btc5m-web/docs/STRATEGIES-GUIDE.md

Strategy Guide

Version: v4.2.0 Author: Penguin Sensei · 岳 · @x_188888_x

⚠ Important Disclaimer

The strategies built into this tool are examples only, intended to demonstrate how to use the strategy framework, and cannot guarantee profits. The Polymarket BTC 5-minute market is highly volatile, and any strategy with fixed parameters carries the risk of becoming ineffective.

Recommendations:

• Run one or two windows with the smallest amount and observe whether the entry/exit logic matches your judgment
• Hover in the frontend to see each strategy's entry/exit conditions
• If you have good entry/exit ideas, new data patterns, or want to work on backtest optimization together, feel free to contact the author and refine them jointly, to achieve a 1+1 > 2 effect

---

Overview

There are currently 3 built-in example strategies (only the diff and momentum types are shown; the prob-chase type is not included as an example):

Key	Name	Type	Summary
D1	Diff 1 · Standard Enhanced	Diff	diff cross entry + trailing stop + drawdown take-profit + stepped take-profit
D2	Diff 2 · Tail Sweep	Diff	large-diff entry at the window tail + stepped take-profit
M1	Momentum 1	Momentum	6-factor scoring entry, holds to window end and is decided by settlement

Core principles:

• The authoritative state of automated strategies lives in the backend server.ts
• Buy confirmation relies on the local position localSize advanced by UserWS
• API positions are used only for reconciliation, releasing timed-out buy orders, and clearing residual positions after a sell
• Closing the frontend does not affect the backend strategy from continuing to run

---

Common Terms

• diff — Binance latest price - (PriceToBeat - BinanceOffset); the core indicator for diff-strategy entry
• upPct / dnPct — the current up/down order book implied probability
• rem — seconds remaining in the current 5-minute window
• localSize — the local position advanced by UserWS; both buy confirmation and sell tracking rely on it
• apiVerified — the API and local positions are aligned

---

Backend State Machine

• IDLE — no strategy is enabled
• SCANNING — scanning for entry conditions
• BUYING — buy triggered, order being sent
• WAIT_FILL — the first 10 seconds after the buy order is sent, only waiting for UserWS fill confirmation
• RECONCILING_FILL — not confirmed within 10 seconds, entering the deferred-confirmation state; after 15 seconds, only if the API also confirms no position does it return to SCANNING
• HOLDING — position confirmed, starting to run take-profit/stop-loss
• SELLING / WAIT_SELL_FILL — selling / waiting for sell confirmation
• DONE — round ended; when the position is not reconciled, it waits for API reconciliation before checking for residual positions

---

Strategy D1 · Standard Enhanced (Diff Type)

Entry Window

• Detected between 210s ~ 50s remaining

Entry Conditions

• Buy up: previous tick diff ≤ +35, current tick diff > +35, up probability < 80%
• Buy down: previous tick diff ≥ -35, current tick diff < -35, down probability < 80%

("Re-cross above/below triggers," not "buy whenever the current value is met")

Cooldown Lock (Prevents Chasing Highs and Flip-Flopping)

Neutral reset: |diff| ≤ 25 sustained for 3 seconds → release all cooldown locks

Single-direction lock (locks that direction if any is met, until returning to neutral):

• High-probability contamination seen first: while diff is within the trigger threshold, the up/down probability is already ≥ 80%
• Overheated: diff ≥ +55 and up probability ≥ 85% (buy-up direction) / diff ≤ -55 and down probability ≥ 85% (buy-down direction)

Exit Mechanisms (Multiple)

1. Stepped take-profit — rises linearly from 90% at 210s to 100% at 10s; sells when the current probability reaches the threshold of the moment
2. Drawdown take-profit — after the probability peak during holding reaches ≥ 85%, sells once it pulls back 8 percentage points
3. Trailing stop — enabled after a minimum holding of 3 seconds; triggered when diff pulls back 20 points from its peak
4. Backstop stop-loss — buy-up diff ≤ +5 / buy-down diff ≥ -5, stop out immediately
5. Forced close — when rem ≤ 10s: take profit if probability ≥ 70%, otherwise stop out

---

Strategy D2 · Tail Sweep (Diff Type)

Entry Window

• Detected between 60s ~ 1s remaining

Entry Conditions

• Buy up: diff > +50 and up probability < 95%
• Buy down: diff < -50 and down probability < 95%

Exit Mechanisms

Stepped take-profit (tightened in tiers by time remaining):

• rem ≥ 40s: probability ≥ 98%
• 20s ≤ rem < 40s: probability ≥ 99%
• 10s ≤ rem < 20s: probability ≥ 100%
• rem < 10s: hold to the end, decided by settlement

Stop-loss:

• Buy-up diff ≤ +5
• Buy-down diff ≥ -5

---

Strategy M1 · Momentum (Momentum Type)

Entry Window

• Detected when more than 60s remain (the final segment of the window does not participate in momentum evaluation)

Entry Logic

Based on 6-factor momentum scoring (see strategies/_core/s6-core.ts for details):

• RSI deviation
• Volume expansion
• 1-minute candle direction
• Price change magnitude
• Candle body ratio
• Number of consecutive same-color candles
• MA7 position
• (Auxiliary filter) MA120 long-term trend + 5-minute structure

An UP threshold triggers buy up, a DOWN threshold triggers buy down (the short threshold is stricter).

Exit Mechanisms

No take-profit, no stop-loss, no forced close; holds to the window end and the win/loss is decided by Polymarket settlement.

This is a "pure settlement" style strategy example: verifying "whether the momentum direction judgment is accurate" rather than "agonizing over mid-window take-profit/stop-loss."

---

Buy/Sell Confirmation and Residual-Position Handling

Buy Confirmation Flow

1. Strategy triggers → BUYING sends order → WAIT_FILL waits for UserWS
2. Not confirmed within 10 seconds: enter RECONCILING_FILL, keep waiting for UserWS
3. After 15 seconds, only if the API also confirms no position is the buy order released and it returns to scanning

Selling

• When selling an unaligned position, reserve a 0.05-share buffer to avoid insufficient balance
• If residual positions remain after API alignment, clear them again

---

Configuration Source

• On startup, strategy config is read from .env (STRATEGY_{D1,D2,M1}_ENABLED, etc.)
• Frontend changes only affect the current process and are not persisted across restarts
• After restart, .env still takes precedence

---

Usage Recommendations

• We recommend running on the premise that "the account has no position in the current window at startup"
• If you need to view status remotely, prefer APP_MODE=full + an SSH tunnel
• The example strategy parameters are all empirical values under historical data; please backtest and verify them yourself before live trading

---

Co-Development

Got a good strategy? Let's optimize it together! Contact the author: @x_188888_x