(量化研究平台)
Discover · Validate · Compose · Deliver// One platform is enough
Find strategies, backtest with real rules, and check for overfitting with statistics. No curve cosmetics, no return promises — you can see whether results are trustworthy. Strategy files go to accounts you control.
我们的服务器 7×24 不停 地生成、复算、 淘汰策略。下面这些数字不是宣传，是这台机器正在做的事。
你买到的不是今天这 0 条， 而是这台机器明天、下个月还在继续为你造新的。
✓最近一次通过检验4 小时前PAXGUSDT · H1夏普 1.55区间收益 +5.5%查看这条策略 →Top Return Strategies
View all strategiesScreened strategies passed profitable replay, signal-health checks and the baseline robustness policy. Advanced checks are disclosed separately and can be combined as filters. Historical results do not predict future returns.
挑选规则：按「收益 ÷ 最大回撤」排序，且只陈列交易数 ≥ 30 笔、同一品种最多 9 条。不按收益最高排 —— 避免只有几笔交易的「假优秀」占住主推位。
0 screened strategies · ranked by annualized return; showing the top 0
No strategy matches the current filters. Loosen them to browse more candidates.
View all strategies24h 发现 229 轮 · 新增 167 · 精选 0（M30 0 · M15 0 · M5 0 · D1 0）
START HERE · 30 SECOND PATH
先选市场，再做一件事
不需要先学习术语。选择你更关心的市场，平台会把这个选择带到生成策略和模拟观察中。
- 01市场快照XAU/USD · H1
- 02策略检查规则、数据与压力测试
- 03模拟观察先看信号与虚拟成交
DECISION PLAYBACK · 过程示意
一眼看懂：策略不是“看到行情就下单”
先读取市场快照
把 XAU/USD 的周期、波动和成本放在同一张图里。
先验证，再模拟；不代表未来收益
stack · integrations · web3
Works with
Compatible terminals, exchange APIs, and research tools; USDT settlement supported. Logos mean compatibility — not endorsements or custody.
Marks indicate technical compatibility, research tooling, or optional delivery paths. All trademarks belong to their owners. No equity or exclusive agency is implied unless separately disclosed in writing.
Strategies discovered & verified by our system
Our servers continuously generate and re-test strategies. Candidates that pass the baseline enter the library; advanced checks stay visible so you choose what must pass.
Auto-discovery
Multiple algorithms rotate 24/7 across gold, bitcoin and timeframes; new strategies keep landing, deduped to keep the best.
Transparent validation
OOS, PSR, DSR, Walk-Forward and Monte Carlo are shown separately. Amber means attention and grey means pending; one overly strict gate does not decide for you.
One-click delivery
Filter a strategy → download it to your trading software for paper testing → decide whether to deploy small after observation, or replace it with another.
Many modules. One complete OS.
Use each feature alone or follow the suggested path. We build research tools — we do not trade for you or hold your funds.
EasyQuant · main path
Not a “winning formula” — a way to see the tablecloth
Markets punish participants without tools and statistical literacy — information asymmetry is the harvest. We turn institutional overfitting tests, cost rails, and risk gates into visual evidence so ordinary researchers can use serious methods instead of being sold a curve. No return promises · not investment advice.
Three stories that travel
See the path from discovery to validation and paper observation. No return promises — only inspectable evidence.
Pretty in-sample still needs full evidence
A real gold dual-MA backtest: negative return and low Sharpe shown as-is. Grey means not yet tested; results are never presented as a return promise.
Open demo backtest →Validation is visible — the decision stays yours
Signal health, OOS, statistical checks, Walk-Forward and Monte Carlo are disclosed separately. Green, amber and grey show the current evidence state.
See validation pipeline →Factory delivers, you execute
Filter strategies, inspect their evidence, then download to paper-test in your own trading software. Keys and live execution stay with you.
Open System Forge →A complete strategy research loop
From discovery to honest backtests to overfitting filters — fully verifiable.
Multi-algorithm strategy search
Genetic search, rule search, reinforcement learning, and more. Compare methods on the same market. Jobs run in the background with live progress.
- 10+ methods — not a single black box
- Live progress & equity curves
- Factor style packs + templates / features
Backtests that cannot peek ahead
Advances one bar at a time. Returns, Sharpe, drawdown, and trades are computed for real — no future data leaks.
- Bar by bar · no peeking ahead
- Batch retests · jobs isolated
- In/out-of-sample · walk-forward · Monte Carlo
Optional robustness evidence
Statistical checks help identify possible overfitting. Amber highlights evidence that needs attention; combine it with OOS, drawdown and paper observation.
- Overfitting & significance checks
- Pass, attention and pending states stay distinct
- Paper follow-through (not live)
Technical foundation
Results earn trust because the architecture refuses self-deception.
Look-ahead-free engine
Time advances bar by bar. At any moment only past and present are visible — no predicting the past with the future.
Multi-algorithm discovery
Multiple search methods side by side. The same validation rules apply — switching methods does not switch standards.
Overfitting statistical tests
Picking the best of many trials inflates luck — we discount that. Making money and being trustworthy are different questions.
Async job queue
Heavy work runs in the background queue with live progress. Closing the browser does not stop the job.
Signal health checks
Empty shells and always-on signals are filtered. Only strategies with real logic enter the library.
XAUUSD / MT5 data
Built-in data, file upload, and optional local MT5 quotes. Near-real-time charts; live orders stay on your terminal — we do not custody funds.
量化交易实战指南
我们把自己在策略发现、回测与验证里踩过的坑，写成可以直接照做的技术文章。文章说明操作步骤、适用条件，以及验证方法的限制。
Original research: how a 1,001-bar test reused 620 discovery bars
Minimum account size for trading (how to work out what you need)
Account size is not just a comfort question. It decides whether you can size correctly, whether your drawdown percentage is survivable, and whether fixed costs leave anything behind. Here are the three mechanisms, the arithmetic for each, and how to work out the smallest account a specific strategy can actually be run on.
阅读全文howtoHow to reduce slippage in algo trading: order types, timing and sizing
Slippage cannot be eliminated, because it is the price of demanding immediacy. It can be reduced, and the levers are specific: order type, when you send the order, how you split it, and what you trade. Here is each lever with what it costs you in exchange.
阅读全文conceptYour backtest passed. Then a real terminal disagreed.
A backtest is code you wrote, running on data you chose. A live terminal is a different program, a different data feed and a different set of assumptions about fills. We re-ran 1,307 of our own strategies on a real terminal. 105 of them failed, diverged or mismatched.
阅读全文howtoBacktest with Python: the cost assumptions that decide your result
Writing a backtest in Python is easy. Writing one whose number means something is not, because the assumptions you leave out are the ones that flatter the result. Here are the seven assumptions to make explicit, with the code shape for each — spread, commission, intrabar order, slippage, position sizing, warm-up and the equity series itself.
阅读全文conceptCAGR formula: how to calculate annual return from a backtest
Total return and CAGR describe the same account and can differ by a factor of four. One ignores time; the other assumes compounding. Here is the conversion, the two ways the number is misused, and why you need the window and the bar count before either figure means anything.
阅读全文howtoMT5 EA works in backtest but not live: a diagnostic checklist
Diagnose an MT5 EA that does not trade or differs from its backtest. Compare signals, order return codes, execution settings and independent evidence.
阅读全文Validation is visible — the decision stays yours
Black-box EAs call every curve “stable profit.” We show pass, attention and not-yet-tested states for each check instead. Advanced tests are evidence, not a substitute for your paper observation.
Of 3796 strategies, with 10 robust-certified ( 0.3%). Review the other evidence, apply your own filters, and paper-test before deciding.
Three memberships · one OS
Free · Premium · Diamond. Daily quotas. Strategies are generated here; live trading stays on your accounts.
Free Member
- Linear-GP + GA · 15 discoveries / day
- 50 backtests / day · validation
- Paper · live prerun shelf · direction preview
- All engines & export packs need a paid plan
Premium Member
- All algorithms · 100 discoveries / day
- 1000 backtests · portfolio & lifecycle
- MT5 EA · strategy packs (no-model strategies)
- Decay monitor · all Free features
Diamond Member
- Everything in Premium + higher limits
- Full packs for strategies with model files
- Local live limits · auto kill-switch
- Legal consent · no custody
Prices in USD. Upgrade with USDT. Research tools do not promise returns; live keys and capital stay with you.
Ready to research strategies that survive scrutiny?
Sign up free and run your first honest backtest report in minutes.
Get started