Hello,
I have always wondered why the backtest results of an EA differ from its live market results. Throughout my trading experience, I have tested thousands of EAs in both backtest and live market conditions. While most EAs yield impressive results in backtests, they often result in losses when applied to the live market, whether with a demo or real account.
Even when I test with a modeling quality of 99% or higher during backtesting, the EA tends to fail in the live market. Why does this discrepancy occur?
That sounds like scam EAs to me at first. Scammers like to show backtest results that show that you can earn millions of dollars in a very short time 
The values are adjusted so that it looks very nice.
As @ProfesorPips already said, this is called overfitting. The bad scam EAs are even bad on the demo account, the better ones bring good values there too.
In the end, however, you have to run an EA on a live account to see whether it is actually any good.
Because there are always differences between backtest, demo and live. In the backtest and demo, for example, you have no slippage. Or if you have a market maker as a broker, then this also influences the execution of your EAs and some other issues.