# From Simulated to Live: The Promotion Gate
> **Draft — S8.3 · series: CONTENT-ROADMAP-99 S8 (Simulation → Live) · status: DRAFT · grounding: wiki simulation-mode · edge-detection-cli · dashboard · task-tracker · skills: research/simulation-mode, edge-detection-cli, kanban-orchestrator, lucidhive-wp-publish · tags: simulation-mode, edge-detection, promotion-gate, paper-trading, kot-scoring, proven-edge, prediction-markets, polymarket, dashboard, task-tracker, ai-agents, digital-architecture, sovereign-infrastructure**
—
The previous articles in this series established two things: simulation layers let you test strategies without spending a dollar (S8.1), and content pipelines can be simulated end-to-end before a single word goes public (S8.2). Both articles share an assumption that deserves scrutiny: that the simulation itself is the hard part. It is not. The hard part is the moment after the simulation finishes and you have to decide whether to turn it on for real.
This article is about that moment. The promotion gate.
## What the promotion gate is
The promotion gate is a decision mechanism that sits between a simulation result and a live action. It answers one question: has this simulation demonstrated enough evidence to justify real-world execution?
This is not a formality. In the Kingdom of Truth’s architecture, every live action costs something — API credits, compute cycles, reputation, or money. The simulation-mode engine exists precisely because you should never spend those resources on an untested strategy. But a simulation that produces a result and then passes it straight to execution without a gate is just a faster way to lose money.
The promotion gate is the checkpoint. It takes simulation output — trades, content drafts, agent behaviors, strategy backtests — and applies a set of criteria before allowing promotion to live. The criteria are explicit, measurable, and non-negotiable.
## How KOT scoring implements the gate
The simulation-mode engine built on 2026-08-03 demonstrates a concrete implementation. Its KOT scoring system is a promotion gate for trading strategies. The mechanics:
A strategy runs against resolved Polymarket markets — real data, real timestamps, real outcomes. The execution engine fills orders at tape prices with configurable slippage (25 basis points default) and fees. Settlement happens against the market’s actual result: 1.0 if you held the winner, 0.0 if you did not.
After N trades (default 20), the scoring engine evaluates. Three criteria must all pass:
1. **Minimum sample size.** At least N trades — you cannot grade a strategy on 3 data points.
2. **Positive net expectancy.** After fees and slippage, the strategy must have made money on average per trade.
3. **Profit factor above 1.0.** Gross wins divided by gross losses must exceed 1.0 — you need more money coming in than going out.
If all three pass, the verdict is `PROVEN_EDGE`. If any fail, the verdict is `INCONCLUSIVE` — not `FAILURE`, because the strategy might work with more data or different parameters, but certainly not ready for live capital.
This is the promotion gate in action. The dual_trend strategy from the initial run scored PROVEN_EDGE with 40 trades, +$568 net, and a profit factor of 5.9. The mean_reversion strategy scored INCONCLUSIVE with 1 trade and -$100. The gate let one through and held the other. No human judgment required — the criteria are architectural.
## Why you need a gate, not just a threshold
A threshold is a number. A gate is a process. The difference matters.
A threshold says “only trade when edge > 7%.” A gate says “the edge detection CLI scans the market, calculates p_true from spot momentum, compares against the Polymarket price, and only signals when the edge exceeds the threshold — and then the simulation engine runs the strategy against historical data, and only when KOT scoring returns PROVEN_EDGE does the result pass to the next stage.”
The gate has multiple checkpoints. The edge-detection-cli provides the first filter: real-time market scanning with the edge formula (`edge = p_true – market_price`), expected value calculation (`EV = (p × profit) – (q × loss)`), and Kelly Criterion sizing (`f* = (p × b – q) / b`). A market that does not pass the edge threshold never enters the simulation. A market that passes the edge threshold enters the simulation, but the simulation’s own scoring gate must also pass before anything goes live.
This layered approach prevents two failure modes:
**False positives from thin data.** A single trade that wins does not mean the strategy works. The gate enforces minimum sample size before promotion.
**False confidence from backtesting.** A strategy that looks good on historical data but has not been tested against real execution conditions — slippage, fees, interleaved fills — fails the simulation gate because the simulation engine includes those real conditions.
## The dashboard layer
The task-tracker concept from the wiki describes a Dataview-powered dashboard that aggregates task metadata into auto-updating tables. The same pattern applies to simulation monitoring.
A promotion gate without observability is a black box. You need to see:
– Which strategies are currently in simulation
– How many trades each has completed
– What the current KOT score looks like
– Which strategies have passed the gate and are awaiting promotion
– Which strategies failed and why
This is where the dashboard becomes essential. The kanban board itself functions as a promotion gate dashboard — tasks move from `simulation` to `ready` to `running` to `done`, and each transition represents a gate check. The Obsidian task-tracker extends this with Dataview queries that pull frontmatter from simulation reports, strategy logs, and KOT scores into a single monitoring view.
Without the dashboard, promotion decisions become ad hoc. Someone remembers that the dual_trend strategy looked good and pushes it live. The dashboard makes the decision data-first: the score is PROVEN_EDGE, the sample is 40 trades, the profit factor is 5.9, and the gate criteria are met. The promotion is a consequence of evidence, not intuition.
## The live transition
When a simulation passes the promotion gate, the transition to live is not a single switch flip. It is a staged process:
**Stage 1: Paper trade with live data.** The strategy runs against real-time market data but executes with simulated money. This tests whether the simulation’s assumptions hold when the data is no longer historical — when you cannot see the outcome in advance.
**Stage 2: Shadow mode.** The strategy generates real signals and real position sizes, but does not execute. A human reviews the signals alongside the live market. This is the sanity check — does the strategy’s behavior in production match what the simulation predicted?
**Stage 3: Limited live.** The strategy executes with a fraction of the intended capital. Kelly Criterion sizing naturally limits this — the formula says “bet X% of bankroll,” and at 25% Kelly, that is a small number. The first live trades validate the entire pipeline: edge detection, simulation, KOT scoring, promotion gate, and execution.
**Stage 4: Full live.** Capital scales to the Kelly-optimal position. The dashboard monitors continuously. If the live performance diverges from the simulation’s predicted performance, the gate can reverse — demoting the strategy back to simulation or halting it entirely.
This staged approach is why the promotion gate is not a one-time check. It is an ongoing relationship between simulation and live. The gate does not open once and stay open. It monitors, scores, and can close again if conditions change.
## Why this matters beyond trading
The promotion gate pattern applies to any domain where simulation precedes action. Content pipelines: simulate an article’s SEO performance, engagement预测, and brand alignment before publishing. Agent behaviors: simulate a specialist’s task completion in an isolated workspace before allowing it to operate on production systems. Business models: simulate revenue projections with real market data before committing resources.
The simulation-mode engine proved this on Polymarket. The dual_trend strategy survived 40 real-data replays with positive expectancy. The mean_reversion strategy did not. The gate made the distinction without human bias. The same architecture — simulation, scoring, gate, staged promotion — works wherever the cost of being wrong justifies the cost of being careful.
The promotion gate is not about being slow. It is about being right the first time, because the second time costs real money.
—
