trade-prediction-markets

交易预测市场

快速入门

此技能可在 Polymarket 预测市场上交易现实世界事件的 YES/NO 代币。

首先加载工具：

Use MCPSearch to select: mcp__workbench__get_all_prediction_events
Use MCPSearch to select: mcp__workbench__get_prediction_market_data
Use MCPSearch to select: mcp__workbench__create_prediction_market_strategy

基本工作流：

1. Browse markets:
   get_all_prediction_events(market_category="crypto_rolling")
   → See BTC/ETH price prediction markets

2. Analyze market data:
   get_prediction_market_data(condition_id="0x123...")
   → Study YES/NO token price history

3. Create strategy:
   create_prediction_market_strategy(
       strategy_name="PolymarketArb_M",
       description="Buy YES when price <40%, sell at 55%"
   )

4. Test strategy:
   run_prediction_market_backtest(
       strategy_name="PolymarketArb_M",
       ...
   )

适用场景：

• 交易现实世界事件（选举、美联储决策、体育赛事）
• 希望获得二元结果敞口（YES/NO）
• 对基于概率的交易感兴趣
• 探索预测市场机会

可用工具（6 个）

get_all_prediction_events

用途：浏览可用的 Polymarket 预测市场

参数：

• active_only（可选，布尔值）：仅显示活跃事件（默认：true）
• market_category（可选，字符串）：按类别筛选

类别：

• crypto_rolling：加密货币价格预测（下一个小时 BTC 会高于 $100k 吗？）
• politics：选举、政策决策
• economics：GDP、通胀、美联储决策
• sports：比赛结果、锦标赛
• entertainment：奖项、票房结果

返回：包含事件名称、类别、市场、条件 ID 和结算状态的事件列表

定价：$0.001

适用场景：发现交易机会、浏览可用市场

get_prediction_market_data

用途：分析特定市场中 YES/NO 代币的价格历史

参数：

• condition_id（必填）：Polymarket 条件 ID
• start_date（可选）：筛选起始日期（YYYY-MM-DD）
• end_date（可选）：筛选结束日期（YYYY-MM-DD）
• timeframe（可选）：K 线时间周期（1m、5m、15m、30m、1h、4h，默认：1m）
• limit（可选，1-10000）：每种代币的最大 K 线数量（默认：1000）

返回：市场元数据、YES 代币价格时间序列、NO 代币价格时间序列

定价：$0.001

适用场景：分析市场价格历史、研究代币行为、验证策略概念

create_prediction_market_strategy

用途：生成包含 YES/NO 交易逻辑的 Polymarket 策略代码

参数：

• strategy_name（必填）：策略名称（遵循格式：Name_RiskLevel）
• description（必填）：YES/NO 逻辑、退出标准和仓位规模的详细要求

返回：完整的 Python PolymarketStrategy 代码

定价：实际大模型成本 + 加价（最高 $4.50）

执行时间：约 30–60 秒

适用场景：构建新的 Polymarket 策略

run_prediction_market_backtest

用途：使用历史数据测试预测市场策略

参数：

• strategy_name（必填）：要测试的 PolymarketStrategy
• start_date（必填）：开始日期（YYYY-MM-DD）
• end_date（必填）：结束日期（YYYY-MM-DD）
• condition_id（单一市场）：特定条件 ID
• asset（滚动市场）：资产代码（"BTC"、"ETH"）
• interval（滚动市场）：市场周期（"15m"、"1h"）
• initial_balance（可选）：初始 USDC（默认：10000）
• timeframe（可选）：执行时间周期（默认：1m）

返回：回测指标（盈亏、胜率、仓位历史）

定价：$0.001

执行时间：约 20–60 秒

适用场景：验证预测市场策略

get_data_availability

用途：检查 Polymarket 市场的可用数据范围

参数：

• data_type："polymarket" 或 "all"
• asset（可选）：按资产筛选
• include_resolved（可选）：包含已结算市场

返回：包含日期范围的数据可用性信息

定价：$0.001

适用场景：回测前检查（确认数据充足）

get_latest_backtest_results

用途：查看近期预测市场回测结果

参数：

• strategy_name（可选）：按策略筛选
• limit（可选）：结果数量

返回：近期回测记录

定价：免费

适用场景：查看现有回测结果

核心概念

预测市场机制

YES/NO 代币结构：

Event: "Will BTC exceed $100,000 by end of hour?"

YES Token:
- Pays $1.00 if event occurs
- Pays $0.00 if event doesn't occur
- Current price = Market's implied probability
- Example: YES token at $0.65 = 65% implied probability

NO Token:
- Pays $1.00 if event DOESN'T occur
- Pays $0.00 if event occurs
- Current price = 1 - YES price
- Example: NO token at $0.35 = 35% implied probability

Total: YES price + NO price ≈ $1.00 (arbitrage if not)

交易方式：

Scenario: YES token at $0.40

Buy YES token:
- Pay $0.40 now
- If event occurs: Receive $1.00 (profit $0.60 = 150% return)
- If event doesn't occur: Lose $0.40 (-100% return)

Risk/Reward:
- Risking $0.40 to make $0.60
- 1.5:1 reward:risk ratio
- Need >40% win rate to break even

市场类别

加密货币滚动市场（高频）：

Type: Continuous prediction markets
Frequency: Every 15m, 1h, 4h, etc.
Question: "Will BTC price increase next [interval]?"

Example:
- 1h BTC rolling market
- New market every hour
- Predict if BTC closes higher than current price

Use case: Short-term price speculation
Trading style: Active, high frequency

政治（事件驱动）：

Type: One-time events
Frequency: Varies (elections, policy decisions)
Timeline: Days to months until resolution

Examples:
- "Will candidate X win election?"
- "Will bill Y pass Congress by date Z?"
- "Will Fed cut rates in next meeting?"

Use case: Event speculation
Trading style: Position trading, hold until resolution

经济（数据发布）：

Type: Scheduled data releases
Frequency: Monthly, quarterly
Timeline: Fixed resolution dates

Examples:
- "Will CPI exceed 3.5% next month?"
- "Will GDP growth exceed 2% this quarter?"
- "Will unemployment rate decrease?"

Use case: Economic data predictions
Trading style: Position before release, exit at resolution

体育（预定赛事）：

Type: Game outcomes, championships
Frequency: Varies by sport
Timeline: Hours to months

Examples:
- "Will Team X win game tonight?"
- "Will Player Y score >25 points?"
- "Will Team Z win championship?"

Use case: Sports betting alternative
Trading style: Event-based positions

策略类型

概率套利（均值回归）：

Concept: Buy underpriced probabilities, sell when corrected

Example:
- Event has ~60% true probability
- YES token priced at $0.45 (implies 45%)
- Buy YES (underpriced)
- Sell when price reaches $0.60 (fair value)

Advantages: Mathematical edge if probability estimation accurate
Disadvantages: Requires good probability estimation

趋势跟踪（动量）：

Concept: Follow YES/NO token price momentum

Example:
- YES token price rising from $0.30 → $0.45
- Buy YES (momentum continuing)
- Exit when momentum fades

Advantages: Captures strong moves
Disadvantages: Late entries, whipsaws

均值回归（区间交易）：

Concept: Fade extreme probability movements

Example:
- YES token spikes to $0.85 (85% implied)
- Seems too high, buy NO token ($0.15)
- Exit when reverts toward mean

Advantages: Profits from overreactions
Disadvantages: Catching falling knives (sometimes market is right)

事件驱动（催化剂交易）：

Concept: Trade based on news/catalysts

Example:
- Positive news for candidate X
- Buy YES token before market fully reacts
- Exit after market prices in news

Advantages: Early mover advantage
Disadvantages: Requires fast news reaction

滚动市场

滚动市场的运作方式：

BTC 1h Rolling Market:

Hour 1 (12:00-13:00):
- Market created at 12:00
- Question: "Will BTC close higher at 13:00 than 12:00?"
- YES/NO tokens trade 12:00-13:00
- Resolves at 13:00 based on price change

Hour 2 (13:00-14:00):
- New market created at 13:00
- Previous market resolved
- Profits/losses settled
- Process repeats

Strategy rolls from market to market automatically

滚动市场的优势：

• 持续的交易机会
• 更多回测数据（许多市场）
• 可预测的结算时间
• 适合算法交易

缺点：

• 频率越高，手续费越多
• 需要主动监控
• 结算时间更短（留给你判断正确的时间更少）

Polymarket 策略框架

必需方法：

class MyPolymarketStrategy(PolymarketStrategy):
    def should_buy_yes(self) -> bool:
        """Check if conditions met for YES token purchase"""
        # Return True to buy YES token

    def should_buy_no(self) -> bool:
        """Check if conditions met for NO token purchase"""
        # Return True to buy NO token

    def go_yes(self):
        """Execute YES token purchase with position sizing"""
        # Calculate position size
        # Buy YES token

    def go_no(self):
        """Execute NO token purchase with position sizing"""
        # Calculate position size
        # Buy NO token

可选方法：

    def should_sell_yes(self) -> bool:
        """Exit YES position"""
        # Return True to sell YES tokens

    def should_sell_no(self) -> bool:
        """Exit NO position"""
        # Return True to sell NO tokens

    def on_market_resolution(self):
        """Handle market settlement"""
        # Called when market resolves
        # Settle P&L

最佳实践

市场选择

选择流动性强的市场：

High liquidity: >$50k volume
- Tight spreads
- Easy entry/exit
- Reliable pricing

Low liquidity: <$10k volume
- Wide spreads
- Difficult exits
- Slippage risk

Recommendation: Start with high-volume markets

优先选择结算标准明确的市场：

GOOD: "Will BTC close above $100k at 5pm EST on Jan 1, 2025?"
- Objective resolution source (price data)
- Specific date and time
- No ambiguity

BAD: "Will crypto have a good year in 2025?"
- Subjective ("good" is undefined)
- Ambiguous resolution criteria
- Dispute risk

避免结果模糊的市场：

Check resolution source:
- Data-driven (prices, scores, votes) → Good
- Subjective judgment → Bad
- "Community decides" → High dispute risk

Research past market resolutions:
- Were resolutions fair?
- Any disputed outcomes?
- Market maker credibility

策略开发

定义明确的概率阈值：

Example: Probability arbitrage strategy

Entry logic:
- Buy YES if price <40% (undervalued)
- Buy NO if price <40% (YES >60%, overvalued)

Exit logic:
- Sell YES at 55% (15% profit target)
- Sell NO at 55% (symmetric)
- Stop loss at 25% (37.5% loss, preserve capital)

纳入头寸规模管理：

Fixed percentage:
- 5% of capital per market
- Max 10 simultaneous positions = 50% deployed
- Conservative, predictable

Kelly Criterion:
- Size based on edge and odds
- More aggressive, optimal growth
- Requires accurate probability estimation

设置退出标准：

Profit targets:
- Sell at X% gain (e.g., 15% above entry)

Time-based exits:
- Close position Y hours before resolution
- Avoid last-minute volatility

Stop losses:
- Sell if price drops below Z% (e.g., 60% of entry)
- Preserve capital on wrong predictions

风险管理

头寸限制：

Per market: 5-10% of capital
- Limits single-market exposure
- Diversifies risk

Total exposure: 50-70% of capital
- Leaves cash buffer
- Allows for new opportunities
- Prevents overtrading

市场多元化：

Don't concentrate in one category:
- 3 crypto markets
- 2 politics markets
- 2 sports markets
→ Diversified across event types

Avoid:
- 10 BTC rolling markets
→ All correlated, high concentration risk

流动性监控：

Check before entry:
- Current volume
- Bid/ask spread
- Order book depth

If liquidity drops:
- May be unable to exit
- Accept mark-to-market loss
- Or hold until resolution

常见工作流

工作流 1：探索滚动市场

目标：寻找 BTC 滚动市场交易机会

1. Browse crypto rolling markets:
   get_all_prediction_events(market_category="crypto_rolling")
   → Lists BTC, ETH rolling markets with intervals

2. Check data availability:
   get_data_availability(data_type="polymarket", asset="BTC")
   → Verify sufficient history for backtesting

3. Analyze specific market:
   get_prediction_market_data(
       condition_id="0x123...",
       timeframe="1m",
       limit=5000
   )
   → Study YES/NO token price patterns

4. Identify strategy:
   - YES token often overshoots (>60%)
   - Mean reversion opportunity
   - Buy NO when YES >65%, exit at 55%

5. Create strategy:
   create_prediction_market_strategy(
       strategy_name="BTCRollingMeanRev_M",
       description="Buy NO token when YES >65%, exit at 55%..."
   )

6. Backtest strategy:
   run_prediction_market_backtest(
       strategy_name="BTCRollingMeanRev_M",
       asset="BTC",
       interval="1h",
       start_date="2024-01-01",
       end_date="2024-12-31"
   )

成本：约 $2.50（$0.003 数据 + $2.50 策略创建）

工作流 2：事件驱动的政治交易

目标：在选举预测市场中交易

1. Browse politics markets:
   get_all_prediction_events(market_category="politics")
   → Find election markets

2. Analyze candidate X market:
   get_prediction_market_data(condition_id="election_123")
   → Study YES token price leading up to election

3. Identify pattern:
   - YES token very volatile
   - Spikes on good news, drops on bad news
   - Opportunities to buy dips, sell spikes

4. Create strategy:
   create_prediction_market_strategy(
       strategy_name="ElectionDipBuy_M",
       description="Buy YES when price drops >15% in 24h,
                   sell when recovers to pre-drop level..."
   )

5. Backtest (limited data for one-time events):
   - May have insufficient data for thorough backtest
   - Analyze manually or use similar past events

6. Trade carefully:
   - Event markets have less data
   - Higher uncertainty
   - Start with smaller position sizes

成本：约 $2.50

工作流 3：多市场投资组合

目标：构建多元化的预测市场投资组合

1. Identify multiple opportunities:
   - BTC 1h rolling (crypto)
   - Fed decision (economics)
   - Championship game (sports)

2. Create strategies for each:
   - Strategy 1: BTC rolling mean reversion
   - Strategy 2: Fed decision probability arbitrage
   - Strategy 3: Sports underdog value

3. Backtest all strategies:
   run_prediction_market_backtest(...) for each

4. Allocate capital:
   - BTC rolling: 15% (more data, higher confidence)
   - Fed decision: 10% (one-time event, moderate confidence)
   - Sports: 5% (less data, lower confidence)
   Total: 30% deployed, 70% cash

5. Monitor performance:
   - Track each strategy independently
   - Rebalance based on results
   - Stop underperformers

成本：约 $7.50（3 个策略）

故障排除

“未找到预测事件”

问题：get_all_prediction_events 返回空结果

解决方案：

• 尝试使用 active_only=False 查看已结算的市场
• 检查其他 market_category
• 市场可能暂时不可用

“市场数据不足”

问题：历史数据不足，无法回测

解决方案：

• 预测市场的历史数据比加密货币市场短
• 使用更短的回测周期（1-3 个月）
• 专注于滚动市场（数据点更多）
• 某些事件只发生一次（数据有限）

“策略表现不佳”

问题：回测显示亏损

解决方案：

• 预测市场效率较高（难以跑赢市场）
• 检查概率估算是否准确
• 确认策略逻辑是否合理
• 考虑手续费和滑点
• 可能需要更复杂的方法

后续步骤

创建预测市场策略后：

进行全面测试：

• 使用 test-trading-strategies 进行回测
• 在多个市场中验证
• 检查胜率和利润因子

优化策略：

• 使用 improve-trading-strategies 进行优化
• 优化阈值和参数
• 测试改进效果

实盘部署（支持后）：

• 目前仅支持模拟
• Polymarket 实盘部署即将推出
• 功能可用后将使用 deploy-live-trading

概要

此技能提供 Polymarket 预测市场交易：

• 6 个工具：浏览事件、分析数据、创建策略、回测
• 费用：数据费用为 $0.001，创建策略费用为 $1-$4.50
• 市场：政治、经济、体育、加密货币滚动市场
• 状态：仅支持模拟（即将支持实盘部署）

核心原则：预测市场针对现实世界事件交易 YES/NO 代币。要取得成功，需要准确估算概率并严格管理风险。

最佳实践：选择流动性充足且结果判定标准明确的市场，在不同事件类型间分散投资，合理控制仓位（每个市场为 5-10%），并设定止盈目标和止损位。

当前限制：暂不支持实盘部署。可用于回测和策略开发。未来更新将支持实盘交易。

注意：预测市场的定价效率很高。持续跑赢市场并不容易。请从模拟开始，在充分验证优势后再投入资金（待实盘部署可用时）。