Please note that there are many factors that can affect strategy calculation results and even the tiniest of discrepancies may lead to a big difference in the outputs. See some examples of this below:
Chart reloading/reopening. Historical data may differ slightly from the data that is collected in real-time. Even if they match, in some cases, historical data is altered by data vendors. Different data = different backtesting results.
Intrabar price movement. When a strategy is calculated in real-time, we have a clear understanding of how the current price was moving between OHLC. When a strategy is applied to a chart, we don't have this information, thus we use assumptions. These assumptions might be wrong and they are definitely not 100% accurate even if the price was indeed moving exactly this way.
Specific functions in the source code. Some functions behave differently in real-time vs. historical calculation due to their internal logic. An example of such a function would be 'security'. This peculiarity is called Repainting, you can find out more about it in our user manual.
Strategy calculation frequency. During historical backtesting, a strategy is always calculated only on close of every candle, while when the data updates in real-time, the same strategy might be set to be calculated with every price update.
If you are sure that none of the above-mentioned issues relate to your case, please send a report with the Pine Script issue type you are experiencing attach a strategy code and describe your problem in detail.