1. If you have more than a few weeks of very small bars (2 minutes or less), the process becomes slower. If you have already built the model, you won’t need a lot of historical to go forward trading. Start the chart at a later date (just a little older than the longest lookback in your indicators). Also, if you zoom in on only the most recent data while trading the system runs faster.
2. If you have a really intensive model (Trading strategy or Prediction) or quite a few of them, things can slow down. Sometimes people leave models in a chart they were experimenting with earlier, perhaps hidden. Remove all models and indicators you don’t actually need to trade.
If your data feed is monitoring a number of issues you aren’t trading, stick to the ones you are trading, especially if you are using small bars. Remove the others.
3. Range, tick, second, and volume bars are REALLY slow because the system is gathering each tick and computing from the whole stream. If you are using these bars it is especially important to evaluate the things above.
4. Since NeuroShell keeps all charts and data in RAM, you should have at least 3 GIGs of RAM if you have multiple charts, or many issues in the same chart.
5. If you are using bars more frequent than minute bars, you need to make sure your models calculate in much less time that it takes a new bar for complete. Some really large, complex models can take more than 60 seconds to complete, especially on a slow machine.
6. If you have built up some really involved and complex indicators with the Indicator Wizard, it will probably be much more efficient to code them in PowerBasic, IBasic, or C, since the Indicator Wizard is interpretative and doesn’t compile.
7. Make sure you aren’t running other programs when you trade. If you have a dual core machine, you can run another program as long as it runs on the other CPU and doesn’t share your intraday datafeed.
8. Fewer subcharts will graph faster. Hide subcharts (subgraphs) that you don’t really need to see.
9. Straight lines graph faster than more complex o/h/l/c graphics