Skip to content

QuantStrategyLab/UsEquityStrategies

BranchesTags

Repository files navigation

UsEquityStrategies

English | 中文

English

Standalone us_equity strategy repository for QuantStrategyLab platforms.

This repository is the strategy layer: it owns pure signal, allocation, and target-computation logic plus strategy metadata. Downstream platform repositories still own broker adapters, order routing, schedule, secrets, and notifications.

Contract boundary

The current integration path is:

  • runtime profiles expose manifest-backed unified entrypoints
  • downstream platforms load those entrypoints through QuantPlatformKit
  • strategy outputs stay inside the shared StrategyDecision contract
  • broker-specific execution order, UI rows, and notification layout stay in platform repositories

Legacy strategy functions may still exist as internal adapters, but downstream runtimes should treat entrypoints/ and manifests as the supported integration surface.

Four-runtime authoring standard

Greenfield us_equity profiles should be authored once against the shared contract and be structurally portable across all four downstream runtimes from the first PR:

  • InteractiveBrokersPlatform
  • CharlesSchwabPlatform
  • LongBridgePlatform
  • PaperSignalPlatform

This means:

  • keep strategy math, required inputs, and target semantics shared here
  • add runtime adapters upstream instead of creating platform-local strategy forks
  • treat rollout enablement as a downstream platform decision, not a reason to omit shared portability by default

If one runtime is intentionally unsupported in a PR, call it out explicitly and keep the portability gap visible in review notes. Do not plan to “backfill paper/compatibility later” as the default workflow.

Authoring and portability guides

Strategy index

Canonical profile Display name Compatible platforms Cadence Benchmark Role Status
global_etf_rotation Global ETF Rotation InteractiveBrokersPlatform, CharlesSchwabPlatform, LongBridgePlatform, PaperSignalPlatform quarterly + daily canary VOO defensive_rotation runtime_enabled
global_etf_confidence_vol_gate Global ETF Confidence Vol Gate InteractiveBrokersPlatform, CharlesSchwabPlatform, LongBridgePlatform, PaperSignalPlatform quarterly + daily canary VOO defensive_rotation_research_candidate runtime_enabled
russell_1000_multi_factor_defensive Russell 1000 Multi-Factor InteractiveBrokersPlatform, CharlesSchwabPlatform, LongBridgePlatform, PaperSignalPlatform monthly SPY defensive_stock_baseline runtime_enabled
tech_communication_pullback_enhancement Tech/Communication Pullback Enhancement InteractiveBrokersPlatform, CharlesSchwabPlatform, LongBridgePlatform, PaperSignalPlatform monthly QQQ parallel_cash_buffer_branch runtime_enabled
mega_cap_leader_rotation_top50_balanced Mega Cap Leader Rotation Top50 Balanced InteractiveBrokersPlatform, CharlesSchwabPlatform, LongBridgePlatform, PaperSignalPlatform monthly QQQ balanced_leader_rotation runtime_enabled
tqqq_growth_income TQQQ Growth Income InteractiveBrokersPlatform, CharlesSchwabPlatform, LongBridgePlatform, PaperSignalPlatform daily QQQ offensive_dual_drive runtime_enabled
soxl_soxx_trend_income SOXL/SOXX Semiconductor Trend Income InteractiveBrokersPlatform, CharlesSchwabPlatform, LongBridgePlatform, PaperSignalPlatform daily SOXX sector_offensive_income runtime_enabled

runtime_enabled strategies are consumed by platform repositories through QuantPlatformKit strategy contracts and component loaders. Canonical profile keys are the runtime-facing layer; display names are the human-facing layer. Compatibility here means the strategy is structurally usable on that runtime stack, including brokerless paper/notify runtimes. Each deployment explicitly selects its strategy with STRATEGY_PROFILE; platform repositories own runtime-specific wiring.

Cadence here is the strategy-level intent. Platform repositories own the actual Cloud Scheduler / GitHub Actions cron settings:

  • daily profiles: run once per trading day near the US close.
  • daily profiles now also publish a runtime execution-timing contract through shared metadata:
    • signal_date
    • effective_date
    • execution_timing_contract
    • signal_effective_after_trading_days
  • the current daily runtime profiles (global_etf_rotation, global_etf_confidence_vol_gate, tqqq_growth_income, soxl_soxx_trend_income) are tagged as next_trading_day strategies at the strategy/runtime contract layer, so downstream runtimes and audit reports do not need to infer this from prose.
  • global_etf_rotation: evaluate canary risk daily, but perform normal rotation only on the last NYSE trading day of March, June, September, and December.
  • monthly snapshot profiles: publish feature snapshots monthly from UsEquitySnapshotPipelines, then execute once in the downstream runtime's monthly window.

Account-size suitability

Platform runtimes must adapt order sizing to broker capability. Schwab remains integer-share only; LongBridge supports fractional-share execution. IBKR remains whole-share by default because TWS API fractional order support must be verified per account/API path before enabling the runtime quantity step. Small accounts can still diverge from weight-based research backtests on integer-share platforms or symbols that do not support fractional execution. Live entrypoints do not hard-block small accounts, but they emit small_account_warning=true in diagnostics when account equity is below the suggested minimum.

Canonical profile Suggested minimum equity Small-account behavior
tqqq_growth_income 500 USD Most suitable for small accounts; TQQQ can usually trade, but BOXX/cash targets may drift.
soxl_soxx_trend_income 1000 USD Can run with drift on integer-share platforms; fractional-share runtimes can express the small SOXX/BOXX legs more closely.
global_etf_rotation (global_etf_confidence_vol_gate legacy alias) 3000 USD Top-2 ETF rotation can drift when selected ETFs are too expensive for the account; the legacy alias shares the same execution caveats.
mega_cap_leader_rotation_top50_balanced 10000 USD The fixed 50% Top2 / 50% Top4 sleeve blend can drift when integer shares cannot represent the intended unequal weights.
tech_communication_pullback_enhancement (qqq_tech_enhancement legacy alias) 10000 USD Small accounts reduce position count and single-name concentration rises.
russell_1000_multi_factor_defensive 30000 USD The default 24-stock basket is not suitable for small accounts.

The warning is advisory. It is meant to make dry-runs, Telegram messages, and reports explicit about the gap between account size and backtest assumptions.

Research candidates and archive

  • mega_cap_leader_rotation_top50_balanced: runtime-enabled monthly profile for the current Top50 balanced candidate. It consumes a transparent Top50 monthly snapshot and runs a fixed 50% Top2 cap50 sleeve plus 50% Top4 cap25 sleeve, with no broad QQQ trend de-risking by default.
  • global_etf_confidence_vol_gate: legacy alias and comparison name for the canonical global_etf_rotation runtime profile. It keeps the same universe and canary defense, uses the same SMA250 / z-gap / relative-volatility gate parameter set, and remains available for explicit regression checks.
  • mega_cap_leader_rotation: umbrella research/backtest name for the static and dynamic variants; see docs/research/mega_cap_leader_rotation.md.

global_etf_rotation

Objective

  • Keep a broad, lower-beta rotation framework for US equity accounts.
  • Stay open to leadership from tech and semiconductors without concentrating only in high-beta products.
  • Fall back to a short-duration safe haven when the cross-asset risk picture is weak.

Universe

  • 22 rotation ETFs: EWY, EWT, INDA, FXI, EWJ, VGK, VOO, XLK, SMH, GLD, SLV, USO, DBA, XLE, XLF, ITA, XLP, XLU, XLV, IHI, VNQ, KRE
  • Canary basket: SPY, EFA, EEM, AGG
  • Safe haven: BIL

Indicators and rules

  • Momentum uses Keller-style 13612W monthly momentum: (12×R1M + 4×R3M + 2×R6M + R12M) / 19.
  • Trend filter: candidate ETF must be above its 250-day SMA.
  • Hold bonus: an existing holding receives +2% score bonus to reduce turnover.
  • Daily canary check: if all 4 canary assets have negative or missing momentum, the strategy goes 100% BIL immediately.

Rebalance behavior

  • Normal rotation only happens on the last NYSE trading day of March, June, September, and December.
  • On a rebalance day, the strategy ranks the eligible universe and selects the top 2 ETFs.
  • Selected ETFs are normally equally weighted (50 / 50), but the default confidence gate can tilt them to 75 / 25 when the Top1 lead is strong and not materially more volatile than Top2.
  • If fewer than 2 names survive, the unused slot is parked in BIL.
  • On non-rebalance days, the strategy returns no target change unless the canary emergency path is triggered.

Confidence-volatility alias

  • global_etf_rotation is the canonical runtime profile and defaults to the SMA250 confidence-gated configuration.
  • global_etf_confidence_vol_gate now resolves to the same runtime profile and remains only as an explicit comparison alias for regression checks.
  • The gate concentrates only when Top1 is clearly ahead and not materially more volatile than Top2; otherwise it remains equal-weight Top2.

Why it exists

  • Compared with a pure tech or leveraged-Nasdaq approach, this profile is meant to be steadier.
  • It still allows VOO, XLK, and SMH to win their way into the rotation instead of hard-coding them out.

russell_1000_multi_factor_defensive

Objective

  • Provide a first stock-level US equity strategy that stays close to the current platform architecture.
  • Start with a price-only factor stack before adding fundamentals or ML reranking.
  • Keep execution realistic by consuming a precomputed feature snapshot instead of fetching 1000 symbols live during the rebalance run.

Universe

  • Point-in-time Russell 1000 constituent snapshot supplied by an upstream data task.
  • Benchmark row: SPY
  • Safe haven: BOXX

Signals and rules

  • Current V1 factors are price-only:
    • mom_6_1
    • mom_12_1
    • sma200_gap
    • vol_63
    • maxdd_126
  • Factors are standardized within sector, then combined into one total score.
  • Existing holdings receive a configurable hold bonus.
  • Market defense uses:
    • SPY trend (sma200_gap > 0)
    • breadth = share of eligible universe above 200MA

Portfolio behavior

  • Rebalance cadence is monthly in the downstream runtime.
  • Default stock exposure:
    • 100% in risk_on
    • 50% in soft_defense
    • 10% in hard_defense
  • Default position count is 24.
  • Unused capital is parked in BOXX.

Feature snapshot schema

  • Required price-history input columns:
    • symbol, as_of, close, volume
  • Required universe input columns:
    • symbol, sector
    • optional: start_date, end_date for point-in-time membership during backtests
  • Generated snapshot columns:
    • as_of, symbol, sector, close, volume, adv20_usd, history_days
    • mom_6_1, mom_12_1, sma200_gap, vol_63, maxdd_126, eligible

Snapshot pipeline ownership

Feature-snapshot generation, Russell 1000 input preparation, ranking artifacts, and the research backtest CLI now live in ../UsEquitySnapshotPipelines. This repo only owns the runtime strategy logic and catalog metadata.

Use the upstream repo for artifact jobs:

cd ../UsEquitySnapshotPipelines
PYTHONPATH=src:../UsEquityStrategies/src:../QuantPlatformKit/src python scripts/update_russell_1000_input_data.py \
  --output-dir data/input/refreshed/r1000_official_monthly_v2_alias \
  --universe-start 2018-01-01 \
  --price-start 2018-01-01 \
  --extra-symbols QQQ,SPY,BOXX
PYTHONPATH=src:../UsEquityStrategies/src:../QuantPlatformKit/src python scripts/build_russell_1000_feature_snapshot.py \
  --prices data/input/refreshed/r1000_official_monthly_v2_alias/r1000_price_history.csv \
  --universe data/input/refreshed/r1000_official_monthly_v2_alias/r1000_universe_history.csv \
  --output-dir data/output/russell_1000_multi_factor_defensive
PYTHONPATH=src:../UsEquityStrategies/src:../QuantPlatformKit/src python scripts/backtest_russell_1000_multi_factor_defensive.py \
  --prices data/input/refreshed/r1000_official_monthly_v2_alias/r1000_price_history.csv \
  --universe data/input/refreshed/r1000_official_monthly_v2_alias/r1000_universe_history.csv \
  --output-dir data/output/russell_1000_multi_factor_defensive_backtest

The backtest output directory still includes summary.csv, portfolio_returns.csv, weights_history.csv, and turnover_history.csv.

tqqq_growth_income

Objective

  • Run the default no-income QQQ / TQQQ dual-drive growth profile.
  • Keep the legacy income and BOXX symbols in the managed universe so existing holdings can be reduced cleanly.

Portfolio layers

  • Growth layer: QQQ and TQQQ; broker runtimes can replace the unlevered growth sleeve with a lower-price proxy such as QQQM while keeping QQQ as the signal source.
  • Default active reserve: 2% cash plus 8% BOXX
  • Legacy / cleanup layer: BOXX, SPYI, QQQI

Signals and indicators

  • Uses daily QQQ history as the signal source.
  • dual_drive_unlevered_symbol controls the tradable unlevered growth sleeve and defaults to QQQ.
  • The default configuration uses MA200, MA20, and positive MA20 slope.
  • Retired ATR-staged sizing has been removed from the TQQQ profile; fixed_qqq_tqqq_pullback is the only supported allocation mode.

Default dual-drive rules (QQQ / TQQQ)

  • Entry requires QQQ > MA200 and positive MA20 slope.
  • Once risk is active, the profile keeps QQQ 45% / TQQQ 45% / BOXX 8% / cash 2% while QQQ remains above MA200; a short-term negative MA20 slope alone does not force an exit.
  • If QQQ falls below MA200, the profile exits QQQ and TQQQ, keeps 2% cash, and parks the rest in BOXX by default.
  • A below-MA200 pullback state can still re-enable risk when QQQ > MA20, MA20 slope is positive, and QQQ has rebounded from its rolling 20-day low by more than the dynamic volatility-scaled gate. The default gate is 2.0x the recent 20-day QQQ daily return volatility, which avoids a fixed 3% constant while still filtering weak MA200 chop without changing the normal above-MA200 trend rule.

Income-layer rules (SPYI / QQQI)

  • The default configuration sets income_threshold_usd = 1_000_000_000, so the income layer is disabled for normal account sizes.
  • Lowering that threshold opts back into the legacy income sleeve.
  • QQQI_INCOME_RATIO still decides the split between QQQI and SPYI when the income layer is enabled.

Defense behavior (BOXX and cash)

  • The fixed dual-drive configuration keeps a small cash buffer and uses BOXX for the remaining idle capital.
  • BOXX remains a managed symbol so old BOXX holdings can be traded down if present.
  • Downstream execution decides whether the gap to target is large enough to trade via a rebalance threshold.

Default runtime profile settings

  • ATTACK_ALLOCATION_MODE = fixed_qqq_tqqq_pullback
  • DUAL_DRIVE_QQQ_WEIGHT = 0.45, DUAL_DRIVE_TQQQ_WEIGHT = 0.45
  • DUAL_DRIVE_UNLEVERED_SYMBOL = QQQ
  • DUAL_DRIVE_CASH_RESERVE_RATIO = 0.02
  • DUAL_DRIVE_PULLBACK_REBOUND_WINDOW = 20
  • DUAL_DRIVE_PULLBACK_REBOUND_THRESHOLD_MODE = volatility_scaled
  • DUAL_DRIVE_PULLBACK_REBOUND_VOLATILITY_MULTIPLIER = 2.0
  • DUAL_DRIVE_PULLBACK_REBOUND_THRESHOLD = 0.0 (fixed-mode fallback only)
  • INCOME_THRESHOLD_USD = 1000000000
  • CASH_RESERVE_RATIO = 0.02
  • EXECUTION_CASH_RESERVE_RATIO = 0.0
  • REBALANCE_THRESHOLD_RATIO = 0.01

soxl_soxx_trend_income

Objective

  • Use the optimized SOXX-gated tiered blend profile for semiconductor exposure.
  • Keep a dedicated income sleeve for larger accounts without forcing that sleeve to shrink during normal trading-layer changes.

Portfolio layers

  • Trading layer: SOXL, SOXX, BOXX
  • Income layer: QQQI, SPYI

Trading-layer rules

  • The default runtime mode uses a tiered SOXX trend gate to avoid relying on one all-or-nothing threshold.
  • If SOXX > MA140 * 1.08, the core sleeve targets SOXL 70% + SOXX 20%.
  • If SOXX > MA140 * 1.06, or an existing SOXL sleeve has not broken MA140 * 0.98, the core sleeve targets SOXL 65% + SOXX 20%.
  • If the gate is off, the core sleeve holds defensive SOXX 15%.
  • Overheat controls are active on the live profile: when the base tier is full or mid, SOXX RSI14 above the effective threshold and/or a break above the upper Bollinger band downgrade the tier by one step per trigger.
  • The live RSI threshold is dynamic: max(70, prior 252 trading days RSI14 90th percentile), with 70 as the fallback floor when the dynamic indicator is unavailable.
  • Unused trading-layer capital is parked in BOXX.

Sizing behavior

  • The tiered gate directly sets core-sleeve exposure: full, mid, or defensive.
  • There is no separate account-size deploy-ratio decay in the SOXL/SOXX profile.
  • The downstream runtime also keeps a cash reserve and only trades when the rebalance gap is large enough.

Income-layer rules

  • The income layer starts only after total strategy equity crosses income_layer_start_usd.
  • It ramps linearly to income_layer_max_ratio by 2x that threshold.
  • Existing income holdings are locked with max(current_income_layer_value, desired_income_layer_value), so the layer only adds capital instead of force-selling down.
  • New income allocation is split by configurable QQQI / SPYI weights.

Default runtime profile settings

  • TREND_MA_WINDOW = 140
  • CASH_RESERVE_RATIO = 0.03
  • MIN_TRADE_RATIO = 0.01, MIN_TRADE_FLOOR = 100 USD
  • REBALANCE_THRESHOLD_RATIO = 0.01
  • ATTACK_ALLOCATION_MODE = soxx_gate_tiered_blend
  • BLEND_GATE_SOXL_WEIGHT = 0.70, BLEND_GATE_MID_SOXL_WEIGHT = 0.65
  • BLEND_GATE_ACTIVE_SOXX_WEIGHT = 0.20, BLEND_GATE_DEFENSIVE_SOXX_WEIGHT = 0.15
  • RSI overheat enabled with dynamic threshold max(70, rolling 252d RSI14 q90)
  • Bollinger overheat enabled; stacked RSI + Bollinger triggers can downgrade full directly to defensive
  • Gate buffers: entry 8%, mid 6%, exit 2%
  • Income layer starts at 150000 USD, caps at 15%
  • Income split: QQQI 70%, SPYI 30%

中文

这是 QuantStrategyLab 的独立美股策略仓。

这个仓库负责纯策略层：信号、仓位、目标权重计算，以及策略元数据。下游平台仓库继续负责券商适配、下单方式、调度、密钥和通知。

契约边界

当前主线集成方式已经固定为：

  • runtime profile 暴露 manifest 驱动的统一 entrypoint
  • 下游平台通过 QuantPlatformKit 加载这些 entrypoint
  • 策略输出保持在共享 StrategyDecision 契约内
  • 券商专属执行顺序、UI 展示行和通知布局继续留在平台仓库

旧策略函数可以继续作为仓库内部 adapter 存在，但下游运行时应把 entrypoints/ 和 manifest 当成正式接入面。

编写与可移植性文档

策略索引

Canonical profile 显示名 兼容平台仓库 策略频率 核心思路
global_etf_rotation 全球 ETF 轮动 InteractiveBrokersPlatform, CharlesSchwabPlatform, LongBridgePlatform, PaperSignalPlatform 季度调仓 + 每日 canary 22 只全球 ETF 的季度 Top 2 轮动，默认使用 SMA250 置信度 + 相对波动门控
global_etf_confidence_vol_gate 全球 ETF 置信度波动过滤 InteractiveBrokersPlatform, CharlesSchwabPlatform, LongBridgePlatform, PaperSignalPlatform 季度调仓 + 每日 canary 与 global_etf_rotation 同源的显式对照入口，保留 75/25 规则与门控参数
russell_1000_multi_factor_defensive 罗素1000多因子 InteractiveBrokersPlatform, CharlesSchwabPlatform, LongBridgePlatform, PaperSignalPlatform 月频 Russell 1000 个股月频 price-only 选股，带 SPY + breadth 防守和 BOXX 停泊
tech_communication_pullback_enhancement 科技通信回调增强 InteractiveBrokersPlatform, CharlesSchwabPlatform, LongBridgePlatform, PaperSignalPlatform 月频 tech-heavy 月频个股选择，做受控回调，并显式保留 BOXX 缓冲
mega_cap_leader_rotation_top50_balanced Mega Cap Top50 平衡龙头轮动 InteractiveBrokersPlatform, CharlesSchwabPlatform, LongBridgePlatform, PaperSignalPlatform 月频 当前 Top50 平衡候选，固定 50% Top2 cap50 + 50% Top4 cap25，不因 QQQ 趋势默认降仓
tqqq_growth_income TQQQ 增长收益 InteractiveBrokersPlatform, CharlesSchwabPlatform, LongBridgePlatform, PaperSignalPlatform 日频 QQQ / TQQQ 双轮增长，默认 45% / 45% / 8% BOXX / 2% 现金
soxl_soxx_trend_income SOXL/SOXX 半导体趋势收益 InteractiveBrokersPlatform, CharlesSchwabPlatform, LongBridgePlatform, PaperSignalPlatform 日频 SOXL / SOXX 趋势切换，剩余资金停在 BOXX，并叠加收入层

runtime_enabled 策略通过 QuantPlatformKit 提供的策略契约和组件加载接口被各个平台仓库引用；research_only profile 保留定义、manifest、entrypoint 和 adapter 作为研究/回放存档，但不会进入平台 rollout allowlist。运行时和部署配置统一使用 canonical profile key。这里的“兼容平台”既包括三个 live broker runtime，也包括 brokerless 的 PaperSignalPlatform。 这里的策略频率表达的是策略层意图；实际 Cloud Scheduler / GitHub Actions cron 配置由各个平台仓库负责：

  • 日频策略：每个美股交易日临近收盘运行一次。
  • global_etf_rotation：每日检查 canary 风险，但正常轮动只在 3 / 6 / 9 / 12 月最后一个 NYSE 交易日触发；默认采用 SMA250 置信度 + 相对波动门控。
  • 月频 snapshot 策略：由 UsEquitySnapshotPipelines 按月发布 feature snapshot，再由下游运行时在月度窗口内执行一次。

小资金适用性

平台运行时必须按券商能力适配下单数量。Schwab 仍按整数股执行；LongBridge 支持碎股执行。IBKR 默认仍按整数股执行，因为 TWS API 是否能接受碎股单需要按账户/API 路径实测确认后才能打开运行时数量步进。小账户在整数股平台或不支持碎股的标的上仍会明显偏离按权重回测得到的收益和回撤。live entrypoint 不会硬性禁止小账户运行，但当账户净值低于建议资金时，会在 diagnostics 里输出 small_account_warning=true。

Canonical profile 建议最低资金 小资金表现
tqqq_growth_income 500 USD 最适合小账户；通常能买到 TQQQ，但 BOXX / 现金层会有偏差。
soxl_soxx_trend_income 1000 USD 整数股平台会有偏离；支持碎股的运行时可以更接近小额 SOXX / BOXX 目标仓位。
global_etf_rotation 3000 USD 默认档已切到 SMA250 + 置信度门控；Top2 ETF 轮动遇到高价 ETF 时仍会偏离。
global_etf_confidence_vol_gate 3000 USD 与 global_etf_rotation 同一执行约束；保留 75/25 对照入口时，小账户和整数股平台会更容易产生偏离。
mega_cap_leader_rotation_top50_balanced 10000 USD 固定 50% Top2 / 50% Top4 袖珍组合需要不等权持仓，小账户整数股会产生明显偏离。
tech_communication_pullback_enhancement（历史别名 qqq_tech_enhancement） 10000 USD 小账户会降低持仓数，单票集中度上升。
russell_1000_multi_factor_defensive 30000 USD 默认 24 只股票组合，不适合小账户。

这个提示只是软警告。目的是让 dry-run、Telegram 通知和报告明确显示：当前账户资金量和研究回测假设之间存在差距。

研究候选与存档

  • mega_cap_leader_rotation_top50_balanced：已注册为 runtime-enabled 月频 profile，消费透明 Top50 月度 snapshot，运行固定 50% Top2 cap50 + 50% Top4 cap25 的组合，不默认使用宽基趋势降仓。
  • global_etf_confidence_vol_gate：global_etf_rotation 的显式对照入口，保留同一标的池和 canary 防守。它使用同样的 SMA250 / z-gap / 相对波动门控参数，方便做回归对照和小样本验证。
  • mega_cap_leader_rotation：静态池和动态池的研究/回测总称；说明见 docs/research/mega_cap_leader_rotation.md。

global_etf_rotation

策略目标

  • 给美股账户提供一个更分散、波动更低的轮动框架。
  • 不把科技和半导体硬排除在外，但也不把风险全部集中到高弹性品种上。
  • 当跨资产风险明显转弱时，退回短久期避险仓位。

标的池

  • 22 只轮动 ETF：EWY、EWT、INDA、FXI、EWJ、VGK、VOO、XLK、SMH、GLD、SLV、USO、DBA、XLE、XLF、ITA、XLP、XLU、XLV、IHI、VNQ、KRE
  • Canary 篮子：SPY、EFA、EEM、AGG
  • 避险资产：BIL

指标和规则

  • 动量使用 Keller 风格的 13612W 月频动量：(12×R1M + 4×R3M + 2×R6M + R12M) / 19。
  • 趋势过滤：候选 ETF 必须站上 200 日均线。
  • 持有加分：当前持仓会获得 +2% 分数加成，用来降低换手。
  • 每日 canary 检查：如果 SPY / EFA / EEM / AGG 这 4 个资产的动量全部为负，或缺失到全部失效，就立刻切到 100% BIL。

调仓行为

  • 正常轮动只在 3 / 6 / 9 / 12 月最后一个 NYSE 交易日触发。
  • 到调仓日后，对合格标的打分，选出前 2 名。
  • 前 2 名等权配置，默认 50 / 50。
  • 如果合格标的不满 2 个，空出来的部分停到 BIL。
  • 非调仓日默认不改目标仓位，除非触发 canary 应急防守。

置信度 + 波动过滤变体

  • global_etf_rotation 现在默认使用 sma_period=250、confidence_threshold=1.0、confidence_weighting_enabled=True、confidence_top1_weight=0.75、confidence_volatility_gate_enabled=True、confidence_volatility_window=126、confidence_volatility_max_ratio=1.3。
  • global_etf_confidence_vol_gate 保留为显式对照入口，参数与默认档一致，便于回放和回归比较。
  • 只有 Top1 明显领先且相对 Top2 不显著更高波动时，才切到 75 / 25；否则仍保持 Top2 等权。

这套策略的定位

  • 相比纯科技或者杠杆纳指路线，这个档位更稳。
  • 但它仍然允许 VOO、XLK、SMH 靠表现进入组合，而不是事先把它们排除。

russell_1000_multi_factor_defensive

策略目标

  • 作为第一版个股策略，先尽量复用现有平台边界。
  • 第一阶段只用价格因子，不急着上基本面和机器学习。
  • 运行时只消费预先算好的 feature snapshot，不在调仓时现场拉 1000 只股票历史数据。

股票池

  • 上游数据任务提供的 Russell 1000 点时成分快照
  • 基准行：SPY
  • 防守资产：BOXX

当前 V1 因子

  • mom_6_1
  • mom_12_1
  • sma200_gap
  • vol_63
  • maxdd_126

策略先在行业内做标准化，再合成总分。当前持仓可以拿到一小段 hold bonus。

防守规则

  • SPY 的 sma200_gap > 0 代表 benchmark 趋势正常
  • breadth = 合格股票里站上 200MA 的比例
  • 默认风险暴露：
    • risk_on：100%
    • soft_defense：50%
    • hard_defense：10%

组合规则

  • 下游运行时按月调仓
  • 默认持仓数 24
  • 剩余资金停在 BOXX

feature snapshot 输入/输出约定

  • 价格历史输入列：
    • symbol、as_of、close、volume
  • 股票池输入列：
    • symbol、sector
    • 可选：start_date、end_date（用于回测时按日期启用 / 退出成分股）
  • 生成后的 snapshot 列：
    • as_of、symbol、sector、close、volume、adv20_usd、history_days
    • mom_6_1、mom_12_1、sma200_gap、vol_63、maxdd_126、eligible

Snapshot 流水线归属

Feature snapshot 生成、Russell 1000 输入数据准备、ranking 产物和研究回测 CLI 已迁移到 ../UsEquitySnapshotPipelines。 本仓库只保留运行时策略逻辑和策略目录元数据。

产物任务请在上游仓库执行：

cd ../UsEquitySnapshotPipelines
PYTHONPATH=src:../UsEquityStrategies/src:../QuantPlatformKit/src python scripts/update_russell_1000_input_data.py \
  --output-dir data/input/refreshed/r1000_official_monthly_v2_alias \
  --universe-start 2018-01-01 \
  --price-start 2018-01-01 \
  --extra-symbols QQQ,SPY,BOXX
PYTHONPATH=src:../UsEquityStrategies/src:../QuantPlatformKit/src python scripts/build_russell_1000_feature_snapshot.py \
  --prices data/input/refreshed/r1000_official_monthly_v2_alias/r1000_price_history.csv \
  --universe data/input/refreshed/r1000_official_monthly_v2_alias/r1000_universe_history.csv \
  --output-dir data/output/russell_1000_multi_factor_defensive
PYTHONPATH=src:../UsEquityStrategies/src:../QuantPlatformKit/src python scripts/backtest_russell_1000_multi_factor_defensive.py \
  --prices data/input/refreshed/r1000_official_monthly_v2_alias/r1000_price_history.csv \
  --universe data/input/refreshed/r1000_official_monthly_v2_alias/r1000_universe_history.csv \
  --output-dir data/output/russell_1000_multi_factor_defensive_backtest

回测输出目录仍然会包含 summary.csv、portfolio_returns.csv、weights_history.csv、turnover_history.csv。

tqqq_growth_income

策略目标

  • 默认配置采用不带收入层的 QQQ / TQQQ 双轮增长策略。
  • 继续把旧收入层和 BOXX 资产留在管理列表里，方便把已有持仓平滑降下来。

资产层级

  • 增长层：QQQ、TQQQ；券商运行时可以把非杠杆增长袖子换成低单价代理，例如 QQQM，但主信号仍使用 QQQ。
  • 默认激活时：2% 现金加 8% BOXX
  • 旧持仓清理 / 兼容层：BOXX、SPYI、QQQI

信号和指标

  • 以 QQQ 的日线数据作为主信号源。
  • dual_drive_unlevered_symbol 控制实际交易的非杠杆增长袖子，默认是 QQQ。
  • 默认配置使用 MA200、MA20 和正向 MA20 斜率。
  • 旧 ATR 分段仓位已经从 TQQQ profile 移除；当前只支持 fixed_qqq_tqqq_pullback。

默认双轮规则（QQQ / TQQQ）

  • 入场需要 QQQ > MA200 且 MA20 斜率为正。
  • 一旦进入风险状态，只要 QQQ 仍在 MA200 上方，就维持 QQQ 45% / TQQQ 45% / BOXX 8% / 现金 2%；短期 MA20 斜率转负不会单独触发离场。
  • 如果 QQQ 跌破 MA200，默认退出 QQQ 和 TQQQ，保留 2% 现金，其余转入 BOXX。
  • 在 MA200 下方也保留一段回调参与逻辑：当 QQQ > MA20、MA20 斜率为正，且 QQQ 较滚动 20 日低点的反弹幅度超过动态波动率门槛时，可重新打开风险仓位。默认门槛是最近 20 日 QQQ 日收益波动率的 2.0x，避免使用固定 3% 常数，同时继续过滤较弱的 MA200 附近震荡，不改变 MA200 上方的主趋势规则。

收入层规则（SPYI / QQQI）

  • 默认配置把 income_threshold_usd 设为 1_000_000_000，普通账户规模下等于关闭收入层。
  • 如果以后要重新启用收入层，可以把这个阈值调低。
  • QQQI_INCOME_RATIO 仍然决定收入层启用时 QQQI 和 SPYI 的拆分比例。

防守行为（BOXX 与现金）

  • fixed dual-drive 默认配置只保留一小部分现金，剩余闲置资金进入 BOXX。
  • BOXX 仍保留为管理资产，方便清理旧 BOXX 持仓。
  • 是否真的下单，由下游执行层再结合再平衡阈值判断。

默认运行 profile 配置值

  • ATTACK_ALLOCATION_MODE = fixed_qqq_tqqq_pullback
  • DUAL_DRIVE_QQQ_WEIGHT = 0.45，DUAL_DRIVE_TQQQ_WEIGHT = 0.45
  • DUAL_DRIVE_UNLEVERED_SYMBOL = QQQ
  • DUAL_DRIVE_CASH_RESERVE_RATIO = 0.02
  • DUAL_DRIVE_PULLBACK_REBOUND_WINDOW = 20
  • DUAL_DRIVE_PULLBACK_REBOUND_THRESHOLD_MODE = volatility_scaled
  • DUAL_DRIVE_PULLBACK_REBOUND_VOLATILITY_MULTIPLIER = 2.0
  • DUAL_DRIVE_PULLBACK_REBOUND_THRESHOLD = 0.0（仅作为 fixed 模式 fallback）
  • INCOME_THRESHOLD_USD = 1000000000
  • CASH_RESERVE_RATIO = 0.02
  • EXECUTION_CASH_RESERVE_RATIO = 0.0
  • REBALANCE_THRESHOLD_RATIO = 0.01

soxl_soxx_trend_income

策略目标

  • 使用优化后的 SOXX 趋势分层闸门半导体策略。
  • 给大账户保留收入层，但不因为交易层切换就强制把收入层减回来。

资产层级

  • 交易层：SOXL、SOXX、BOXX
  • 收入层：QQQI、SPYI

交易层规则

  • 默认运行配置使用 SOXX 趋势分层闸门，避免仓位完全依赖单一开关。
  • 如果 SOXX > MA140 * 1.08，核心层目标为 SOXL 70% + SOXX 20%。
  • 如果 SOXX > MA140 * 1.06，或已有 SOXL 仓位尚未跌破 MA140 * 0.98，核心层目标为 SOXL 65% + SOXX 20%。
  • 如果趋势闸门关闭，核心层防守目标为 SOXX 15%。
  • 线上 profile 启用过热控制：基础档位为 full 或 mid 时，如果 SOXX RSI14 高于有效阈值，和/或价格突破布林上轨，会按触发项逐级降档。
  • 线上 RSI 阈值为动态阈值：max(70, 过去 252 个交易日 RSI14 的 90% 分位数)；动态指标缺失时以 70 作为 fallback floor。
  • 交易层没有部署出去的资金停在 BOXX。

仓位规则

  • 分层闸门直接决定核心层风险暴露：full、mid 或 defensive。
  • SOXL/SOXX profile 不再保留单独的账户规模 deploy-ratio 衰减。
  • 下游运行层另外还会保留现金储备，并且只有偏离目标足够大时才触发调仓。

收入层规则

  • 总策略权益超过 income_layer_start_usd 才启动收入层。
  • 到 2 倍阈值 时，收入层线性抬升到 income_layer_max_ratio。
  • 收入层采用 max(current_income_layer_value, desired_income_layer_value) 锁定已有收入资产，所以默认只增配，不主动减配。
  • 新增收入资金按可配置的 QQQI / SPYI 比例拆分。

默认运行 profile 配置值

  • TREND_MA_WINDOW = 140
  • CASH_RESERVE_RATIO = 0.03
  • MIN_TRADE_RATIO = 0.01，MIN_TRADE_FLOOR = 100 USD
  • REBALANCE_THRESHOLD_RATIO = 0.01
  • ATTACK_ALLOCATION_MODE = soxx_gate_tiered_blend
  • BLEND_GATE_SOXL_WEIGHT = 0.70，BLEND_GATE_MID_SOXL_WEIGHT = 0.65
  • BLEND_GATE_ACTIVE_SOXX_WEIGHT = 0.20，BLEND_GATE_DEFENSIVE_SOXX_WEIGHT = 0.15
  • RSI 过热已启用，动态阈值为 max(70, rolling 252d RSI14 q90)
  • 布林带过热已启用；RSI + 布林带双触发时，full 可直接降到 defensive
  • 闸门缓冲：入场 8%，中档 6%，退出 2%
  • 收入层起点 150000 USD，上限 15%
  • 收入层配比：QQQI 70%，SPYI 30%

About

Shared us_equity strategy implementations and execution metadata for QuantStrategyLab platforms, including ETF rotation and income-oriented allocation strategies.

Topics

python trading quant strategies asset-allocation etf-rotation us-equity

Resources

Readme

License

MIT license

Contributing

Contributing

Security policy

Security policy
Activity
Custom properties

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

37 tags

Packages

 
 
 

Contributors

Languages