Fetching Strategies

The primary role of the AI Backtesting Assistant is to help users find strategies quickly without having to iterate through various conditions, tickers, and timeframes on TradingView charts. Instead, users can simply ask the AI to look for the strategy that would best match the criteria they want. The AI Backtesting Assistant has access to thousands of unique strategies, half using our S&O toolkit and the other half using our PAC toolkit.

​

Strategies Configuration

Strategies accessible by the AI Backtesting Assistant are backtested using an initial capital of 10,000$, without commissions/slippage, and using default settings (except for the PAC which uses a different invalidation behavior). Unit position sizing is used across all tickers, except forex pairs which use a standard lot (100,000 units). The strategies the AI can access don’t use every single one of the features within the toolkits, see the section below to learn more about the supported tickers, timeframes, and backtesting conditions.

​

Supported Tickers

The following tickers are supported:

Forex

Stocks

ETFs

Crypto (Coinbase)

Commodities

Futures

​

Supported Timeframes

5m, 15m, and 60m timeframes are supported.

---

​

Classics Conditions

Our AI can fetch strategies constructed from a wide variety of classical indicators including:

• EMA (Exponential Moving Average, using periods 20, 50, 200)
• SuperTrend
• VWAP (Volume Weighted Average Price)
• Ichimoku
• Linear Regression
• Bollinger Bands
• Donchian Channel
• Aroon
• RSI (Relative Strength Index)
• MFI (Money Flow Index)
• Stochastic
• MACD (Moving Average Convergence Divergence)
• Volume
• Volume Delta

AI strategies constructed from classical indicators use a single trigger and a combination of up to two filters for the long and short entry conditions. These conditions are listed out below.

​

Triggers

Triggers determine the initial conditions used by a strategy. These tend to be less persistent events on a chart, occurring for a single bar.

Bullish Triggers

Bearish Triggers

​

Filters

Filters add precision to a strategy, potentially filtering out bad trades that would have been taken by simply using triggers. Filters tend to be more persistent, lasting for multiple bars on a chart.

Bullish Filters

Bearish Filters

---

​

Signals & Overlays Conditions

AI strategies constructed from our Signals & Overlays (S&O) toolkit use a single trigger and a combination of up to two filters for the long and short entry conditions. These conditions are listed out below.

​

Triggers

Triggers determine the initial conditions used by a strategy. These tend to be less persistent events on a chart, occurring for a single bar.

Bullish Triggers

Bearish Triggers

​

Filters

Filters add precision to a strategy, potentially filtering out bad trades that would have been taken by simply using triggers. Filters tend to be more persistent, lasting for multiple bars on a chart.

Bullish Filters

Bearish Filters

Neutral Conditions

​

Exits

Strategies using the S&O toolkit can make exit trades which are outlined below:

Exit Conditions

---

​

Oscillator Matrix Conditions

AI strategies constructed from our Oscillator Matrix (OSC) toolkit use a single trigger and a combination of up to two filters for the long and short entry conditions. These conditions are listed out below.

​

Triggers

Triggers determine the initial conditions used by a strategy. These tend to be less persistent events on a chart, occurring for a single bar.

Bullish Triggers

Bearish Triggers

​

Filters

Filters add precision to a strategy, potentially filtering out bad trades that would have been taken by simply using triggers. Filters tend to be more persistent, lasting for multiple bars on a chart.

Bullish Triggers

Bearish Triggers

---

​

Price Action Concepts Conditions

Strategies based on our Price Action Concepts toolkit (PAC) are based on sequences, entering a position when a sequence is completed without any repeated steps. PAC strategies can use a maximum of three steps.

​

Steps

Bullish Steps

Bearish Conditions

​

How To Fetch Strategies

Interacting with the AI Backtesting Assistant is straightforward. All users need to do is type a prompt in the chat box located at the bottom of their screen. Here are examples of prompts that will trigger the AI Backtesting Assistant to retrieve a trading strategy:

• What is the best performing strategy across all crypto tickers on the 5-minute timeframe?
• Give me three strategies using fair value gaps that have a winrate above 80%.
• Find a strategy with above average net profit but below average drawdown.

Users can also click on one of the suggested prompts in the search group located above the input prompt. The AI can access the following backtesting results for strategies:

• Net Profit
• Max Net Profit
• Closed Trades
• Winning Trades
• Losing Trades
• Winrate
• Gross Profit
• Gross Loss
• Profit Factor
• Max Drawdown
• Max Drawdown Percent
• Average Trade
• Average Winning Trade
• Average Losing Trade

These results will be returned in a table display within the chat. If multiple strategies are returned, they will all be located within that table. The AI can occasionally fetch strategies in the middle of a response; this will return another table containing the strategies fetched. The AI has also access to the following information from a strategy:

• Strategy start time: When the strategy started being evaluated
• Strategy current position (long, short, or none)

​

Fetch Statistics Across The Entire Strategy Population

The AI can also get general statistics and data across the whole population of strategies being backtested. Example of prompts doing this are:

• What is the total net profit of all bitcoin strategies you have?
• On BTCUSD, are strategies using normal Confirmation signals better than the ones using strong Confirmation signals overall?
• What is total net profit of profitable strategies across stocks?

This can be useful to get big picture information about strategies.

​

Add a Strategy to Your Chart

Once you get a strategy you like from our AI, you may be inclined to add it to your chart to visualize it & further optimize its results on TradingView. To do so, follow these steps:

​

Saving Strategies

Users can export strategies displayed within a table as a csv using the download icon located under it.

​

Prompting Tips

The AI can make mistakes when fetching strategies, as this is a very complex process depending on the user query. We will always aim to improve the AI Backtesting Assistant over time; however, the following tips can help you get the most from our AI.

​

Make Detailed Queries

Make sure to detail every aspect of the strategy you want to fetch if you have a precise one in mind, for example:

Original	Improved
Fetch a strategy that uses Confirmation Signals	Fetch a strategy that only uses Confirmation Signals and no other conditions
I want a PAC strategy that goes long on a new FVG, and CHoCH	I want a PAC strategy that goes long when: - Step 1: a new bullish FVG occur - Step 2: a bullish CHoCH occur
I want the best strategy	I want the strategy with the highest net profit on BTCUSDT 5 minute

Context and details will help the AI avoid making guesses about your query.

​

Encourage the AI to Improve

The AI Backtesting Assistant can be fairly self-aware if a fetched strategy doesn’t meet a user criteria, and will then retry. But if the AI response is clearly mistaken, then it can be good to let it know using a prompt encouraging re-evaluation of the fetching process:

The strategy your returned is not the one I want. Correct the issue in your fetching process and retry to fetch the strategy in accordance with my query.

​

Split Complex Queries Into Multiple Tasks

The AI will be more prone to mistakes if the query is complex, as such, it is best to do follow-up messages asking about the details of a strategy rather than asking the AI to give those alongside other information. For example:

Return a good strategy on BTCUSD, alongside the average net profit of all strategies on BTCUSD that use Confirmation signals, then tell me what the highest winrate strategy is, using the same conditions as the first one

The AI Backtesting Assistant might provide more relevant answers if we let context accumulate by splitting and re-ordering our initial query into multiple ones: