; favor smaller $M$ and higher reliance on honest passes for penalty income.

4. Sheriff playbook — exact, per-phase

4.1 Core decision

For each bag compute:

$$ EV_{\text{open}} = p_T(-P) + (1-p_T)M $$

Open if $EV_{\text{open}} > B$. If no bribe, open if $EV_{\text{open}}>0$.

Round-count impact:

  • In short games (compressed 5-player), prioritize immediate EV and bonus denial; less tolerance for reputation manipulation.
  • In long games (3-player, 9 rounds), use signaling early, conserve coins mid-game, then ramp inspections late.

4.2 Prioritization (when multiple merchants offer bribes)

Order by:

  1. Player who threatens your lead (highest current score / nearest bonus swing).
  2. Player uninspected for multiple rounds.
  3. Player with low $P$ and suspected high $M$.
  4. Pattern-breakers (card type/count sudden changes).
  5. Low bribe offers when your EV indicates inspection profitable.

4.3 Bribe anchoring and signaling

  • Early: accept small bribes to seed coins and set baseline.
  • Mid: raise minimum bribe acceptance threshold publicly (e.g., “Under 3 isn’t worth my time”) to increase expected merchant payments.
  • Randomly inspect a truthful bag occasionally to signal unpredictability — lowers overall lying at modest coin cost.

4.4 Endgame adjustments (explicit)

  • If leading: accept bribes more often; deny large swings only if inspection secures a direct denial of an opponent’s bonus that would overtake you.
  • If trailing: inspect aggressively; accept short-term coin losses for higher variance to flip standings.
  • Always treat potential King/Queen flip as additional $M$ when computing EV: add the expected point swing converted to coin-equivalent to $M$.

4.5 Multi-inspector (6 players / Deputies)

  • Bribe must satisfy two inspectors or be structured so one deputy accepts and acts alone. Anchor bribe higher. Overall inspection probability increases; tighten acceptance thresholds.

5. Meta layers, bookkeeping, and calibration

5.1 Per-player ledger (must do this)

For each opponent keep three quick counters:

  • TruthfulCount $T$ (times inspected and truthful)
  • LieCount $L$ (times inspected and lying)
  • LastInspectedRound (round index)

Compute $p_T = T/(T+L)$. Update after each inspection. For small samples, smooth toward global prior (add 1 pseudo-count if you want conservative estimates).

5.2 Liquidity management

  • Merchant: keep ~10–15 coins mid-game to sustain bribes and penalties.
  • Sheriff: keep ~10 coins to credibly threaten inspection. If broke, anchoring and signaling lose power.

5.3 Tactical psychological plays (use sparingly)

  • Public math: state your estimate of $p_T$ and the computed EV to shift perceptions.
  • Praise someone you’ve just caught (signals impartiality).
  • Token bribes: merchants offering 1 coin often expect leniency; accept occasionally to build funds and maintain unpredictability.

6. Full worked scenario (4-player, 8 rounds, midgame example)

Context: Round 4 of 8. You are Merchant, behind by one bonus. Sheriff is conservative (has paid once earlier). You estimate $p_T$ for you ≈ 0.45.

Bag: declare 5 Apples → $P=10$. Actual: 3 Apples, 1 Cheese (penalty 2), 1 Silk (penalty 4) → $M=6$.

  • Break-even $p_T^c = 6/16 = 0.375$. Your $p_T = 0.45 > p_T^c$ → Sheriff loses EV opening. Offer 0 (or token 1 to be safe). If you need the Apple set for a King bonus next round and $\Delta_{\text{bonus}}$ is large, you can offer 1 coin to secure pass certainty.

Alternate (if Sheriff looks suspicious; $p_T$ estimated 0.30):

  • Compute $B^* = 0.30(-10) + 0.70(6) = -3 + 4.2 = 1.2$. Round up → offer 2. Compare to $B_{\max} = 13$ → safe. Offer 2, keep a chip reserve.

Sheriff reaction (if you are Sheriff instead):

  • With $p_T=0.30$, $EV_{\text{open}} = 1.2$. Accept no bribe <2; inspect if merchant offers 1. If merchant offers 2, accept it.

7. Short cheat sheets (memorize)

Merchant:

  • Compute $P, M$. Compute $p_T^c = M/(M+P)$. If estimated $p_T &gt; p_T^c$, offer 0. Else offer $\lceil p_T(-P)+(1-p_T)M\rceil$ capped at $B_{\max}$. Repack if $B^* &gt; B_{\max}$.

Sheriff:

  • Compute $EV_{\text{open}} = p_T(-P) + (1-p_T)M$. Inspect if $EV_{\text{open}} &gt; B$. Prioritize high-score threats, uninspected streaks, and pattern-changers. Adjust inspection frequency by round-budget position.

Round budgeting:

  • 3-player (9 rounds): build trust early, one midgame push, aggressive endgame.
  • 4-player (8 rounds): single deception window midgame, two endgame rounds.
  • 5-player (10 rounds): build trust, two deception windows, last two rounds decisive.
  • 5 compressed/6 players: accelerate deception; expect heavy inspections.

8. Final operational checklist (run fast each round)

  1. Update ledger $T, L$. Compute $p_T$.
  2. For your candidate bag compute $P, M, V_{\text{pass}}, V_{\text{caught}}$.
  3. If Merchant: compute $B^{}$, $B_{\max}$. If $B^{} > B_{\max}$ → repack. Else offer $\min(\lceil B^{*} \rceil, B_{\max})$.
  4. If Sheriff: compute $EV_{\text{open}}$. If there’s a bribe $B$, accept when $B \ge \lceil EV_{\text{open}}\rceil$; otherwise open. Use priority tiebreakers.
  5. Adjust based on round number and standings (apply endgame bias to deny bonuses when necessary).

This guide is the full operational manual: the formulas, the heuristics, the round-aware timing, and the examples to execute at the table. Use the ledger and the round budgets to convert the abstract EV rules into exact, repeatable actions.


Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment