Gekko backtest results change if you adjust time.

[mazing7]

This does not seem normal, if I change the data set's time by just 3 hours, the trades being executed 3 weeks later are different than the original backtest. This was a backtest using 3 min candles and my own strategy using a tulip indicator. I've created several strategies with amazing backtest results but when i run the strategy live its hit or miss, usually miss. Is this due to tulip indicators not working properly? 
Also when I removed all indicators and just created a strategy based off statistics/ price movement using no indicators this problem did not reappear.

[askmike]

Can you tell me more about the markets, and the exact strategy (indicators + logic)?