ASK THE BACKTESTER
Describe a trading rule. See how it did on old prices.
Type a rule in plain words. An AI agent builds it from basic steps, runs it on real past prices, and shows every trade next to SPY. It is a test, not advice.
Each question can take up to a minute. Results use daily prices and include fees you set. Past results do not predict future returns.
FEATURED BACKTEST · CALENDAR
QQQ seasonal strategy
Hold QQQ from November to May, then stay in cash. Replay its historical signals and fills.
From 2016-11-01 to 2026-10-30, this rule made 10 trades and returned +130.9%. Buying and holding SPY returned +258.1% over the same dates. The largest drop from a past high was 28.6%.
Loading trades...
HOW IT WORKS
Every strategy runs the same five steps
There is no special code for any one strategy. Each rule is built from the same small set of steps, so every result can be checked the same way.
- 1
Get data
Daily prices for each ticker, each with the date it became known. A test only sees prices known on that day.
- 2
Build series
New lines from old prices, using basic steps: average, high, low, change, lag, ratio. RSI and MACD are just these steps combined.
- 3
Build signals
Yes or no tests over time, like "price is above its average" or "one line crosses another".
- 4
Simulate fills
Trades in time order, with the fees and slippage the run sets. A stop on a day runs before a new buy on that day.
- 5
Count the money
Cash, holdings, return, and drawdown on each day, next to buying and holding SPY.
STRATEGY LIBRARY
81 saved backtests, by type
Each one is a saved run on real past prices. Open one to replay its trades, or ask the chat to change it.
USE IT IN YOUR AI AGENT
The same tools work in Claude, ChatGPT, and other agents
Backtester is an MCP server. MCP is a standard way for an AI agent to call outside tools. Add the address below, then ask your agent to test a rule. It gets the trades, the reasons, and the chart.
SERVER ADDRESS
https://backtester.daenadigital.com/mcp CLAUDE CODE
claude mcp add --transport http backtester https://backtester.daenadigital.com/mcp CLAUDE OR CHATGPT
Add a custom connector or app, and paste the server address.
WHAT THE DATA SUPPORTS
Honest limits
A test only runs on data that exists. These are the same limits every AI agent reads before it runs a test.
- Prices are raw provider prices scaled at each stock split, so a split does not change the price level. Prices and share counts are in share units at the first session of the run. The report lists each split. Cash dividends are not included. Numbers you supply, such as price levels, dollar distances, option strikes and premiums, and per-share observations, are not scaled. Set them in run-start share units.
- Intraday runs use the daily feed as the exchange calendar. Intraday bars on a day with no daily bar, such as a market holiday, are left out and listed by ticker in excludedSessions.
- Equity is checked at each close. If it is zero or less, every holding in that book is closed at that close with reason liquidation, and the book makes no new trades.
- Runs need prices for each ticker on the same session dates.
- Options use an explicit premium and modeled expiry payoff. There is no live option chain.
- Fundamental, earnings, and calendar event values can be supplied as observations with their release times. The price feeds do not provide these values.
QUESTIONS
Frequently asked questions
What is a backtest?
A backtest runs a trading rule on old prices to see what would have happened. It shows each trade, why it happened, and the result next to SPY. It does not place trades.
Is this investment advice?
No. A backtest is a test on old prices only. Past results do not predict future returns.
Can a test use future prices by mistake?
No. Every value carries the date it became known, and a rule only sees values known on or before its decision date. Adding new prices never changes earlier signals or trades.
Where do the prices come from?
Daily prices come from a market data provider. They are adjusted for stock splits. Cash dividends are not included, so results for both the rule and SPY leave them out.
Why is every result shown against SPY?
SPY is a fund that tracks 500 large US stocks. Buying and holding it is the simplest choice an investor has, so each rule is shown next to it over the same dates.
Which rules can it test?
Any rule made from the basic steps: averages, highs and lows, changes, ranks, ratios, crosses, calendar dates, stops, and targets. That covers trend, mean reversion, breakouts, rotation, pairs, and calendar rules. If a rule needs data that does not exist, it says so.
Do fees count?
Each run sets its own fees and slippage, and every fill pays what the run sets, except when a modeled option expires. Slippage means each fill is at a slightly worse price than the market price that day. Each strategy page shows the fees its run used. Some saved examples use zero. The chat is told to pick 5 basis points of each unless you ask for something else. A basis point is one hundredth of one percent.
Can I use it from my own AI agent?
Yes. The same tools the chat uses are open to any AI agent that supports MCP, such as Claude or ChatGPT. Add the server address shown above.
Risk warning
Past results do not predict future returns. Not investment advice.