SignalFounder 只有一个闭环：发现一个思路，把它编排成规则，用历史数据压力测试，再实时盯住它。本指南按顺序讲解每一个界面——它是做什么的、该点哪里，以及在你进入下一步之前，怎样才算"够好"。SignalFounder is one loop: discover an idea, compose it into rules, stress-test it on history, then watch it live. This guide walks each screen in order — what it does, where to click, and what "good enough" looks like before you move on.
引导流程不会问你的收入。它只问你想先做什么、关心哪个市场、想让编辑器多亲力亲为——然后据此调配 Discover、Build 和 Monitor。这里的一切都可以在设置里改回来。The setup flow won't ask your income. It only asks what you want to do first, which market you care about, and how hands-on you want the editor to be — then tunes Discover, Build, and Monitor to match. Everything here can be changed back in settings.
Discover 是供给侧：一个可搜索的信号、指标和社区策略库。信号和指标是有意分开的——信号是"触发/不触发"的决策，指标是你读取的一个数值。找到与你思路接近的东西，把它 Fork 下来，而不是从空白开始。Discover is the supply side: a searchable library of signals, indicators, and community strategies. Signals and indicators are deliberately separate — a signal is a fire/no-fire decision, an indicator is a number you read. Find something close to your idea and fork it instead of starting blank.
Build 负责策略工作。引导式编辑器用表单和标签——上手最快。逻辑图编辑器则展开嵌套规则、自定义指标和版本历史。两者写的是同一份数据，所以在引导式里起步的策略能在逻辑图里干净地打开。在这里设定你的入场和出场逻辑、仓位大小和成本。Build is where the strategy work happens. The guided editor uses forms and chips — fastest to start. The logic-graph editor unfolds nested rules, custom indicators, and version history. Both write the same data, so a strategy started in the guided editor opens cleanly in the graph. Set your entry and exit logic, position size, and costs here.
entry: when RSI(14) crosses_below 30 and close > EMA(200) exit: when RSI(14) crosses_above 55 or bars_held > 24 size: vol_target("0.35") stop: 1.2%
回测描述的是过去的表现，不是承诺。SignalFounder 默认按贴近真实的成本运行，并把可信度一栏放在最前：扣费后的净收益、最大回撤、敞口和成交笔数。先读这些，再看资金曲线。一条漂亮但只有五笔成交的曲线只是噪声。Backtests describe past performance, not a promise. SignalFounder runs with realistic costs by default and puts the trust column first: net-of-fees return, max drawdown, exposure, and trade count. Read those before the equity curve. A pretty curve with only five trades is just noise.
Monitor 负责实时状态。绑定一个钉住的策略版本，只有当规则真正触发时，app 才会即时提醒。健康面板会追踪信号新鲜度、与回测的漂移和回撤趋势——让你在回撤之前就知道该暂停，而不是之后。免打扰时段会被尊重。Monitor handles live status. Bind a pinned strategy version, and the app only alerts when a rule actually fires. The health panel tracks signal freshness, drift from the backtest, and drawdown trend — so you know to pause before a drawdown, not after. Quiet hours are respected.
Factor Lab 是预验证因子研究的网页归档——每个实验在发布前都通过了紧缩夏普、FDR 校正、置换检验和成本真实性。无需账号即可浏览；当某个实验契合你的思路时，保存或 Fork 它，几秒内它就会落到你 iOS 上的 Build 工作区。Factor Lab is a web archive of pre-validated factor research — each experiment cleared deflated Sharpe, FDR correction, permutation tests, and cost realism before publishing. Browse without an account; when an experiment fits your idea, save or fork it and within seconds it lands in your Build workspace on iOS.
一些小规则，把可交易的策略和一张讨喜的图表区分开来。A few small rules that separate a tradable strategy from a flattering chart.
净收益、回撤和成交笔数比资金曲线的形状告诉你更多。寥寥几笔成交撑起的漂亮曲线只是噪声。Net return, drawdown, and trade count tell you more than the shape of the equity curve. A pretty curve on a handful of trades is just noise.
大改动前先保存一个版本，这样 Monitor 始终清楚到底什么在运行，你也能并排对比各次运行。Save a version before big changes, so Monitor always knows what's running and you can compare runs side by side.
当实时表现与回测背离时，健康卡会标记出来。把第一次漂移标记当作去看一眼的理由，而不是恐慌的理由。The health card flags when live performance diverges from the backtest. Treat the first drift flag as a reason to look, not to panic.
创建一个工作区，从 Discover 或 Factor Lab Fork 一个东西，一个下午就能让一个回测过的策略跑上实时监控。Create a workspace, fork something from Discover or Factor Lab, and get a backtested strategy onto a live monitor in an afternoon.