Mixed-Strategy Policy Table¶
Policy artifact — instantiates Strategic Randomization and Exploitability Reduction
A written register that lists each viable action, its probability weight, its safety constraints, and the exceptions that override the draw — the declarative source of truth for a mixed strategy.
A Mixed-Strategy Policy Table is the plain artifact at the center of the whole archetype: a written register in which each row is a viable action, and its columns record the action's probability weight, the constraints it must satisfy to be legitimate, the trigger that revises its weight, the accountable owner, and the exception rule that lets a human override the draw. Its defining move is declaration, not computation or execution: it fixes what the support set is, what each action is allowed to weigh, and when a person may set the dice aside — turning an implicit habit into an explicit, reviewable specification. Everything downstream (a service that draws from it, a dashboard that watches it, a red team that tests it) points back to this table as the source of truth.
Example¶
An American-football offensive coordinator builds a play-calling table for third-and-short. The rows are the plays actually in the game plan and drilled that week: inside run, play-action pass, quarterback sneak, quick out. Each row carries a weight — inside run 40%, sneak 25%, play-action 20%, quick out 15% — set so the defense can't sit on the favorite. The constraint column records what makes a play legitimate to call: the sneak is struck out inside the opponent's five-yard line where a stuffed play risks a safety, and play-action is barred if the tackle is injured.
The exception column is where the human stays in charge: if the defense shows an obvious all-out blitz, the coordinator overrides the draw and audibles to the quick out regardless of what the weights say. On game day an assistant literally reads the weighted card, but the table is what makes the mix disciplined rather than a hunch — the defense sees a genuinely unpredictable call, while the staff can point to exactly why each play is in the mix and why the audible was allowed.
How it works¶
- Fix the support. Enter only actions that are safe, authorized, resourced, and worth choosing — the pruning step that keeps the table from padding the mix with dominated options.
- Assign the weights. Record an explicit probability per row, summing to one, chosen so no single action is cheap to anticipate.
- Attach the guards. Each row carries the constraints that make it legitimate and the conditions under which it is temporarily struck from the draw.
- Write the exceptions. Name the specific situations in which a designated owner may override the weighted draw, so overrides are principled rather than ad hoc.
Tuning parameters¶
- Support width — how many rows are in the table. Wider is harder to predict but risks admitting weak actions; narrower is sharper but more readable.
- Weight spread — how flat or peaked the probabilities are. Flat weights maximize unpredictability; peaked weights protect performance but leave a favorite to exploit.
- Exception breadth — how many situations license an override. Generous exceptions keep humans safe and adaptive but reopen the predictability the table was meant to close.
- Revision cadence — how often weights and rows are revisited. Frequent revision tracks a shifting game but risks the staff drifting into a readable rhythm of its own.
When it helps, and when it misleads¶
Its strength is legitimacy made visible: the table is what lets an organization randomize on the record, showing a reviewer that every action in the support is viable, every weight is intentional, and every override is bounded — the accountability half of the archetype's central trade-off. It is also the shared spec that every other mechanism consumes, so keeping it explicit prevents the mix from decaying into folklore.
Its failure mode is dominated-action padding: the temptation to add weak rows purely to look unpredictable, which lowers predictability and performance at once — folding a dominated strategy into the support to seem sophisticated.[n1] A subtler failure is that a written table is only as random as its execution — if the exceptions are invoked on convenience, or a person "draws" by feel instead of by the weights, the paper mix and the real one diverge (the archetype's pseudo-random-habit trap). The classic misuse is treating the table as decoration while decisions are actually made by gut. The guarding discipline is to prune every row against a dominated-action check, keep the exception list short and specific, and pair the table with an external draw and an independent look at the realized frequencies.
How it implements the components¶
strategic_action_repertoire— the row set is the declared support: the pruned list of actions that may actually be chosen.probability_policy— the weight column states the explicit, bounded distribution over those actions, with owner and update trigger recorded alongside.constraint_and_safety_boundary— the constraint column encodes what keeps each action safe, legal, and authorized, so safety dominates the draw.exception_override_rule— the exception column names when a human may legitimately override the weighted draw.
It does not add a stratified_randomization_layer — dividing the population into risk tiers is the job of Randomized Patrol or Route Schedule — and it names no private_randomness_source; producing tamper-resistant draws from the table is left to Random-Seeded Assignment Service.
Related¶
- Instantiates: Strategic Randomization and Exploitability Reduction — this is the archetype's simplest, most direct mechanism: the declared probability policy itself.
- Sibling mechanisms: Random-Seeded Assignment Service · Randomized Patrol or Route Schedule · Exploitability Matrix Review · Adversarial Bandit Exploration Policy · Entropy Budget Dashboard · Red-Team Predictability Test · Commit-Reveal Random Draw · Randomized Decoy Rotation · Stochastic Challenge or Audit Timing
Editorial Notes¶
Form Classification¶
Form family: Rule, Policy & Commitment
Rationale: The table establishes standing action weights, safety constraints, and override conditions that govern every future mixed-strategy draw.
Nearest alternative: Decision, Gate & Allocation — A later draw chooses one action, but this mechanism is the persistent policy under which those bounded selections occur.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Economics & Finance
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Mixed strategies and dominated-action reasoning were canonically formulated in game theory, housed here with economics and finance.
Related originating lineages:
- Mathematics — Probability and formal game theory supply the mathematical representation of action weights.
- Operations Research — Decision-policy tables and constrained randomized policies were operationalized in optimization and decision analysis.
Review resolution: Both independent reviews agree on primary origin economics_finance; reconciliation resolves secondary fields (origin_mode_disagreement, encyclopedia_synthesis_disagreement). Alternate origins retained (mathematics, operations_research) are the union of reviewer-supported formative lineages with explicit rationales, not a list of later application domains. Present-day breadth is represented separately as domain_reach=multi_domain; origin_mode=cross_disciplinary_synthesis records the historical relationship among lineages. Confidence is conservatively reconciled to high, and encyclopedia_synthesis=true preserves either reviewer's finding that the encyclopedia generalized the mechanism.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
[n1] In game theory a dominated strategy is one that yields a worse payoff than some alternative under every opponent response; a rational mixed strategy never places weight on a strictly dominated action. Padding a policy with dominated rows to look unpredictable is the failure the archetype names as "dominated-action padding." ↩