mpg

MPG / Mainline-Path Game

Core Claim / 核心判断

MPG / Mainline-Path Game（主线-路径博弈）处理的不是“主趋势是否存在”，而是：

看对主线，不等于穿过路径。

SELA and MPG are sibling strategic lenses. Common order: SELA calibrates direction before MPG tests carrier/path; sequence, not hierarchy. SELA calibrates direction; MPG stress-tests how an actor crosses volatility, resistance, carrier fragility, and timing.

SELA ↔ MPG Twin-Lens Handshake

MPG direct-load companion check: when a system-efficiency/trend-based mainline appears, run SELA support check before path strategy; must read mindthus:sela; Do not treat this as an internal memory-only check.

Direct-load output obligation: do not wait for using-mindthus to carry this handoff. When MPG is loaded directly, the first visible sentence must be a plain-language thesis that separates long-term direction from current carrier/path in ordinary language. Default answer must not start with debug labels. Good first sentence: 长期保留内部模型能力是对的，但当前全栈自建平台这条路不该继续。 Internal labels (direction, current carrier/path, qualified_mainline, carrier_vehicle, etc.) are debug/audit support; hide unless asked. State whether the direction is supported, failed, or evidence-limited; then choose path posture.

Use SELA support + MPG dominate when direction is plausible but carrier, exposure, vehicle, or commitment still decides action. MPG still owns the action posture.

If the direction check fails, return to SELA or EDSP before path strategy. If variables visibly pull in different directions, include a short AQM visibility map or say why it is skipped.

Primary deliverable:

Path-Carrying Strategy / 主线承载方案

This is not a risk/opportunity score; it is an action-bearing plan: hold, stage, hedge, switch vehicle, pause, or challenge the mainline claim.

Internal audit fields (hide by default):

• qualified_mainline
• carrier_vehicle
• counter_force_map
• exposure_budget
• optionality_design
• action_posture
• trigger_conditions
• mainline_challenge

Mainline / 主路径

When To Use

Use MPG only when all three conditions are present:

1. A mainline, strategic target, convergence claim, or long-term direction already exists.
2. The route is nonlinear because counter-forces can reshape the path before the mainline resolves.
3. A real actor must configure carrier, exposure, timing, optionality, authority, or vehicle before the mainline is realized.

If the user is only asking whether a prior answer followed MPG or MPG-AQM rules, treat the method name as review scope and inspect the transcript/evidence directly. Do not load MPG as the current judgment owner for that evidence-review task.

Structural trigger:

mainline + proxy/carrier + concentrated exposure + now/continue/commit decision

Use MPG when a long-term mainline meets scarce resource (稀缺资源) exposure on the purest carrier (最纯载体), proxy, or vehicle. This is not domain-specific.

Human-Readable First / 先讲人话

MPG output must be readable before it is structured.

Short rule:

先给人看的判断，再给机器看的字段。

Do not expose MPG internal field names before the plain-language conclusion.

Default order: 一句话判断;为什么;怎么办;校准字段 internal unless asked.

The first paragraph should be 普通人能复述. If the answer starts with qualified_mainline: or a dense field table, rewrite it.

Reasoning Durability / 推演耐久性

MPG playback should evaluate whether the reasoning held under the 当时信息面, not whether the final outcome happened to match.

Short rule:

结果不准不必然失败；信息不全时，触发条件比预测命中更重要。

later outcome is not the score. Ask whether reasoning respected evidence limits, separated mainline from carrier, named path risks, and left usable triggers.

In playback review: trigger conditions should explain when a later result changed the judgment.

MPG-AQM Visibility Layer / 主线-路径显影层

Use Approximate Quantified Mapping / 非精准量化显影 only after the plain-language MPG judgment is clear and only when the path relationship is complex enough to hide the dominant variables.

Short rule:

数字是假设，关系才是重点。

MPG owns the judgment. AQM only makes variables visible: mainline strength, path resistance, carrier fragility, information gap, trigger strength.

Visible snapshot trigger: when the user says variables are many, asks for the dominant factor, or rejects generic balance / not generic balance, add a short visible AQM snapshot / 显影快照 after the one-sentence thesis: mainline strength / path resistance / carrier fragility / information gap / trigger strength -> stage/probe, not commit.

Boundary: do not calculate the decision. If the discussion starts defending exact digits, stop AQM and return to evidence boundaries or plain-language MPG reasoning.

Fidelity Support

For high-impact, replayed, or regression-tested MPG runs, use the fidelity contract in resources/fidelity-contract.md and, when useful, validate a structured audit artifact with scripts/validate_mpg_output.py.

The validator emits an MPG Shape & Evidence Risk Report. It checks whether required path-carrying moves are present, but it does not judge mainline truth, forecast accuracy, or outcome hit rate.

Operating Flow

Qualify mainline -> map counter-forces -> separate mainline/carrier/vehicle -> set exposure_budget -> design optionality -> choose action_posture -> set trigger_conditions -> state mainline_challenge.

Guardrails / 从属补漏

Do Not Collapse MPG Into SELA Timing Check

Use SELA for long-term system efficiency versus local advantage. Use MPG when direction must cross path volatility.

Short relation:

Common order: SELA calibrates direction before MPG tests carrier/path; sequence, not hierarchy.

Do Not Use MPG Without A Carrier

Do not use MPG without a carrier. Without actor, vehicle, exposure, timing, authority, or path decision, use explanation, SELA, EDSP, or fact-gathering.

Keep Evidence Honest

MPG can use hypothetical numbers, not measurements. Use Evidence / Claim Ceiling when facts, probabilities, markets, policy, medical/legal/safety, or financial advice require evidence.

Counter-Forces Are Not Villains

counter-forces are not bad by default. Incentives, cashflow, legacy needs, regulation, drawdowns, or resistance can carry real information; treat them as path-shaping forces.

Survival Alone Is Not Success

survival alone is not success. Surviving by abandoning mainline, core advantage, or future upside fails MPG.

Boundaries / 边界

• If the mainline itself is unclear, use EDSP or SELA before MPG.
• If the issue is “which layer should control this work,” Use WAE.
• If the missing input is facts, domain research, runtime proof, stakeholder judgment, or fresh market/legal/medical evidence, gather evidence first.
• If the task is just a bounded artifact that looks thin, use TVG.
• If a Mission needs durable task state and decision hooks, use tplan; route to MPG only when active object fits.
• Do not use MPG as a generic risk matrix, opportunity list, or narrative essay.

Runtime Support / 支撑材料

• resources/methodology.md — full MPG methodology, contract, boundaries, mixed cases.
• resources/fidelity-contract.md — MPG fidelity contract for v0.9.
• templates/fidelity-output.json — structured audit artifact template.
• scripts/validate_mpg_output.py — MPG shape and evidence-risk validator.
• tests/mpg/pressure_tests.md — scenario pressure tests for effect, misuse, and routing boundaries.