Free Options Data
for Backtesting用于回测的
免费期权数据

TL;DR核心摘要

Choose free options data that fits the test选择适合回测需求的免费期权数据

Define the market, underlyings, date range, interval, quote or trade requirement, and whether the strategy selects by delta, moneyness, or days to expiration. That specification determines whether a free route is usable.

先明确目标市场、标的资产、日期范围和数据时间粒度，以及需要报价数据还是成交记录；同时还要确定策略是依据 Delta、价内外程度还是距到期日天数来筛选合约。上述条件将决定免费数据渠道能否满足需求。

Control the biases that break options backtests控制会破坏期权回测的关键偏差

Rebuild the point-in-time contract universe重建严格时点的可选合约全集

At each decision time, select only contracts that were listed and observable then. Retain expired, delisted, and adjusted contracts instead of joining a present-day chain backward. Record listing date, expiration, root, multiplier, deliverable, settlement, and adjustment status so corporate actions do not silently change the instrument.

在每个决策时点，只能从当时已经挂牌且可见的合约中筛选。必须保留已到期、退市和调整后合约，不能把今天的期权链向历史日期反向拼接。记录挂牌日、到期日、root 代码、乘数、交割物、结算方式与调整状态，避免公司行动悄悄改变合约含义。

Keep every market-data clock separate分别保存每一种市场数据时间

A row can contain a fresh quote, an older last trade, prior-session open interest, and Greeks calculated later by a vendor. Store quote, trade, open-interest, underlying, analytics, and receipt times separately. The strategy may use a field only if its source time was available before the simulated decision.

同一行可能包含刚更新的报价、更早的最近成交、上一交易日未平仓量，以及供应商稍后计算的希腊值。报价、成交、未平仓量、标的行情、分析值和接收时间都应分开保存；只有源时间早于模拟决策时刻的字段，策略才可以使用。

Model contract selection before fills先模拟合约筛选，再模拟成交

If a rule chooses the nearest 30-delta call or 45-day contract, compute that choice from the contemporaneous chain and the model inputs available then. Do not use a vendor's revised Greek or a contract that became liquid later. Save the candidate set and rejection reasons with every signal.

如果规则选择最接近 30 Delta 的认购期权或距到期 45 天的合约，就必须基于当时期权链和当时可用的模型输入完成筛选。不能使用供应商后来修订的希腊值，也不能选择之后才获得流动性的合约。每个信号都应保存候选集和排除原因。

Use executable fill assumptions采用可执行的成交假设

• Reject crossed, locked, stale, or zero-sided quotes unless the strategy defines a special rule.
• Model buys from the ask side and sells from the bid side, then test midpoint and adverse-slippage scenarios separately.
• Cap order size by displayed size or a documented participation rule.
• Include contract multiplier, commissions, exchange fees, assignment, exercise, and early-exercise logic.
• Do not fill after the last valid quote or outside the simulated market session.

• 除非策略另有明确规则，否则应拒绝交叉报价、锁定报价、陈旧报价和单边为零的报价。
• 买入从卖价侧建模，卖出从买价侧建模，再单独测试中间价和不利滑点情景。
• 订单规模受显示挂单量或明确的参与率规则约束。
• 计入合约乘数、佣金、交易所费用、指派、行权和提前行权逻辑。
• 最后一条有效报价之后或模拟交易时段之外不得成交。

Separate research, validation, and test periods严格划分研究、验证和测试区间

Tune contract filters, volatility rules, and slippage assumptions only on the research period. Freeze the strategy before validation and keep a final untouched test period. Report results by regime, liquidity bucket, expiration type, and spread cost—not only one aggregate return.

合约筛选、波动率规则和滑点假设只能在研究区间调整；进入验证区间前冻结策略，并保留最终未接触的测试区间。结果应按市场状态、流动性分组、到期类型和价差成本分别报告，而不能只展示一个汇总收益率。

Python example: reject unrealistic option rowsPython 示例：剔除不合理的期权数据

Normalize the chain before selecting a contract. This local example keeps valid markets, computes a midpoint, and rejects crossed quotes.

选择合约前，应先将期权链标准化。以下本地示例仅保留有效报价、计算中间价，并剔除买价高于卖价的异常数据。

import pandas as pd

df = pd.read_csv("options_quotes.csv")
df["timestamp"] = pd.to_datetime(df["timestamp"], utc=True)
df = df.dropna(subset=["timestamp", "expiration", "strike", "bid", "ask"])
df = df[(df["bid"] >= 0) & (df["ask"] >= df["bid"])]
df["mid"] = (df["bid"] + df["ask"]) / 2
df = df.sort_values(["timestamp", "contract"]).drop_duplicates(
    ["timestamp", "contract"], keep="last"
)
print(df[["timestamp", "contract", "bid", "ask", "mid"]].tail())

Use QVeris to compare options-data capabilities使用 QVeris 比较期权数据能力

• Use the QVeris Playground to inspect one known symbol and expiration before mapping a schema.
• Ask explicitly for historical dates and expired contracts; never infer history from a current result.
• Save the tool ID, request, raw response, retrieval time, validation decision, and rejected assumptions.

• 映射数据结构前，先在 QVeris Playground 中检查一个已知标的和到期日。
• 明确测试历史日期和已到期合约，绝不能由当前结果推断历史能力。
• 保存工具 ID、请求、原始响应、抓取时间、校验结论和被否定的假设。

FAQ常见问题

References and next steps参考资料与下一步

QuantConnect plans
QuantConnect options documentation
Historical Option Data samples
Alpha Vantage historical options

QuantConnect 套餐
QuantConnect 期权文档
Historical Option Data 样本
Alpha Vantage 历史期权