BTC K14 Paper Trader

Run the complete BTC K14 strategy in paper mode with dynamic exits, a persistent ledger and a separate Simmer virtual SIM execution mirror.

This upgrades the former observation-only skill to the full, frozen K14 strategy in paper trading only. Read DISCLAIMER.md. No live trading switch, wallet signing, or real-money order endpoint is included. Never present hypothetical fills as exchange receipts or API connection-test trades as strategy performance.

Run

Provide SIMMER_API_KEY through the runtime's secret environment and a persistent K14_STATE_DIR. Never put secrets in this package, messages, logs or ClawHub. Install the pip dependencies from clawhub.json, then run:

sh

python paper_agent.py --action run

Default duration is 1800 seconds; K14_RUN_SECONDS accepts 10–86400. --state-dir and --seconds override those defaults. Reuse the same state directory across runs. A process lock rejects overlapping runs on the same filesystem. Managed runners with isolated disks must not run multiple copies for one API key; local locking does not establish distributed exclusivity.

python paper_agent.py --action status only reads the Simmer virtual portfolio and receipts. --action smoke spends up to 1 virtual SIM, buys then sells on an active Bitcoin simulation market to verify connectivity. Smoke uses sdk:k14-connection-test, distinct from strategy sdk:k14-paper-v020; it is never strategy profit evidence. Do not run smoke on a schedule. No user money is spent.

Create a file named STOP inside the state directory to stop the running process cleanly. Remove it only when explicitly resuming. Publication and an automaton manifest do not prove a Simmer-hosted process exists: retrieve its actual run report first.

Complete strategy

The bundled, hash-verified TrialSession and ParallelExit implement K14 on the last 14 completed Binance BTCUSDT 15-minute candles; frozen q=0.05 monthly thresholds; official Polymarket PTB; price/volatility/clock/book/fee validation; post-only Maker entry in the first 120 seconds; at most 5 shares and one entry window per persistent test account. Initial model cash is 50, daily risk 6 and total risk 20. Maximum entry price .48, minimum .30; opposite-leg price cap .50 and fee-inclusive paired cost cap .95.

Exit profile continuous-decay-v1 compares passive sell, direct FOK sell and opposite-side FOK buy then Merge. Target net profit per share at window ages 0/180/360/540/720 seconds is .12/.09/.06/.03/.01. Replacement waits for terminal receipts; partial fills retain exact quantity. No new buying after 840 seconds. At 900 seconds cancel and reconcile, then Redeem only after an official resolved outcome. Losing unresolved single-sided positions can settle to zero; this strategy has no automatic below-cost stop-loss.

The frozen October 2026 threshold file is not valid in November. Missing or expired thresholds block new signals. Never silently retune gates to make an entry happen.

Two distinct ledgers

1. paper.sqlite: complete-strategy model. Maker fills require a later fresh public book to cross strictly through the limit after a modeled 250 ms delay; touching a quote is not enough. Visible depth limits partial quantity. FOK is all-or-none on displayed depth. Cancellation/confirmation and Merge are simulated locally. Taker fees use the frozen model; Merge reserve is .00001/share. All receipts are labeled LOCAL_HYPOTHETICAL_TRADE_THROUGH_V1. These assumptions do not establish actual queue fills, latency or profitability.
2. simmer.sqlite: actual virtual SIM API requests and responses. On modeled BUY/SELL fills, the mirror submits venue="sim" with execute=false; it uses actual LMSR receipt quantities for later proportional sells. Source-market tokens must match. LMSR price, quantity, rounding and fills differ from Polymarket. Native Simmer simulation does not provide this strategy's Maker queue, cancel or Merge operations. A modeled Merge is recorded only in the local ledger; mirrored paired SIM positions remain for platform settlement. Do not add local Merge proceeds to the Simmer cash balance.

The mirror allows at most 2.5 SIM per buy and 6 SIM cumulative buys for this initial test, including the optional 1 SIM smoke. It pauses after ambiguous or rejected mirror execution for reconciliation; it never automatically resends an uncertain POST. Keys are attached only to api.simmer.markets; redirects are refused. Wallet/private-key endpoints and real venues are not exposed.

Inspect evidence

report.json shows process status, public feed health, K14 signal, pending modeled orders, model cash/fees/payouts, mirror status and zero real-money orders. simmer-status.json contains the virtual portfolio and real API receipts for virtual trades. paper.sqlite retains commands, hypothetical fills and settlements atomically with checkpoints. simmer.sqlite persists each request before transmission and records idempotency keys and unknown outcomes. Restart preserves cash, open positions and the one-window budget.

The reviewed core is bundled in vendor/, so it needs no local source repository. source-pins.json verifies integrity. portability-change.json records the only platform-import adjustment; entry and exit logic remain identical. Network disconnects, stale feeds, unknown settlement, virtual API rejection and lack of a current K14 signal are observable states, not permission to fabricate trades.