Gold Sparrow
- Uzmanlar
- Sürüm: 1.80
- Güncellendi: 2 Eylül 2026
Required reading: Live trading may involve significant drawdowns. Always trade with money you can afford to lose, and avoid setting the LotFactor too high (recommended 0.1–0.2).
By setting the parameter "UseRecentModel" to true, this EA can automatically download newly trained parameter values based on recent market data.
■Timeframe used 1H
■Minimum account balance --- Recommended 650-750 USD per 0.01 lot
■Gold Sparrow is a trading robot trained on 15 years of historical XAUUSD data. Please use it on the 1-hour timeframe. By incorporating multiple channel-based indicators, the EA is designed to better understand the current price position within market waves.
■Since Gold Sparrow employs a strategy of buying dips and selling rallies in line with the trend, it performs strongly when the trend persists but incurs losses if the trend suddenly reverses.
To help prevent drawdowns caused by sudden trend reversals, it is recommended to set the MaxHoldingHours parameter within the range of 5 to 8.
Recommended broker: A broker with GMT+3 in summer and GMT+2 in winter, such as IC markets.
(Exness is not recommended, but if you use Exness, please adjust the parameters as follows: TakeProfit=300000, StopLoss=30000, MaxSpread=600, Slippage=300, and WeekEndCloseTime=18)
■How to install the new model --- In MT5, go to Tools → Options → Expert Advisors and add
https://greentrailns-stack.github.io/MetaTrader/gold_sparrow1.1_model.json
as an allowed URL. After attaching the EA to a chart, a file named gold_sparrow1.1_model.json will be downloaded into the Files folder inside the Common directory.
When UseRecentModel is set to true, the EA will trade using the newly downloaded parameters (backtesting is not possible unless the parameters have been downloaded). While Gold Sparrow is running, it will automatically update the JSON file between 1:00–2:00 MT5 time on weekdays if a new version is available.
■This EA requires 500 H1 bars. In MT5, go to “Tools” → “Options” → “Charts” and set “Max bars in chart” to 500 or more. Then temporarily disable “Auto Scroll” and drag the chart to load and display at least 500 past candles. Thank you for your cooperation.
Parameter descriptions
■TakeProfit, StopLoss --- Values are in points.
■UseBuyLimitOrSellLimitLoss --- Places a pending order OffsetPoint points away from the current price.
■WPR_Filter --- Uses WPR as an entry filter. The EA will not enter trades when WPR is above or below the values set in WPR_Buy and WPR_Sell.
■OrderIntervalMinutes --- Prevents the EA from placing a new order until the specified number of minutes has passed since the previous order.
■OrderDelay (MilliSecond) --- Delays order execution by the specified number of milliseconds after a signal is generated.
■This EA is programmed to close a position if it is in profit after HoldingHours. If HoldingHours has passed and the position is still negative, it will close the moment the profit reaches (lot size × MinProfit). For example, with a 0.01‑lot order and MinProfit set to 100, the EA will close at 100 yen for JPY accounts or 100 cents for USD accounts. For micro accounts, please adjust MinProfit accordingly. If MaxHoldingHours has passed, the EA will close the position even if it is at a loss.
If the latest version does not work on MT5, please try v1.4. https://www.mql5.com/ja/market/product_version/371062/download
Note: Backtest results do not guarantee future profits. Please run the EA on a demo account for a while before using it live.

Hello, I updated version 1.7 today, but during my backtesting, I found that when UseRecentModel was set to true, the backtesting performance was much lower than when it was set to false. Is the new parameter actually more unsuccessful? Looking forward to your explanation to clarify my confusion, thank you!
Edit:Thank you for your reply! So, in the long run, do you recommend setting it to 'true'? Or rather, set it to 'false' normally, and when the model is updated, set it to 'true' to download the latest model, and then change it back to 'false'. As you said, choose the high-performance one from multiple candidates (including the latest model just downloaded). But this seems a bit troublesome haha.