An idea from a chat or a book looks convincing until you test it on history with commission and slippage.
Fill in the template with your rules, and the agent computes the result after costs and lists all of its assumptions separately.
Test a trading idea on history in ATAS.
Instrument and period: my BTCUSDT 5m chart, all loaded history.
Setup: the bar closes above its open, its volume is more than twice the 20-bar average, and delta is negative. US session only, 9:30-16:00 New York time.
Entry: long at the open of the next bar.
Stop: beyond the low of the signal bar.
Target: 1R.
Time exit: 16:00 New York time.
Costs: commission 0.05% of the trade value per side and slippage of 0.01% per side.
Use code. Don't add filters on your own, and list everything you assumed. Show the number of trades, win rate, expectancy in R after costs and drawdown. If there are fewer than 30 trades, say that no conclusions can be drawn. Don't change my charts and don't open any trades.
Put in your own costs.
What happens
The agent computes directly from the bars and writes out its assumptions. The idea in this example gave about −1 R per trade after costs over 40 trades: 17 trades reached the target, and 11 stayed positive after commission and slippage. Stops on BTC five-minute bars are short, and costs charged as a percentage of the trade value eat a noticeable part of each R. Look at the list of assumptions: it has everything the agent decided on its own, for example a stop 0.1 USDT below the low and the handling of the switch to standard time.
Test a trading idea on history in ATAS.
Instrument and period: my BTCUSDT 5m chart, all loaded history. Setup: the bar closes above its open, its volume is more than twice the 20-bar average, and delta is negative. US session only, 9:30-16:00 New York time. Entry: long at the open of the next bar. Stop: beyond the low of the signal bar. Target: 1R. Time exit: 16:00 New York time. Costs: commission 0.05% of the trade value per side and slippage of 0.01% per side.
Use code. Don't add filters on your own, and list everything you assumed. Show the number of trades, win rate, expectancy in R after costs and drawdown. If there are fewer than 30 trades, say that no conclusions can be drawn. Don't change my charts and don't open any trades.
Computed with code on the history of your BTCUSDT Binance Futures, 5m: Sep 4 to Nov 5, 2026, 18,108 closed bars, nothing missing. The forming bar is excluded.
| Metric | After costs |
|---|---|
| Number of trades | 40 |
| Win rate, trades with a positive result | 27.5% |
| Expectancy | −1.013 R per trade |
| Max drawdown on closed trades | 40.536 R |
| Total result | −40.536 R |
17 trades reached the target, but only 11 stayed profitable after costs. Under these assumptions the idea showed a negative result.
Calculation assumptions:
- Average volume is the simple average of the previous 20 bars, including bars outside the session; signal: volume strictly greater than twice the average.
- Session: daily 09:30-16:00
America/New_York, with the switch to standard time. Weekends and holidays were not excluded. - The signal bar starts no earlier than 09:30; entry at the next open strictly before 16:00.
- "Beyond the low" means 0.1 USDT below. This is an assumed tick size; I couldn't confirm it through the API.
- 1R = entry bar open − stop, before costs; the target is the same distance above the entry.
- Each signal is a separate trade; overlaps allowed, the risk of each trade is 1R.
- If the stop and the target are both reached within one bar, the stop counts first. There is one such case.
- On an open below the stop, the exit is at the open; on an open above the target, at the target. The 16:00 exit is at the close of the 15:55 bar.
- Slippage worsens the entry and the exit by 0.01% each; commission is 0.05% of the actual value of each side.
- Drawdown is computed on the realized result in order of exits, without intra-trade mark-to-market. A percentage drawdown needs the capital size and the risk.
I didn't change your charts and didn't open any trades.
Calculation code · All trades · Results
What you need
- A crypto exchange connection
- An agent that can run code
- Your real costs. Without them, the result looks better than it is.
This is a historical test. Past results do not guarantee future results, and the trading decision is yours.