Why is an EA profitable in backtests but blows the account in a real account?
Hello,
There are several EAs in the market claiming profitable results by showing backtest data. I have tried tons of free and paid EAs with amazing backtest results. With backtest results, EAs show amazing returns and very low drawdown—even several years of backtest data show positive results.
But when the same EA is applied to a real account with the same settings, the result is totally different—it blows the account in a day, week, or at most a month. Are backtest data not reliable or not legitimate?
Its the overfit issue.
Whether the ea creator is aware of it or not.
Whether they developed the strategy using neural nets , python , strategy tester , genetic algorithms , paper trading , manual development over years , systems are vulnerable to overfits.
To be able to detect overfits you would need a closed environment where development->testing->publishing occurs on the company's servers something both expensive and not feasible from a rights to the code perspective.
For instance , if i am not aware of the overfit problem , i test a strategy over 12 years in the tester , it works , i really believe i found something valuable and i publish it.
The least likely to overfit is a person who is really really bad at memorizing stuff who decides to develop a strategyMany reasons. Overfitting is a huge one, as EA creators want to show you the best possible outcome. So, they set the back test to do something like "find the optimal SL between 10 and 200 ticks". Yeah, the results will look perfect, because that exact setting worked for that exact scenario. But it's not going to work in the future as it's too exact.
They also don't use tick data. Many people, from EA builders to users, don't understand that "every tick" in MetaTrader is actually made-up ticks. You have to set it to "every tick based on real ticks" if you want to use actual tick data. That setting alone can change the results very drastically as one of them is made up data, the other is actual market data.
Now that's not to say EA's can't be profitable, you just need to find a reputable seller that actually knows what they're doing. I've used a few from AlgoDatabase successfully but I find the seller is likely an actual professional, not a guy churning out EA's for a quick buck. Anyone can make any EA look profitable in a back test, as there is always going to be some combination of things that would have worked. Find the seller who offers back tests based on tick data (you can tell by the results page, as it says this) as if it's not using actual market tick data then it's useless. You can't back test on made up data and then expect it to work with real markets.
This industry as a whole is full of a lot of grifters who want to make a quick buck, most EA sellers are going to fall under that category. It's sad, but it might not even always be intentional. Some people just don't possess the knowledge, like the tick setting above. Probably 9 out of 10 people I talk to don't know it's using made up ticks from OHLC.
Many reasons. Overfitting is a huge one, as EA creators want to show you the best possible outcome. So, they set the back test to do something like "find the optimal SL between 10 and 200 ticks". Yeah, the results will look perfect, because that exact setting worked for that exact scenario. But it's not going to work in the future as it's too exact.
They also don't use tick data. Many people, from EA builders to users, don't understand that "every tick" in MetaTrader is actually made-up ticks. You have to set it to "every tick based on real ticks" if you want to use actual tick data. That setting alone can change the results very drastically as one of them is made up data, the other is actual market data.
Now that's not to say EA's can't be profitable, you just need to find a reputable seller that actually knows what they're doing. I've used a few from AlgoDatabase successfully but I find the seller is likely an actual professional, not a guy churning out EA's for a quick buck. Anyone can make any EA look profitable in a back test, as there is always going to be some combination of things that would have worked. Find the seller who offers back tests based on tick data (you can tell by the results page, as it says this) as if it's not using actual market tick data then it's useless. You can't back test on made up data and then expect it to work with real markets.
This industry as a whole is full of a lot of grifters who want to make a quick buck, most EA sellers are going to fall under that category. It's sad, but it might not even always be intentional. Some people just don't possess the knowledge, like the tick setting above. Probably 9 out of 10 people I talk to don't know it's using made up ticks from OHLC.
Indeed, there are other reasons as well. It's crucial to thoroughly investigate key metrics such as Profit Factor and the Risk/Reward (RR) ratio. For instance, an EA with an RR below 1 might seem attractive due to a higher win rate, but just a few consecutive stop-losses can significantly impact your account.
Another common issue is unrealistic expectations from users, believing an EA should quickly make them wealthy or generate substantial profits in a short period. However, a genuinely effective EA should prioritize consistent, realistic performance over the long term.
I don't blame sellers who honestly create technically sound EAs. It is ultimately the buyer's responsibility to exercise caution and carefully evaluate settings and real-world performance before purchasing.
Unfortunately, buyers often feel pressured to purchase quickly due to marketing tactics like "Only X copies left before the price increases!" or because a signal has already "20 weeks of profitable results." In reality, it's easy for less scrupulous sellers to launch multiple accounts with varying parameters for the same EA, then showcase only the one that performed positively due purely to luck.
Nevertheless, we can't entirely blame sellers who effectively market their products, provided that results aren't falsified. It's up to the buyer to remain vigilant and informed.
Hello,
There are several EAs in the market claiming profitable results by showing backtest data. I have tried tons of free and paid EAs with amazing backtest results. With backtest results, EAs show amazing returns and very low drawdown—even several years of backtest data show positive results.
But when the same EA is applied to a real account with the same settings, the result is totally different—it blows the account in a day, week, or at most a month. Are backtest data not reliable or not legitimate?