A fully systematic digital-asset fund trading major crypto perpetual futures, long and short, executed end-to-end by machine. Twenty-one production strategies across four independent signal families — from 15-minute order-flow to daily trend — validated on six years of market history under strict out-of-sample discipline.
全系统化数字资产基金，多空双向交易主流加密货币永续合约，全流程机器执行。21 项生产策略、四大相互独立的信号体系——从 15 分钟级订单流到日线趋势——在六年完整市场历史上以严格的样本外纪律完成验证。
Cross-venue spreads capture institutional capital flow — a leading signal across major digital assets. 跨交易所价差，捕捉机构资金流向，领先主流资产行情。
Identify structural trend reversals, filter false breakouts, raise hit rate. 识别趋势结构反转，过滤假突破，提升胜率。
Adapt to market state — activate only in suitable regimes, controlling drawdowns. 自适应市场状态，只在有效环境下开仓，控制回撤。
Our 21-strategy production portfolio. In-sample backtest Sharpe 4.27 (2022–2025), validated out-of-sample on Jun 2020–2021 (Sharpe 2.94) and 2026 YTD forward OOS (Sharpe 5.38). 21 项策略组合。样本内回测夏普 4.27（2022–2025）；样本外覆盖 2020 年 6 月–2021（夏普 2.94）与 2026 年迄今前向样本外（夏普 5.38）。
Targets below are anchored on actual 2026 out-of-sample results through Jul 31 (212 days, true OOS), with conservative haircuts applied. Numbers are presented as target live / in-sample backtest. 以下目标基于 2026 年迄今 212 天真实样本外表现（截至 7 月 31 日），经保守折扣后给出。数据成对呈现：实盘目标 / 样本内回测。
Sharpe retention well above the industry >0.5 threshold for genuine edge in two independent out-of-sample windows places the portfolio in the top decile of systematic strategies. 在两段独立的样本外区间，夏普保持率均显著高于 0.5 的行业真实阿尔法阈值——这一表现跻身系统化策略前十分位。
Backtest-to-live divergence held within 1–2 percentage points. Engineering-grade deployment. 回测与实盘偏差控制在 1–2 个百分点，工程级部署。
Minute-resolution OHLCV across multiple market regimes. Coinbase, Binance, Binance Perp. Funding rate history modeled for directional cost. 横跨多个市场周期的分钟级 OHLCV 数据。来源涵盖 Coinbase、Binance、Binance 永续，包含完整资金费率历史。
At 1.5× nominal leverage, average net market exposure is only ~0.5× (gross ~0.9×) — offsetting long/short positions across 21 strategies internally hedge away nearly half of deployed risk. 在 1.5 倍名义杠杆下，平均净市场敞口仅约 0.5 倍（总敞口约 0.9 倍）——21 项策略的多空头寸相互对冲，内部消化了近一半的已部署风险。
No discretionary overrides. No narrative trades. Every position is a model output. Drawdown limits are pre-committed, not reactive. 不做主观干预，不做叙事交易。每笔仓位皆为模型输出；回撤上限事先约定，而非事后被动响应。
Your capital stays in your own exchange account. We trade it through a trade-only, IP-whitelisted, withdrawal-disabled API key — we can never move your funds. A separately managed account, not a commingled fund: there is no manager-custody risk, and your money never leaves your account. 资金始终留在您自己的交易所账户。我们仅通过「只读交易、IP 白名单、禁止提现」的 API 密钥进行交易，无法转移您的资金。这是独立管理账户（SMA），而非混同基金——不存在管理人托管风险，您的资金从不离开您的账户。
Capital sits in your own Binance / OKX sub-account. You hold the keys and 100% of the balance. 资金存放于您自己的 Binance / OKX 子账户，您持有密钥与全部余额。
You grant a withdrawal-disabled, IP-whitelisted API key — scoped to trading only. Revoke it any time. 您授予一枚禁止提现、IP 白名单的 API 密钥，仅限交易权限，随时可撤销。
We place trades and settle agreed fees. Every fill is exchange-timestamped — a third-party-verifiable record from day one. 我们执行交易并结算约定费用。每笔成交均有交易所时间戳，第一天起即可第三方验证。
Our production integration with Hexon is live, and Latent Meta strategies are now listed on the Hexon marketplace — a new way to access the book.与 Hexon 的生产环境集成已上线，Latent Meta 策略现已登陆 Hexon 交易市场——为投资者提供全新的接入方式。
Guided by our post-ETF market-structure research, an ETH range-regime sleeve was rotated out for a newly validated low-frequency 1H trend-inflection strategy — selected on training data only and 8/8 on the validation battery: parameter stability, walk-forward reselection, Monte-Carlo timing, 5x cost stress, cross-asset transfer. The book stays at 21 strategies.基于我们对 ETF 时代市场结构的研究，一项 ETH 区间策略被轮换为新验证的低频 1H 趋势拐点策略——仅以训练数据选型，并以 8/8 通过验证测试：参数稳定性、滚动重选、蒙特卡洛时序检验、5x 成本压力与跨资产迁移。组合维持 21 项策略。
Two newly validated strategies joined the book — including our first ZEC edge — and one retired, keeping the four signal families regime-balanced.两项新验证的策略加入组合——包括我们首个 ZEC 策略——同时退役一项，保持四大信号体系的状态均衡。
A volatility-aware regime-sizing layer and a counter-trend profit-taking rule now govern position sizing — cutting modeled max drawdown by roughly a third at comparable returns.波动率感知的状态仓位调节层与逆势止盈规则接入仓位管理——在收益相当的情况下，将模型最大回撤压缩约三分之一。
The production bot began trading live capital in April 2026 — realized Sharpe 1.9, exchange-verified.生产环境交易机器人于 2026 年 4 月开始实盘交易——已实现夏普 1.9，可交易所验证。