# Trado Strategy Lab: full text > Trado Strategy Lab is an AI-first algo trading platform for crypto perpetual futures: describe a strategy in plain words, backtest it on years of 1-minute data for 900 Binance perpetuals after every cost, see where it works, paper-trade it, and bring your own AI over MCP. Source: https://algo.trado.trade # The AI-first algo trading platform **Describe it. Test it honestly. Paper-trade it. Bring your own AI.** Trado Strategy Lab is for crypto perpetuals. Say what you want in plain words, backtest it on years of 1-minute data for 900 perpetuals after every cost, see where it works and where it does not, then paper-trade it. ## Backtests lie. Ours prices the cost. > My backtest said +300 %. My account said −40 %. Most backtests ignore fees, funding and slippage, peek at the future by accident, and never check whether the position would have been liquidated. The result looks good because it was never charged for anything. Every Lab trade is priced after costs, liquidation is modelled for isolated and cross margin, optimizations keep a holdout period aside to validate on, and the report's verdict will not give a green light for paper gains. Read [why backtests lie](https://algo.trado.trade/learn/why-backtests-lie.md). ## Coding should not be the gate. > I can see the pattern. I can't write the code. Pine Script, Python and exchange APIs keep many people with good market ideas from ever testing them. Describe the idea in plain words and the Copilot drafts the rules. Use the form builder or the canvas to refine them. When you want code, write Python against a small SDK and the engine runs it. Read about [algo trading without code](https://algo.trado.trade/learn/algo-trading-without-code.md). ## Data should not be a chore. > Getting clean data is a part-time job. Downloading, cleaning and storing years of 1-minute bars is a project before it is a strategy. The Lab already holds years of 1-minute data for 900 Binance perpetuals, with funding rates and instrument rules, refreshed twice a day. Pick coins and dates and run. Read about [crypto data for backtesting](https://algo.trado.trade/learn/crypto-data-for-backtesting.md). ## One lucky backtest is not an edge. > I optimized it and it got worse. A strategy tuned until the past looks perfect usually fails on the future. The optimizer validates on a holdout and walks forward through time. Insights say where a strategy works and where it does not, with regime tables and an evidence meter that reads Not enough, Weak, Moderate or Strong. Read about [overfitting in backtests](https://algo.trado.trade/learn/overfitting-in-backtests.md). ## Going live should not mean starting over. > Backtest to live is a cliff. Rewriting a backtested idea for a bot, wiring webhooks and redoing the sizing maths is where good strategies get lost. The same rules run on paper and live through the Lab's runner, using the same decision and position code as the backtest. One click starts a paper run; live goes through an exchange adapter and your explicit approval. Read about [moving from backtest to live](https://algo.trado.trade/learn/from-backtest-to-live.md). ## Leverage kills quietly. We show the number. > Leverage looked free until it wasn't. At 20×, a move of about 4.5% against you loses the whole margin. Most tools never put that number in front of you. The liquidation guard stops a run whose stop-loss sits beyond the liquidation price, and the leverage audit in every report shows what each coin allowed. Read [leverage and liquidation, explained](https://algo.trado.trade/learn/leverage-and-liquidation-explained.md). ## Pay for work, not for a subscription you barely use. > Another subscription for a tool I use twice a month. Charting and bot tools often bill a monthly fee for something used twice a month. The Lab prices work in credits: you see the price before you run, you pay as you go, and building and editing strategies costs nothing. Bringing your own AI agent is free. Read [what algo trading tools cost](https://algo.trado.trade/learn/what-algo-trading-tools-cost.md) or see [pricing](https://algo.trado.trade/pricing.md). ## There should be someone to ask. > There is no one to ask. Forums and video rabbit holes are a slow way to find out why a strategy behaved the way it did. The Copilot explains every run, and “Ask why” sits on every result. Read about [learning algo trading with AI](https://algo.trado.trade/learn/learn-algo-trading-with-ai.md). ## AI with you at every step The Copilot is built into the Lab. It drafts a strategy from your description, proposes sensible defaults for a run and shows what it will cost before anything is spent, runs the backtest when you confirm, and explains the result. Anything expensive, and anything that touches a live account, waits for your yes. 1. **Describe it.** Write the idea in plain words. The Copilot drafts the rules and reads them back to you; you change them in the form builder, on the canvas, or in Python. Building and editing strategies costs nothing. 2. **Backtest it after every cost.** Fees, funding, slippage and liquidation are priced into each trade, on 1-minute bars for 900 Binance perpetuals. You see the credit price before anything runs. 3. **Understand it.** The report shows the result by coin, by market regime and by period, and says plainly when the evidence is thin. Ask why on any result and the Copilot explains it. A verdict in plain words, never a green light for paper gains. 4. **Paper-trade it.** The same rules run on live prices without risking money, one click from the report. Backtests and paper trading never touch a real account. 5. **Go live, on purpose.** Live trading goes through an exchange adapter that you set up, and waits for a separate approval from you in the browser. Nothing goes live on its own, and an agent cannot approve it. ## Honest by design Three things most backtests skip, and what the Lab does about each of them. - **Costs.** Fees, funding and slippage are charged on every entry and exit, so a strategy that trades often pays for it. Proof: the report flags the cost margin and any liquidation as badges. - **Liquidation.** Isolated and cross margin are modelled the way the exchange does it, and a stop beyond the liquidation price is refused before the run. Proof: the liquidation guard answers before you spend a credit. - **Holdouts.** Optimizations validate on a period they were never tuned on and walk forward through time, so a lucky fit shows up as one. Proof: the evidence meter reads Not enough, Weak, Moderate or Strong. ## Bring your own AI. It costs nothing extra. The Lab speaks MCP, so Claude, Cursor or ChatGPT can create strategies, run backtests and read reports with a personal access token. You pay your AI provider what you already pay them; the Lab charges nothing for the connection. See [bring your own AI](https://algo.trado.trade/bring-your-own-ai.md). - Claude - Cursor - ChatGPT One line connects Claude Code: ```sh claude mcp add --transport http trado-lab https://algo.trado.trade/mcp --header "Authorization: Bearer " ``` ## Pay for the work you run Building and editing strategies is free. Backtests, optimizations, Copilot replies and paper or live deployments use credits, priced by how much work they are, and you see the price before you run. See [pricing](https://algo.trado.trade/pricing.md) for the live price list. Waitlist members start with 500 launch credits on top of the 25 welcome credits every account gets. Founding members also pay the Starter price for their first plan order for any monthly plan above Starter. ## Join the waitlist: founding member benefits The Lab opens on 1 November 2026, 00:00 IST. Waitlist members get in first and get more. The benefits attach when you create your account with the same e-mail address you used on the waitlist and verify it. Joining the waitlist after launch gives no founding benefits. - 500 launch credits, on top of the 25 welcome credits every account gets. - Early access by invitation, about a week before launch. - First month of a higher plan at the Starter price. - A Founding member badge on your account. - The Starter price applies to your first paid plan order for any monthly plan above Starter, never together with a coupon, and is available until 7 January 2027. - Each referral who joins moves you up 5 places. Bronze at 1 referral, Silver at 3 (adds 250 credits), Gold at 10 (adds 1,000 credits). The level you hold when your e-mail is verified is the one that counts. - Launch credits are valid for 90 days from the moment your e-mail is verified. ## Questions ### Is Trado Strategy Lab live yet? Not yet for the public. The Lab launches on 1 November 2026, 00:00 IST. Until then the site takes waitlist sign-ups, and the team already uses the Lab. ### Which markets does it cover? Binance USDⓈ-M perpetual futures: 900 coins with years of 1-minute data, plus funding rates and each coin's instrument rules. Backtests run on 1-minute bars, exact exits on the trade tape, or trade-level. ### Do I need to code? No. Describe the idea in plain words, or build it in the form builder or on the canvas. If you do code, Python strategies run in a sandbox against a small SDK. See [the AI strategy builder](https://algo.trado.trade/ai-strategy-builder.md). ### Will it trade my money for me? Only if you tell it to. Backtests and paper trading never touch a real account. Live trading needs an exchange adapter that you set up and a separate approval from you in the browser. See [paper trading](https://algo.trado.trade/paper-trading.md). ### Is this investment advice? No. The Lab is software for testing and running rules you decide on. A backtest describes the past, trading leveraged derivatives can lose more than you expect, and nothing here is a promise of profit. ### What does it cost? Building and editing strategies is free. Backtests, optimizations, Copilot replies and paper or live deployments use credits, priced by how much work they are, and you see the price before you run. See [pricing](https://algo.trado.trade/pricing.md). ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) - [See pricing](https://algo.trado.trade/pricing) ## Related - [Crypto backtesting after every cost](https://algo.trado.trade/backtesting.md): Backtest crypto perpetuals on years of 1-minute data for 900 coins, after fees, funding, slippage and liquidation, with a verdict that admits weak evidence. - [AI strategy builder, no code needed](https://algo.trado.trade/ai-strategy-builder.md): Describe a trading idea in plain words and the Copilot drafts the rules. Refine them in a form or on a canvas, or write Python. No code required. - [Paper trading for crypto strategies](https://algo.trado.trade/paper-trading.md): Take a backtested strategy to paper trading in one click. The same rules run on live Binance prices; going live is a separate step you approve. - [Bring your own AI via MCP](https://algo.trado.trade/bring-your-own-ai.md): Connect Claude, Cursor or ChatGPT to Trado Strategy Lab over MCP. Your agent builds strategies, runs backtests and reads reports for free; runs use credits. --- Source: https://algo.trado.trade/waitlist # Join the waitlist. Get in first, and get more. Trado Strategy Lab opens on 1 November 2026, 00:00 IST. Waitlist members get 500 launch credits on top of the welcome credits, a Founding member badge and early access by invitation. Add your email, and your WhatsApp number if you want the launch message there. ## What founding members get The benefits attach when you create your account with the same e-mail address you used on the waitlist and verify it. Joining the waitlist after launch gives no founding benefits. - **500 launch credits**, on top of the 25 welcome credits every account gets (525 in total at Bronze). They are valid for 90 days from the moment your e-mail is verified. - **A Founding member badge** on your account. - **Early access:** Early access by invitation, about a week before launch. Invitations go out as addresses are approved, so there is no exact date to promise. - **A first month at the Starter price:** your first paid plan order for any monthly plan above Starter costs the Starter price. It applies to the first paid plan order only, never together with a coupon, and is available until 7 January 2027. ## How it works 1. **Join.** Add your email, and your WhatsApp number if you want the launch message there. 2. **Share your link.** You get a personal link as soon as you join. Send it on WhatsApp or anywhere else. 3. **Move up.** Each friend who joins with your link moves you up 5 places and counts toward your level. ## Move up the list by sharing it Your position is your join order. Each friend who joins with your link moves you up 5 places, and referral levels add credits on top of the launch credits. The level you hold at the moment your e-mail is verified is the one that counts. - **Bronze:** 1 referral. - **Silver:** 3 referrals, adds 250 credits. - **Gold:** 10 referrals, adds 1,000 credits. - A referral made from your own browser does not count, so the list stays fair. ## What we ask for, and what we do with it Only your email is required. We use it to confirm your place and to tell you when the Lab opens. Your WhatsApp number, if you give it, is stored unverified and used for the launch message only. Your name and how you trade are optional and help us decide what to build first. We never publish the list, and the form does not say whether an address or number is already on it. A temporary-mail address is not accepted. ## Waitlist questions ### When does the Lab open? On 1 November 2026, 00:00 IST. Early access is by invitation, about a week before launch, as addresses are approved. ### Do I have to sign up with the same e-mail address? Yes. The benefits attach only when the account is created with the e-mail address you used on the waitlist and that address is verified. Joining the waitlist after launch gives no founding benefits. ### What happens if I join twice? Nothing changes. An address that is already on the list gets its own place back. ### Do I have to give a WhatsApp number? No. It is optional. If you add one, use the international format, for example +91 98765 43210; ten-digit Indian numbers get +91 by themselves. ### Does joining cost anything? No. Joining is free and does not commit you to anything. See [pricing](https://algo.trado.trade/pricing.md) for what the Lab costs once it opens. ## Next steps - [See pricing](https://algo.trado.trade/pricing) ## Related - [Pricing: pay for the work you run](https://algo.trado.trade/pricing.md): Building strategies is free. Backtests, optimizations, Copilot replies and deployments use credits, priced by the work, and you see the price before you run. - [Bring your own AI via MCP](https://algo.trado.trade/bring-your-own-ai.md): Connect Claude, Cursor or ChatGPT to Trado Strategy Lab over MCP. Your agent builds strategies, runs backtests and reads reports for free; runs use credits. - [Algo trading for beginners](https://algo.trado.trade/for-beginners.md): Start algo trading without code or a data pipeline: describe an idea, backtest it honestly, ask the Copilot why, and paper-trade before you risk a rupee. --- Source: https://algo.trado.trade/pricing # Build strategies for free. Pay only for what you run. Creating and editing strategies costs nothing. Backtests, optimizations, Copilot replies and deployments use credits, priced by how much work they are, and you see the price before you run. Prices include GST. ## How credits work One credit ledger pays for everything that uses compute. Each action has a price in credits, shown before you confirm it: a small backtest costs little, a long multi-coin run costs more, and an optimization is reserved at its time budget and charged for the time it actually uses. A backtest or optimization that fails is refunded. Bring-your-own-AI access to the Lab is free; only the runs your agent starts use credits, exactly as if you had started them yourself. - **Free:** creating, editing and versioning strategies, the form builder, the canvas, and reading your results. - **Credits:** backtests, optimizations, Copilot replies, exports and paper or live deployment days. - **Plans** include an allowance of credits each period; **packs** are credits you buy when you need them and use for anything. ## Pay as you go, or take a plan There is no subscription you have to keep. Buy a credit pack when you need one, or take a plan if you run the Lab every week. A plan lasts one period and does not renew by itself: buy it again to keep it, and a few days of grace keep your limits meanwhile. Waitlist founding members get launch credits, and their first plan order for any plan above Starter costs the Starter price. See the [waitlist](https://algo.trado.trade/waitlist). ## Pricing questions ### Will I know the cost before I run something? Yes. Every run, optimization and deployment shows its price in credits before you confirm it, and the Copilot states the cost before it runs anything. ### What happens if a run fails? A backtest or optimization that fails is refunded in credits. ### Do prices include GST? Yes. The prices shown include GST, and invoices are issued for every purchase. ### Does bringing my own AI cost extra? No. Connecting Claude, Cursor or ChatGPT to the Lab is free. The backtests and optimizations your agent starts use credits like any other run. See [bring your own AI](https://algo.trado.trade/bring-your-own-ai.md). ## Current plans Read from the live price list; prices include GST. New accounts also get 25 welcome credits once their email is verified. - **Free:** Free. 20 backtests. Best for trying it out. - **Starter:** ₹499 / month. 400 backtests · 300 optimizations · 150 Copilot. Best for one strategy a week. - **Pro:** ₹1,499 / month. 1,500 backtests · 1,200 optimizations · 600 Copilot · 100 data export. Best for daily research. - **Scale:** ₹4,999 / month. 6,000 backtests · 5,000 optimizations · 2,500 Copilot · 600 data export. Best for teams and agents. ## Credit packs (pay as you go) Pack credits work for anything and last 365 days. - **Small:** 200 credits for ₹199 (₹1 per credit). - **Medium:** 1,250 credits for ₹999 (₹0.80 per credit). - **Large:** 7,500 credits for ₹4,999 (₹0.67 per credit). ## What things cost - Backtest, small (runs on the Lab server): 1 credit. - Backtest, one coin for a year: 3 credits. - Backtest, 30 coins for three years: 6 credits. - Copilot reply, typical: 19 credits. - Optimization, 5 minutes (settled to the time used): 78 credits. - Optimization, 30 minutes (settled to the time used): 403 credits. ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) ## Related - [What algo trading tools cost](https://algo.trado.trade/learn/what-algo-trading-tools-cost.md): Subscriptions for tools you use twice a month add up. What to look for in algo trading pricing, and how credits priced by the work compare. - [Bring your own AI via MCP](https://algo.trado.trade/bring-your-own-ai.md): Connect Claude, Cursor or ChatGPT to Trado Strategy Lab over MCP. Your agent builds strategies, runs backtests and reads reports for free; runs use credits. - [Join the waitlist](https://algo.trado.trade/waitlist.md): Join the Trado Strategy Lab waitlist for 500 launch credits, early access by invitation and referral rewards. Launch: 1 November 2026, 00:00 IST. --- Source: https://algo.trado.trade/backtesting # Backtesting that charges you for everything the market would A backtest is only useful if it is honest. The Lab replays your rules on years of 1-minute data for 900 Binance perpetuals, prices every trade after fees, funding and slippage, models liquidation, and tells you plainly how much to trust the result. ## What every trade is charged A backtest runs under an execution profile: one broker's fees, slippage, funding, leverage caps, lot and tick rules, account currency and liquidation model. The default profile models Binance USDⓈ-M futures. Nothing is optional and nothing is left at zero by default. - **Fees and slippage** on every entry and exit, so a strategy that trades often pays for it. - **Funding** on every position held across a funding time. - **Instrument rules:** each coin's lot size, tick size and leverage cap apply, so an order the exchange would refuse is refused here too. - **Liquidation**, modelled for isolated and cross margin: a position that would have been liquidated is liquidated, not rescued by hindsight. ## The data is already there The Lab holds years of 1-minute bars for 900 Binance USDⓈ-M perpetuals, with funding rates and instrument data, refreshed twice a day. You choose coins and dates; there is nothing to download, clean or store. A run can replay 1-minute bars, use the trade tape for exact stop and target fills, or work at trade level. A run says when a coin did not have enough history for a filter to be ready, instead of quietly trading without it. ## A report that says where it works Each run produces a full strategy-tester report with a chart of every trade. Beyond the headline numbers it breaks the result down so one lucky stretch cannot hide. - **Market regimes:** every trade is grouped by the market it entered in (bull, strong bear, weak bear, sideways, with a high-volatility flag), so you see which conditions paid and which did not. - **By coin and by period:** whether a few coins or a few days carried the result. - **A verdict in plain words:** from “Made money broadly” to “Lost money”, including “Too few trades to tell”. - **An evidence meter** that reads Not enough, Weak, Moderate or Strong, because ten trades are not a result. - **A leverage audit** showing the caps that applied to each coin. ## Optimization that checks itself When you search for better parameters, the optimizer offers grid, random, genetic, TPE, CMA-ES and NSGA-II searches, and validates with a holdout period and walk-forward testing. It shows heatmaps, parameter importance and how stable the neighbourhood around the best settings is, and reports Monte Carlo resampling, the probability of backtest overfitting and the deflated Sharpe ratio. Read about [overfitting in backtests](https://algo.trado.trade/learn/overfitting-in-backtests.md). ## Backtesting questions ### Can a backtest look into the future? The engine runs the same decision code bar by bar that the live runner uses, built-in indicators are checked against published reference values, and custom Python indicators are checked for look-ahead before they can be used. ### Whose fees and funding does it use? The default execution profile models Binance USDⓈ-M futures. A profile is a named, versioned set of fees, slippage, funding, leverage caps, instrument rules and liquidation model, so results are reproducible. ### How long does a backtest take? Small runs finish in seconds; long or many-coin runs go to elastic compute and show an estimate and the credit price before you start. ### Does a good backtest mean it will make money? No. A backtest describes the past. The Lab is built to make that description honest, not to promise a result. Read [why backtests lie](https://algo.trado.trade/learn/why-backtests-lie.md). ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) - [See pricing](https://algo.trado.trade/pricing) ## Related - [Why crypto backtests lie](https://algo.trado.trade/learn/why-backtests-lie.md): Look-ahead, missing costs, survivorship and ignored liquidation flatter backtests. A checklist to test any backtest, and how the Lab answers each point. - [Crypto data for backtesting](https://algo.trado.trade/learn/crypto-data-for-backtesting.md): Clean 1-minute crypto data is a project before it is a strategy. What good backtesting data needs, where DIY pipelines break, and what the Lab holds. - [Overfitting in backtests](https://algo.trado.trade/learn/overfitting-in-backtests.md): Tune a strategy long enough and the past looks perfect. Why overfitting feels like success, and the checks that expose it: holdouts, walk-forward, stability. - [Paper trading for crypto strategies](https://algo.trado.trade/paper-trading.md): Take a backtested strategy to paper trading in one click. The same rules run on live Binance prices; going live is a separate step you approve. --- Source: https://algo.trado.trade/ai-strategy-builder # Describe the idea. The Copilot drafts the rules. Write what you want in plain words, such as “buy when the 4-hour trend is up and RSI dips below 30, stop at 2%”. The Copilot turns it into explicit rules you can read and change, shows what a backtest will cost, and runs it when you confirm. ## What the Copilot does The Copilot is built into the Lab and uses the same tools your own agent would, so what it builds is ordinary, inspectable strategy rules. - **Drafts a strategy** from your description and asks about the things that matter instead of guessing. - **Proposes defaults** for the run: coins, dates and settings. - **Shows the cost** before anything is spent, and runs the backtest only when you confirm. - **Explains the result.** “Ask why” sits on every result: why it lost in a regime, why a stop was hit, what a metric means. ## You stay in control Nothing the Copilot writes is hidden. The strategy is plain rules you can open, edit and version. Anything expensive, and anything that touches a live account, waits for your explicit yes. Creating and editing strategies costs no credits; the Copilot's replies and the runs it starts do, at prices shown first. ## Four ways to build, one engine Every way ends in the same strategy definition, so you can switch between them without changing what the strategy does. - **Plain words** with the Copilot. - **The form builder:** indicators, entries, exits, targets, stops, partial exits, trailing stops and risk, as fields. - **The canvas:** the same rules drawn as a graph, with live sparklines on each node and a “why here?” view of every value at a clicked bar. - **Python:** a small SDK, run in a sandbox with no network and no secrets. Only the idea lives in code; every number is a parameter a run can change. ## What you can build with There are 170 built-in indicators, including averages, structure, momentum, volatility, volume, statistics such as beta and correlation against another coin, and all 61 candlestick patterns, each checked against published reference values. Conditions can use other timeframes and other coins, indicators can be fed by a formula or another coin, and named values and conditions can be reused across rules. Six ready-made examples (EMA crossover, Donchian breakout, RSI fade, squeeze breakout, session and volatility, BTC regime filter) open as editable rules. Strategies are versioned and forkable, and custom indicators written in Python are checked for look-ahead. ## Strategy builder questions ### Do I need to know how to code? No. You can describe a strategy, use the form builder or draw it on the canvas. Python is there for when you want it. ### Can I use my own AI instead of the Copilot? Yes. Claude, Cursor or ChatGPT can build and test strategies in the Lab through MCP, at no extra charge for the AI. See [bring your own AI](https://algo.trado.trade/bring-your-own-ai.md). ### What if the Copilot gets the rules wrong? You see the rules before anything runs, and a strategy is validated before a backtest is queued: every problem is listed at once with where it is. You can correct the rules by hand or ask the Copilot to. ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) - [Bring your own AI](https://algo.trado.trade/bring-your-own-ai) ## Related - [Algo trading without code](https://algo.trado.trade/learn/algo-trading-without-code.md): Pine Script, Python and APIs keep good ideas untested. What real no-code algo trading looks like: plain words in, readable rules out, Python if you want it. - [Algo trading for beginners](https://algo.trado.trade/for-beginners.md): Start algo trading without code or a data pipeline: describe an idea, backtest it honestly, ask the Copilot why, and paper-trade before you risk a rupee. - [Bring your own AI via MCP](https://algo.trado.trade/bring-your-own-ai.md): Connect Claude, Cursor or ChatGPT to Trado Strategy Lab over MCP. Your agent builds strategies, runs backtests and reads reports for free; runs use credits. - [Crypto backtesting after every cost](https://algo.trado.trade/backtesting.md): Backtest crypto perpetuals on years of 1-minute data for 900 coins, after fees, funding, slippage and liquidation, with a verdict that admits weak evidence. --- Source: https://algo.trado.trade/paper-trading # From backtest to paper trade, without rewriting anything The cliff between a backtest and a live bot is usually a rewrite. In the Lab the same rules run on paper and live through one runner, so a strategy you tested is the strategy you watch. Paper first, always; live only when you decide. ## The same rules, the same code A deployment pins a version of your strategy and runs it on the Lab's own runtime. It reads a live Binance kline feed and uses the same decision and position code as the backtest, then turns each decision into a canonical action that an adapter delivers. There is no second implementation to drift away from the first. ## One click to paper Starting a paper deployment is immediate: choose the strategy, the coins and the settings, and it runs on live prices with no real orders. You can watch its positions and events as they happen, pause it, or stop it. A strategy you backtested with leverage caps ignored is marked as research only and cannot be deployed. - Paper deployments start at once and never touch an account. - Positions, actions and events are recorded, so you can compare what happened with what the backtest expected. - Deployments are charged per active day in credits, and your plan sets how many can exist at once. ## Going live is a separate, deliberate step Live trading goes through an exchange adapter. You set up the adapter's credentials yourself and approve the live deployment in a signed-in browser session. A personal access token, and therefore any AI agent connected through one, can create and manage paper deployments but cannot approve a live one. Paper results are evidence, not a guarantee. Real fills, latency and exchange behaviour differ from a simulation, which is one reason to run on paper first and compare it with the backtest before risking money. ## Paper trading questions ### Does paper trading use real money? No. A paper deployment runs on live prices and records what it would have done. It never places an order on an exchange. ### How long should I paper trade before going live? That is your decision. Long enough to see the strategy behave in more than one market condition, and to compare its trades with the backtest's. The Lab does not set a number for you. ### Can my AI agent put a strategy live? No. Live approval needs you, in a browser session. Agents connected with a token can only create and manage paper deployments. See [bring your own AI](https://algo.trado.trade/bring-your-own-ai.md). ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) - [See pricing](https://algo.trado.trade/pricing) ## Related - [From backtest to live trading](https://algo.trado.trade/learn/from-backtest-to-live.md): Why strategies break between a backtest and a live bot, and a sane path through paper trading. The Lab runs the same rules in both, so nothing is rewritten. - [Crypto backtesting after every cost](https://algo.trado.trade/backtesting.md): Backtest crypto perpetuals on years of 1-minute data for 900 coins, after fees, funding, slippage and liquidation, with a verdict that admits weak evidence. - [Leverage and liquidation guard](https://algo.trado.trade/liquidation-and-leverage.md): At 20×, a move of about 4.5% against you loses the whole margin. The Lab models liquidation, refuses stops past it, and audits the leverage each coin allowed. - [Bring your own AI via MCP](https://algo.trado.trade/bring-your-own-ai.md): Connect Claude, Cursor or ChatGPT to Trado Strategy Lab over MCP. Your agent builds strategies, runs backtests and reads reports for free; runs use credits. --- Source: https://algo.trado.trade/liquidation-and-leverage # Leverage kills quietly. The Lab shows you the number. At 20× leverage, a move of about 4.5% against you loses the whole margin. Most tools never put that number next to your strategy. The Lab does, before you run and again in the report. ## The number that matters On an isolated position, the move against you that liquidates it is roughly 100 divided by the leverage, minus the exchange's maintenance margin. With the Lab's default maintenance margin of 0.5% of notional, a position at 20× is liquidated by a 4.5% adverse move, and at 5× by a 19.5% one. Higher leverage does not just magnify gains; it shrinks the room a strategy has to be wrong. The run form states this next to the leverage field, in the form “A move of 4.5% against you would liquidate this position at 20×”, so the figure is in front of you while you choose. ## What the backtest models - **Liquidation**, for isolated and cross margin, inside the simulation: a liquidated position ends, and its loss counts. - **Per-coin leverage caps.** Each coin trades at the lower of your leverage and its own cap, from the execution profile's table or the exchange's value. Backtests accept up to 125×; deployments up to 100×. - **The liquidation guard.** If a stop-loss sits at or beyond the liquidation price at your leverage, the run is refused with an explanation, because that stop could never trigger. You can acknowledge the warning and run anyway. - **The leverage audit** in every report: which caps applied, to which coins, and where they came from. ## Research runs are labelled as research You can run a what-if with leverage caps ignored. The report says so, and a strategy tested that way cannot be deployed, so a research shortcut never becomes a live position by accident. ## What this does not do The Lab models liquidation from an execution profile's maintenance margin and rules. A real exchange account can differ in the details, such as tiered margin or auto-deleveraging in extreme markets. Trading leveraged derivatives can lose more than you expect, including the whole margin. This is software, not advice, and a good backtest is not a promise. Read the long version in [leverage and liquidation, explained](https://algo.trado.trade/learn/leverage-and-liquidation-explained.md). ## Leverage questions ### What leverage should I use? The Lab does not tell you. It shows what each setting means: the adverse move that liquidates you, how often your backtest was liquidated, and what each coin allowed. Lower leverage leaves more room for a strategy to be wrong. ### What does the liquidation guard refuse? A run whose stop-loss lies at or beyond the isolated liquidation price at its leverage. The position would be liquidated before the stop could fire. The refusal explains why; you may acknowledge it and run anyway. ### Does it model cross margin? Yes, backtests model both isolated and cross liquidation. ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) - [See pricing](https://algo.trado.trade/pricing) ## Related - [Leverage and liquidation explained](https://algo.trado.trade/learn/leverage-and-liquidation-explained.md): At 20×, a move of about 4.5% against you wipes out the margin. How liquidation price works on crypto perpetuals, isolated versus cross, and what to check. - [Crypto backtesting after every cost](https://algo.trado.trade/backtesting.md): Backtest crypto perpetuals on years of 1-minute data for 900 coins, after fees, funding, slippage and liquidation, with a verdict that admits weak evidence. - [Paper trading for crypto strategies](https://algo.trado.trade/paper-trading.md): Take a backtested strategy to paper trading in one click. The same rules run on live Binance prices; going live is a separate step you approve. - [Algo trading for beginners](https://algo.trado.trade/for-beginners.md): Start algo trading without code or a data pipeline: describe an idea, backtest it honestly, ask the Copilot why, and paper-trade before you risk a rupee. --- Source: https://algo.trado.trade/for-beginners # Learn algo trading by testing your own ideas You do not need to code, download data or know a trading API to start. Describe an idea, see how it would have done after every cost, ask the Copilot why, and paper-trade it. The Lab keeps the first steps free of risk and says so when the evidence is thin. ## A first week that makes sense A path that gets you from a thought to a tested idea without a wall of setup. - **Start from an example or a sentence.** Six ready-made strategies open as editable rules, or write the idea in plain words and let the Copilot draft it. - **Run a backtest.** Pick a few coins and a period. The price in credits is shown first. - **Read the verdict.** It says whether the strategy made money broadly, in some conditions only, or too rarely to tell, and how strong the evidence is. - **Ask why.** Every result has an “Ask why” button. The Copilot explains what happened in plain language. - **Paper-trade it** before you think about real money. ## What you do not need - **Code.** The builder, the canvas and the Copilot cover the common ground. Python is optional. - **Data.** The Lab already has years of 1-minute data for 900 Binance perpetuals. - **Capital.** Backtests and paper trading do not use real money. - **A subscription.** Building is free, and you pay in credits only for the work you run. ## Guardrails for people still learning Beginners most often lose money by trusting one pretty backtest, using too much leverage, or never testing outside the period they tuned on. The Lab leans against each of those: costs are charged in every run, the evidence meter and verdict call out thin results, the optimizer validates on a holdout, and the liquidation guard and leverage audit put the risk of leverage in front of you. It will not promise that a strategy works. Read more in [why backtests lie](https://algo.trado.trade/learn/why-backtests-lie.md) and [learning algo trading with AI](https://algo.trado.trade/learn/learn-algo-trading-with-ai.md). ## Beginner questions ### Do I need to know how to trade first? It helps to have an idea, even a rough one. You do not need to know how to build a bot. The Copilot helps turn the idea into rules and explains the results. ### How much money do I need? None to start. Backtests and paper trading do not use real money. They use credits, and building strategies is free. ### Is it safe for a beginner? The Lab is built to slow you down before real money: paper first, live only by your explicit approval, and risk figures shown before you run. Trading leveraged crypto can still lose money quickly, and nothing here is advice. ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) - [See pricing](https://algo.trado.trade/pricing) ## Related - [AI strategy builder, no code needed](https://algo.trado.trade/ai-strategy-builder.md): Describe a trading idea in plain words and the Copilot drafts the rules. Refine them in a form or on a canvas, or write Python. No code required. - [Learn algo trading with AI](https://algo.trado.trade/learn/learn-algo-trading-with-ai.md): Forums and videos cannot explain your own strategy; an AI that reads your actual runs can. What to ask, a learning loop, and where AI needs checking. - [Algo trading without code](https://algo.trado.trade/learn/algo-trading-without-code.md): Pine Script, Python and APIs keep good ideas untested. What real no-code algo trading looks like: plain words in, readable rules out, Python if you want it. - [Leverage and liquidation guard](https://algo.trado.trade/liquidation-and-leverage.md): At 20×, a move of about 4.5% against you loses the whole margin. The Lab models liquidation, refuses stops past it, and audits the leverage each coin allowed. --- Source: https://algo.trado.trade/bring-your-own-ai # Prefer your own agent? Connect it. The Lab speaks MCP, the open protocol AI assistants use to call tools. Connect Claude, Cursor, ChatGPT or any MCP client with a personal access token, and your agent can build strategies, run backtests and optimizations, and read the reports. The AI is yours and costs nothing extra. ## What your agent can do The MCP server exposes the same tools the in-app Copilot uses, limited to what the token's scopes allow. - Create, validate, update and version strategies, and write custom indicators. - Run backtests, optimizations and parameter sweeps, and run studies that measure what happens after a named condition across the whole universe. - Read reports, trades, equity curves and insights, and summarise whole batches of runs. - Check what a run would cost before it is queued, and read your credits and limits. - Create and watch paper deployments. ## How to connect Sign in, open Settings → Tokens, and create a token with the scopes you want: `read`, `run` (strategies, runs, optimizations) or `deploy` (paper deployments). The token acts as you, in your own lab. Then add the Lab's MCP endpoint to your client. With Claude Code, for example: `claude mcp add --transport http trado-lab https://algo.trado.trade/mcp --header "Authorization: Bearer "`. Other clients take the same URL and bearer token in their MCP settings. Tell your agent to call `get_agent_guide` first: it is a short briefing on the Lab's data, strategy format, limits and safety rules. ## Safe by design - **Scopes.** A token sees only the tools its scopes allow. A read-only token cannot start a run. - **No live trading.** Tokens can manage paper deployments only. Approving a live deployment needs you, in a browser session. - **Spend checks.** Agents can validate a run or search to see its problems and its credit price before queueing anything. - **Rate limits** per token, so a runaway agent cannot flood the Lab. Use one token per project and revoke it any time. ## What it costs Connecting your own AI is free: you pay your AI provider whatever you already pay them, and the Lab charges nothing for the connection or for reading. The backtests and optimizations your agent starts use credits at the same prices as runs you start yourself. See [pricing](https://algo.trado.trade/pricing.md). ## Bring-your-own-AI questions ### Which AI assistants work? Claude, Cursor and ChatGPT, and any other client that supports MCP over streamable HTTP with bearer-token authentication. ### Why use my own agent instead of the Copilot? If you already pay for an assistant you like, it can use the Lab at no extra AI cost, and it keeps the context of your other work. The in-app Copilot is there for people who want everything in one place. ### How do I stop an agent's access? Revoke its token in Settings → Tokens. The Lab refuses a revoked token from then on. ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) - [See pricing](https://algo.trado.trade/pricing) ## Related - [AI strategy builder, no code needed](https://algo.trado.trade/ai-strategy-builder.md): Describe a trading idea in plain words and the Copilot drafts the rules. Refine them in a form or on a canvas, or write Python. No code required. - [Paper trading for crypto strategies](https://algo.trado.trade/paper-trading.md): Take a backtested strategy to paper trading in one click. The same rules run on live Binance prices; going live is a separate step you approve. - [Learn algo trading with AI](https://algo.trado.trade/learn/learn-algo-trading-with-ai.md): Forums and videos cannot explain your own strategy; an AI that reads your actual runs can. What to ask, a learning loop, and where AI needs checking. - [Pricing: pay for the work you run](https://algo.trado.trade/pricing.md): Building strategies is free. Backtests, optimizations, Copilot replies and deployments use credits, priced by the work, and you see the price before you run. --- Source: https://algo.trado.trade/learn # Learn algo trading without the jargon Eight guides, one for each thing that most often goes wrong between an idea and a strategy you can trust. Each explains the problem with concrete examples and shows how Trado Strategy Lab handles it. ## The eight problems Start with the one that sounds most like you. - [Why most crypto backtests lie](https://algo.trado.trade/learn/why-backtests-lie.md): Look-ahead, missing costs, survivorship and ignored liquidation flatter backtests. A checklist to test any backtest, and how the Lab answers each point. - [Algo trading without code](https://algo.trado.trade/learn/algo-trading-without-code.md): Pine Script, Python and APIs keep good ideas untested. What real no-code algo trading looks like: plain words in, readable rules out, Python if you want it. - [Crypto data for backtesting: the chore before the strategy](https://algo.trado.trade/learn/crypto-data-for-backtesting.md): Clean 1-minute crypto data is a project before it is a strategy. What good backtesting data needs, where DIY pipelines break, and what the Lab holds. - [Overfitting feels like success](https://algo.trado.trade/learn/overfitting-in-backtests.md): Tune a strategy long enough and the past looks perfect. Why overfitting feels like success, and the checks that expose it: holdouts, walk-forward, stability. - [From backtest to live: closing the cliff](https://algo.trado.trade/learn/from-backtest-to-live.md): Why strategies break between a backtest and a live bot, and a sane path through paper trading. The Lab runs the same rules in both, so nothing is rewritten. - [Leverage and liquidation, explained](https://algo.trado.trade/learn/leverage-and-liquidation-explained.md): At 20×, a move of about 4.5% against you wipes out the margin. How liquidation price works on crypto perpetuals, isolated versus cross, and what to check. - [What algo trading tools really cost](https://algo.trado.trade/learn/what-algo-trading-tools-cost.md): Subscriptions for tools you use twice a month add up. What to look for in algo trading pricing, and how credits priced by the work compare. - [Learn algo trading with someone to ask](https://algo.trado.trade/learn/learn-algo-trading-with-ai.md): Forums and videos cannot explain your own strategy; an AI that reads your actual runs can. What to ask, a learning loop, and where AI needs checking. ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) - [See pricing](https://algo.trado.trade/pricing) ## Related - [Crypto backtesting after every cost](https://algo.trado.trade/backtesting.md): Backtest crypto perpetuals on years of 1-minute data for 900 coins, after fees, funding, slippage and liquidation, with a verdict that admits weak evidence. - [Algo trading for beginners](https://algo.trado.trade/for-beginners.md): Start algo trading without code or a data pipeline: describe an idea, backtest it honestly, ask the Copilot why, and paper-trade before you risk a rupee. - [AI strategy builder, no code needed](https://algo.trado.trade/ai-strategy-builder.md): Describe a trading idea in plain words and the Copilot drafts the rules. Refine them in a form or on a canvas, or write Python. No code required. --- Source: https://algo.trado.trade/learn/why-backtests-lie # Why most crypto backtests lie A backtest is a story about the past, and the story is only true if it was charged for everything the market would have charged. Most are not. Here are the five ways a backtest flatters a strategy, and the questions that expose each one. ## Five ways a backtest lies None of these needs bad intent. They are the default behaviour of a simple backtester. - **Look-ahead.** The strategy trades on information it could not have had: acting on a bar's close at that bar's open, or using an indicator that quietly reads future values. Results look excellent, and no real account can reproduce them. - **Missing costs.** Fees on every entry and exit, funding on every position held across a funding time, and slippage when the price moves while your order fills. A strategy that trades often can be profitable before costs and a loser after them. - **Survivorship.** Testing only on the coins that are liquid today leaves out the ones that faded. The universe looks healthier than the one you would have faced. - **Ignored liquidation.** A leveraged position can be closed by the exchange in the middle of a bar. If the test only looks at closing prices, it never sees the wick that ended the trade. - **One lucky period.** A strategy that only ever met one kind of market looks robust until the market changes. ## An example with round numbers Take a hypothetical 20× long entered at 100. With a typical 0.5% maintenance margin it is liquidated near 95.5. Now imagine an hourly bar that dips to 94 and closes at 99. A backtest that only reads closes sees a 1% loss and a position still open. The exchange would have liquidated it, and the loss would have been the whole margin. The numbers are made up to show the mechanism; the mechanism is real on any leveraged perpetual. ## Questions to ask any backtest Before you trust a result from any tool, ask: - Are fees, funding and slippage charged on every trade, and can I see the numbers used? - What is the data granularity? Daily or hourly bars hide wicks that 1-minute bars show. - Is liquidation modelled, for the margin mode I would actually use? - Is there a period the strategy was never tuned on? - How many trades are behind the headline number, and does the tool say when that is too few? - Does it show results by market condition, or only the average? ## How the Lab answers each one - **Costs:** every trade is priced after the execution profile's fees, slippage and funding. - **Wicks:** runs replay 1-minute bars, and exact-exit runs use the trade tape for stop and target fills. - **Liquidation:** modelled for isolated and cross margin, with a guard that refuses a stop placed beyond the liquidation price. - **Held-out data:** optimizations validate on a holdout and walk forward through time. - **An honest verdict:** results are grouped by market regime, the verdict can say “No clear edge” or “Too few trades to tell”, and an evidence meter reads Not enough, Weak, Moderate or Strong. A strategy does not get a green light for paper gains. ## Common questions ### Can any backtest be trusted? A backtest can be honest about the past without predicting the future. Treat a good one as evidence that an idea deserves a paper trade, not as a forecast. ### How many trades are enough? There is no magic number. The Lab's evidence meter weighs the number of trades and the span of days, and says “Not enough” rather than giving a score when there is too little to judge. ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) - [How backtesting works](https://algo.trado.trade/backtesting) ## Related - [Crypto backtesting after every cost](https://algo.trado.trade/backtesting.md): Backtest crypto perpetuals on years of 1-minute data for 900 coins, after fees, funding, slippage and liquidation, with a verdict that admits weak evidence. - [Overfitting in backtests](https://algo.trado.trade/learn/overfitting-in-backtests.md): Tune a strategy long enough and the past looks perfect. Why overfitting feels like success, and the checks that expose it: holdouts, walk-forward, stability. - [Leverage and liquidation guard](https://algo.trado.trade/liquidation-and-leverage.md): At 20×, a move of about 4.5% against you loses the whole margin. The Lab models liquidation, refuses stops past it, and audits the leverage each coin allowed. --- Source: https://algo.trado.trade/learn/algo-trading-without-code # Algo trading without code Plenty of people have a market idea they have never tested because testing meant learning Pine Script, then Python, then an exchange API. No-code algo trading removes that gate, but only if it produces rules you can read, not a black box you have to trust. ## The coding gate To go from an idea to a tested strategy you normally need a scripting language for the chart, a data source, a backtesting library, and later a bot framework and an exchange API to run it. Each is a skill of its own, and each is a place to introduce a bug that makes the result wrong. Many ideas never get past the first step. ## What no-code should mean A tool that hides the logic is not better than code; it is just harder to check. Good no-code has three properties. - **The strategy is explicit.** Indicators, entry conditions, exits, stops and risk settings are written down as rules you can read and edit. - **You can say it in words.** The first draft can come from a sentence, which is then checked rather than trusted. - **Code is an option, not a wall.** When the idea needs something unusual, you can write it, in the same engine. ## From a sentence to rules Say you write: “Buy when price closes above the 200-period EMA on the daily chart and RSI crosses back up through 30 on the 1-hour. Sell when RSI passes 70 or the trade is down 3%.” A good builder turns that into something you can inspect: - **Indicators:** EMA(200) on 1d, RSI(14) on 1h. - **Entry:** daily close above the EMA, and RSI crossing up through 30. - **Exits:** RSI above 70, or a 3% stop-loss. - **Risk:** position size and leverage, shown with the liquidation figure they imply. ## How the Lab does it In the Lab you can describe the idea to the Copilot, which drafts the rules; build them in a form or on a canvas that draws the same rules as a graph; or write Python against a small SDK. All three produce the same strategy definition, so you can start in words and finish by hand. There are 170 built-in indicators and six ready-made examples to start from, and strategies are versioned so you can always go back. Whatever you build is then tested under the same honest rules: costs on every trade, liquidation modelled, a verdict that admits weak evidence. See [the AI strategy builder](https://algo.trado.trade/ai-strategy-builder.md). ## The pitfall: no code does not mean no checking Removing the code removes bugs from your code, not flaws from the idea or from the backtest. Read the rules the tool wrote, run the strategy on periods and coins it was not built around, and look at the result by market condition before you believe it. Read [why backtests lie](https://algo.trado.trade/learn/why-backtests-lie.md). ## Common questions ### Will I ever need Python? Only if you want logic the builder does not express, such as a custom indicator. The Lab supports Python strategies and indicators in a sandbox, and checks custom indicators for look-ahead. ### Can I use ChatGPT or Claude instead? Yes. Your own agent can build and test strategies in the Lab through MCP. See [bring your own AI](https://algo.trado.trade/bring-your-own-ai.md). ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) - [Bring your own AI](https://algo.trado.trade/bring-your-own-ai) ## Related - [AI strategy builder, no code needed](https://algo.trado.trade/ai-strategy-builder.md): Describe a trading idea in plain words and the Copilot drafts the rules. Refine them in a form or on a canvas, or write Python. No code required. - [Algo trading for beginners](https://algo.trado.trade/for-beginners.md): Start algo trading without code or a data pipeline: describe an idea, backtest it honestly, ask the Copilot why, and paper-trade before you risk a rupee. - [Bring your own AI via MCP](https://algo.trado.trade/bring-your-own-ai.md): Connect Claude, Cursor or ChatGPT to Trado Strategy Lab over MCP. Your agent builds strategies, runs backtests and reads reports for free; runs use credits. --- Source: https://algo.trado.trade/learn/crypto-data-for-backtesting # Crypto data for backtesting: the chore before the strategy Before you can test a single idea you need years of clean price data, funding rates and each coin's trading rules. Building that pipeline is a project of its own, and the mistakes in it end up inside your results. Here is what good data needs and where homemade versions go wrong. ## It is bigger than it looks One year of 1-minute bars is 525,600 rows for a single coin. Multiply by hundreds of coins and several years and you are managing a serious dataset: downloading it politely from exchange APIs that rate-limit you, storing it in a format you can query quickly, and keeping it current every day. ## What good backtesting data needs - **1-minute (or finer) bars.** Hourly or daily bars hide the intraday wicks that trigger stops and liquidations. - **Consistent timestamps.** One timezone (UTC) and no silent shifts between files. - **Gaps that are visible.** Missing bars should be known and handled, not interpolated away. - **Funding rates**, because holding a perpetual across a funding time costs or pays money. - **Instrument rules:** tick size, lot size and leverage limits per coin, as they applied at the time. - **Enough history before the test period** for indicators to warm up. A 200-day average is meaningless on day one. - **An honest universe.** Ask which coins are included and whether ones that faded or were delisted are part of what the strategy is tested on. ## A quick audit of any dataset Whether you built it or bought it, a few checks catch most of the damage before it reaches a result. - **Count bars per day.** A full UTC day of 1-minute bars is 1,440. Days well short of that point to gaps. - **Look for flat runs.** Many identical closes in a row usually mean a feed stalled and the last price was repeated. - **Check each bar is possible.** The high must be at least the open and close, and the low at most; the volume must not be negative. - **Check order.** Timestamps must be strictly increasing, with no duplicates after a merge. - **Compare a sample with a second source.** A handful of random days will show a systematic offset. - **Look at the odd coins.** Newly listed ones, renamed symbols and thin markets are where errors cluster. ## Where do-it-yourself pipelines break The common failures are quiet ones: a bad tick that creates a fake trade, a symbol that changed name, a day of missing bars filled with the last price, a timezone off by 5.5 hours. Each can flatter a strategy without raising an error. You find out when a live account disagrees with the backtest. ## What the Lab already holds The Lab keeps years of 1-minute bars for 900 Binance USDⓈ-M perpetuals, with funding rates and instrument data, refreshed twice a day. You choose coins and dates and run; there is nothing to download. History matters at the edges too. Each coin is loaded with the warm-up its indicators need, sized per coin. If a coin was listed too recently for a filter to be ready, the run says which coin and until when, rather than trading without the filter and not telling you. ## Common questions ### Can I bring my own data? The Lab runs on its own curated Binance perpetuals dataset, so results are comparable and reproducible. If you need an instrument it does not cover, the data is the first thing to check. ### Why 1-minute bars? Because stops, targets and liquidations happen inside bars. With 1-minute bars, and the trade tape for exact exits, a backtest sees the wicks that coarser data hides. See [backtesting](https://algo.trado.trade/backtesting.md). ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) - [How backtesting works](https://algo.trado.trade/backtesting) ## Related - [Crypto backtesting after every cost](https://algo.trado.trade/backtesting.md): Backtest crypto perpetuals on years of 1-minute data for 900 coins, after fees, funding, slippage and liquidation, with a verdict that admits weak evidence. - [Why crypto backtests lie](https://algo.trado.trade/learn/why-backtests-lie.md): Look-ahead, missing costs, survivorship and ignored liquidation flatter backtests. A checklist to test any backtest, and how the Lab answers each point. - [From backtest to live trading](https://algo.trado.trade/learn/from-backtest-to-live.md): Why strategies break between a backtest and a live bot, and a sane path through paper trading. The Lab runs the same rules in both, so nothing is rewritten. --- Source: https://algo.trado.trade/learn/overfitting-in-backtests # Overfitting feels like success You adjust a parameter, the equity curve improves, you adjust another, and it improves again. By the end the strategy looks brilliant. It may simply have memorised the past. Overfitting is the most common way a good-looking backtest fails, and it feels exactly like progress. ## What overfitting is Markets contain a repeating signal and a lot of noise. A strategy with enough adjustable parts can be fitted to the noise of one stretch of history, which will never repeat. The fit looks like skill because the test period is the same one you tuned on. ## Why trying many things makes it worse Test enough random settings on the same data and some will look good by luck alone, the way the best result of a thousand coin-flip sequences looks like a streak. The more combinations you try, the better the best one looks, and the less it means. This is why a strategy that is the winner of a big parameter search deserves more suspicion, not less. ## Warning signs in your own results You can often tell before running any formal test. - The best settings sit at the edge of the range you searched, which suggests the real optimum is somewhere you did not look, or that there is none. - The strategy has many rules and parameters for the amount of history, so it has room to memorise. - A tiny change, such as one bar on a moving-average length, swings the result from great to poor. - It works on the coin you tuned on and loses on a similar coin it never saw. - The equity curve is suspiciously smooth for a market that is not. - You can explain the result only after seeing it, never before. ## How to check for it - **A holdout.** Keep a period the search never saw, and judge the winner only on that. - **Walk-forward testing.** Tune on one window, test on the next, slide forward and repeat, so every result comes from data the settings were not fitted to. - **Neighbourhood stability.** If nudging a parameter slightly destroys the result, you found a spike, not an edge. A broad plateau of good settings is a better sign. - **Parameter importance.** Which settings actually drive the result, and are those the ones you can explain? - **Several coins and periods.** An idea that only works on one coin in one year is a story. - **By market condition.** A strategy that earns everything in one regime and bleeds in the rest is exposed when that regime ends. - **Fewer knobs.** Every free parameter is another way to fit noise. ## What the Lab does about it The optimizer supports grid, random, genetic, TPE, CMA-ES and NSGA-II searches and validates with a holdout and walk-forward windows. Results come with heatmaps, parameter importance and a neighbourhood-stability view. For the whole search it reports Monte Carlo resampling, the probability of backtest overfitting and the deflated Sharpe ratio, which discounts a result for the number of things you tried. On any run, Insights say where the strategy works and where it does not, with regime tables, and the evidence meter says how much to read into it. The verdict will say “No clear edge” rather than flatter a lucky run. See [backtesting](https://algo.trado.trade/backtesting.md). ## Common questions ### Is optimizing a strategy a mistake? No. Searching is how you learn which settings matter. The mistake is treating the best in-sample result as the answer without validating it on data the search never touched. ### What is walk-forward testing? You tune on a window of history, test the tuned settings on the period right after it, then slide both windows forward. Every test result comes from data the settings were not fitted to. ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) - [How backtesting works](https://algo.trado.trade/backtesting) ## Related - [Crypto backtesting after every cost](https://algo.trado.trade/backtesting.md): Backtest crypto perpetuals on years of 1-minute data for 900 coins, after fees, funding, slippage and liquidation, with a verdict that admits weak evidence. - [Why crypto backtests lie](https://algo.trado.trade/learn/why-backtests-lie.md): Look-ahead, missing costs, survivorship and ignored liquidation flatter backtests. A checklist to test any backtest, and how the Lab answers each point. - [Learn algo trading with AI](https://algo.trado.trade/learn/learn-algo-trading-with-ai.md): Forums and videos cannot explain your own strategy; an AI that reads your actual runs can. What to ask, a learning loop, and where AI needs checking. --- Source: https://algo.trado.trade/learn/from-backtest-to-live # From backtest to live: closing the cliff The backtest looked good, and then the bot behaved differently. Often the difference is not the market but the rewrite: the strategy that was tested is not quite the strategy that was deployed. A safer path keeps the rules identical and puts a paper stage in the middle. ## Where the cliff comes from Moving from research to a running bot usually means starting over. - The strategy is rewritten in a bot framework, and subtle differences creep in: a candle boundary, a rounding rule, an order type. - Webhooks, API keys and order plumbing have to be wired, each a new place for a bug. - Position sizing and leverage maths are redone by hand. - Nothing sits between “it backtested well” and real money. ## A sane path to live - **1. Backtest after costs,** and read the result by market condition, not only the average. - **2. Validate on data the search never saw:** a holdout or walk-forward windows. - **3. Paper-trade the same rules** on live prices with no real orders. - **4. Compare paper with the backtest:** entries, exits, trade count, and the gap between expected and actual fills. - **5. Start live small,** with limits you can afford to lose entirely, and watch it. - **6. Keep a stop-the-bot rule** you decided on in advance, not in the middle of a drawdown. ## What to compare on paper A paper run will not match the backtest exactly, and it should not. Look for differences you can explain: a trade the backtest took that paper skipped, or fills at slightly worse prices. Differences you cannot explain are the ones to chase before any money is involved. ## What to watch once it runs Decide the warning signs before the strategy is running, so you are not inventing them mid-drawdown. - **Trade frequency.** A strategy that suddenly trades far more or far less than its backtest did has changed, or the market has. - **Slippage and fees against the assumptions.** If real costs are larger than the model's, the edge shrinks with them. - **Funding paid or received,** for positions held across funding times. - **Drawdown against the backtest's worst.** A live drawdown beyond anything the backtest saw is a signal to stop and look. - **A written stop rule.** The point at which you pause the strategy, decided in advance. ## How the Lab closes the gap A deployment pins a version of your strategy and runs it on the Lab's own runtime, using the same decision and position code as the backtest, on a live Binance kline feed. Nothing is rewritten, so the strategy you tested is the strategy that runs. Paper starts in one click and never touches an account. Live goes through an exchange adapter you set up, and needs your explicit approval in a signed-in browser. A token, and therefore an AI agent, can manage paper deployments but cannot approve a live one. See [paper trading](https://algo.trado.trade/paper-trading.md). ## Common questions ### Will paper trading match the backtest exactly? No. Live prices, timing and fills differ from a replay. The aim is to find differences you can explain before real money is at stake. ### Does paper trading cost anything? Deployments are charged per active day in credits, and your plan sets how many can run at once. See [pricing](https://algo.trado.trade/pricing.md). ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) - [See pricing](https://algo.trado.trade/pricing) ## Related - [Paper trading for crypto strategies](https://algo.trado.trade/paper-trading.md): Take a backtested strategy to paper trading in one click. The same rules run on live Binance prices; going live is a separate step you approve. - [Crypto backtesting after every cost](https://algo.trado.trade/backtesting.md): Backtest crypto perpetuals on years of 1-minute data for 900 coins, after fees, funding, slippage and liquidation, with a verdict that admits weak evidence. - [Leverage and liquidation guard](https://algo.trado.trade/liquidation-and-leverage.md): At 20×, a move of about 4.5% against you loses the whole margin. The Lab models liquidation, refuses stops past it, and audits the leverage each coin allowed. --- Source: https://algo.trado.trade/learn/leverage-and-liquidation-explained # Leverage and liquidation, explained Leverage lets a small amount of margin control a larger position. It also means a small move against you can end the trade for you. At 20× leverage, a move of about 4.5% against you loses the whole margin. Here is the maths, so the number is in front of you before you click run. ## How leverage and margin work On a perpetual future you put up margin, and the exchange lets you hold a position several times larger. A position worth 20 times your margin is 20× leveraged. Gains and losses are measured on the full position, so a 1% price move changes your margin by 20%. ## The liquidation maths On an isolated position, the exchange closes it when the loss eats the margin down to a maintenance level. As a rule of thumb, the adverse move that liquidates you is 100 divided by the leverage, minus the maintenance margin. Using a 0.5% maintenance margin, which is the Lab's default, that gives: - **2×:** about 49.5% against you. - **5×:** about 19.5%. - **10×:** about 9.5%. - **20×:** about 4.5%. A move of that size against you loses the whole margin. - **50×:** about 1.5%. - **100×:** about 0.5%. ## Why wicks matter more than closes Liquidation happens on the move itself, not at the end of the hour. A bar that dips 6% and closes down 1% shows a 1% loss on a chart and would have liquidated a 20× long on the way. This is why a leveraged backtest on coarse bars is unreliable, and why the Lab replays 1-minute bars. ## Isolated versus cross margin In **isolated** margin each position has its own margin, so the most it can lose is that margin. In **cross** margin all of your free balance backs the position, which delays liquidation but exposes the whole account to one bad trade. Neither is safer in general; they fail in different ways. The Lab models both. ## Stops, size and leverage are three different things Leverage is not position size, and it is not risk. Risk per trade is set by how far away your stop is and how large the position is. The trap is a stop-loss placed beyond the liquidation price: the exchange will close the position first, and the stop never fires. The Lab's liquidation guard refuses that setup unless you acknowledge it, and every report carries a leverage audit that shows which caps applied to which coins. ## A checklist before you add leverage - What adverse move would liquidate me at this leverage, and have this coin's normal swings gone that far? - Is my stop closer than my liquidation price? - How many times was my backtest liquidated, not just how much it earned? - Does the coin allow this leverage? Each has its own cap. - Could I afford to lose the whole margin on this trade? ## Common questions ### Is the liquidation price exact? The Lab computes it from its execution profile's maintenance margin and rules. A real exchange account can differ in details such as tiered margin, so treat it as a close estimate and not a guarantee. This is software, not advice. ### Is there a safe leverage? No leverage is safe in itself. Lower leverage leaves more room for a trade to be wrong. See [leverage and liquidation in the Lab](https://algo.trado.trade/liquidation-and-leverage.md). ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) ## Related - [Leverage and liquidation guard](https://algo.trado.trade/liquidation-and-leverage.md): At 20×, a move of about 4.5% against you loses the whole margin. The Lab models liquidation, refuses stops past it, and audits the leverage each coin allowed. - [Why crypto backtests lie](https://algo.trado.trade/learn/why-backtests-lie.md): Look-ahead, missing costs, survivorship and ignored liquidation flatter backtests. A checklist to test any backtest, and how the Lab answers each point. - [Algo trading for beginners](https://algo.trado.trade/for-beginners.md): Start algo trading without code or a data pipeline: describe an idea, backtest it honestly, ask the Copilot why, and paper-trade before you risk a rupee. --- Source: https://algo.trado.trade/learn/what-algo-trading-tools-cost # What algo trading tools really cost Charting, data, backtesting, bot hosting and an AI assistant can each be a separate subscription, and most people use them in bursts. Before you pick a tool, work out what you will actually use, then look for pricing that matches it. ## The subscription trap A monthly plan makes sense when you use a tool every day. Algo research is rarely like that. You spend a weekend on a batch of ideas, then a quiet month waiting for something to paper-trade. Paying the same every month for bursty use means paying for capacity you are not using. ## What to compare between tools Headline price tells you little. Ask what is included, and what is metered or locked. - **Data:** is it included, and at what resolution and history? - **Backtest compute:** how many runs, how large, and is anything queued or capped? - **Optimization:** is it available, and how are big searches limited? - **AI help:** included, metered, or bring-your-own? - **Paper and live running:** is hosting a separate cost? - **What happens if you stop paying:** can you still read your results? - **Tax:** are prices shown with GST? ## Estimate your own usage first Before comparing prices, write down how you will really use a tool. The honest answers decide which pricing model suits you. - How many new ideas will you test in a month, and how many backtests does each one take? Most take many. - Will you optimize parameters, which is far heavier than a single run? - How many coins and how many years does a typical run cover? - Will you paper-trade or run live, and for how many days at a time? - Are there months you will not touch it at all? ## Pricing by the work instead The Lab prices work in credits. Building and editing strategies is free. A backtest, an optimization, a Copilot reply or a day of paper deployment costs credits in proportion to how much work it is, and you see the price before you confirm. An optimization is reserved at its time budget and charged for the time it actually uses, and a backtest or optimization that fails is refunded. You can buy credits as packs and use them for anything, or take a plan whose allowance renews each period. Neither renews by itself, so there is no subscription to forget to cancel. See [pricing](https://algo.trado.trade/pricing.md) for the current prices; they are read live from the price list. ## The AI does not have to be a line item If you already pay for Claude, Cursor or ChatGPT, you can connect it to the Lab over MCP and use it at no extra AI cost. The backtests it starts use credits like any other run. See [bring your own AI](https://algo.trado.trade/bring-your-own-ai.md). ## Common questions ### What is the cheapest way to try the Lab? Building strategies is free, and new accounts get welcome credits to run a first few backtests. Waitlist founding members get 500 launch credits on top of the welcome credits. See the [waitlist](https://algo.trado.trade/waitlist). ### Will I see what something costs before I run it? Yes. Runs, optimizations and deployments show their price in credits before you confirm, and agents can check a price without queueing anything. ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) - [See pricing](https://algo.trado.trade/pricing) ## Related - [Pricing: pay for the work you run](https://algo.trado.trade/pricing.md): Building strategies is free. Backtests, optimizations, Copilot replies and deployments use credits, priced by the work, and you see the price before you run. - [Bring your own AI via MCP](https://algo.trado.trade/bring-your-own-ai.md): Connect Claude, Cursor or ChatGPT to Trado Strategy Lab over MCP. Your agent builds strategies, runs backtests and reads reports for free; runs use credits. - [Join the waitlist](https://algo.trado.trade/waitlist.md): Join the Trado Strategy Lab waitlist for 500 launch credits, early access by invitation and referral rewards. Launch: 1 November 2026, 00:00 IST. --- Source: https://algo.trado.trade/learn/learn-algo-trading-with-ai # Learn algo trading with someone to ask Most people learn algo trading from forums and long videos, and none of them can look at your strategy and tell you why it lost in March. An assistant that can read your actual runs can. Here is how to use one well, and where it still needs checking. ## The gap in how people learn General tutorials teach concepts. They cannot tell you why your particular strategy behaves the way it does, so you are left guessing, re-reading and changing three things at once. That is slow, and it teaches the wrong lessons. ## What an AI is good at here - **Explaining a result in plain words:** why a run lost money, what a metric means, why a stop was hit. - **Turning an idea into explicit rules,** which you then read and correct. - **Suggesting the next experiment,** such as testing on other coins or splitting results by market condition. - **Spotting suspicious results:** very few trades, one coin doing all the work, a curve that looks too smooth. ## A learning loop that works Change one thing at a time, and ask why before you change the next. - **State the hypothesis** in a sentence: what should make this strategy earn money? - **Backtest it** on several coins and a period it was not built around. - **Read the verdict and the regime table,** then ask why for anything surprising. - **Change one thing,** rerun, and compare. - **Validate** on held-out data, then paper-trade. ## Good questions to ask - “Which trades did most of the damage, and what did they have in common?” - “Does this only work in bull markets?” - “If I halve the leverage, what happens to the worst drawdown?” - “Is the evidence for this strong enough to paper-trade?” ## An illustration Imagine you ask: “Why did this strategy lose in March?” A useful answer reads the run, not the internet. It might say that most of the losing trades were longs entered in a sideways regime, that they were stopped out quickly and repeatedly, and that the strategy made money in the trending months around it. It would then suggest a test, such as adding a trend filter and rerunning on the same period. That is a hypothesis you can check in minutes, which is the whole point. (This is an invented example, to show the shape of a good answer.) ## Where AI still needs checking An assistant that guesses is worse than none. The value comes from being grounded in your real data: the Lab's Copilot explains runs by reading them with the same tools it uses to build and run strategies, and asks you before anything expensive or live. Still read the rules it writes, treat a confident explanation as a hypothesis to test, and remember that no assistant can promise a strategy will keep working. The same applies to your own agent. See [bring your own AI](https://algo.trado.trade/bring-your-own-ai.md). ## Common questions ### Does the AI replace learning the basics? No. It speeds up the loop between an idea and an answer, which is where most learning happens. You still need to understand costs, leverage and overfitting, which is why the Lab shows them and links to explanations. ### Do I have to use the Lab's Copilot? No. Claude, Cursor or ChatGPT can use the Lab through MCP. See [bring your own AI](https://algo.trado.trade/bring-your-own-ai.md). ## Next steps - [Join the waitlist](https://algo.trado.trade/waitlist) - [Bring your own AI](https://algo.trado.trade/bring-your-own-ai) ## Related - [AI strategy builder, no code needed](https://algo.trado.trade/ai-strategy-builder.md): Describe a trading idea in plain words and the Copilot drafts the rules. Refine them in a form or on a canvas, or write Python. No code required. - [Algo trading for beginners](https://algo.trado.trade/for-beginners.md): Start algo trading without code or a data pipeline: describe an idea, backtest it honestly, ask the Copilot why, and paper-trade before you risk a rupee. - [Overfitting in backtests](https://algo.trado.trade/learn/overfitting-in-backtests.md): Tune a strategy long enough and the past looks perfect. Why overfitting feels like success, and the checks that expose it: holdouts, walk-forward, stability.