Creating own trading system

Do you use database? What do you use if you have one in your system?

 

I’m not sure why you’d want to build a whole system from the ground up? For systems development there are a lot of pre-made programs that have a done of features (backtesting, walk-forward optimization, etc) built in. If this is for curiosity, then I understand, but if you’re looking to make a fully functional system, it seems like you’re reinventing the wheel a bit.

 

For some simple stuff, more "default" kind yes, but not with something totally different from "default".

 

Another point is can you believe the retail level trading system, no matter backtest or real trade. Do the same backtest in different popular retail trading system, you will get different answer, which one should you believe?

 

Can you give an example of this? I tend to only use one backtesting platform so I’m curious if you’ve ever tried importing the same data into two different platforms, running the same code, and getting two different results. Doesn’t seem possible.

 

This is a good post but seem like amost no one seriously talk about the speed different of these two setup.

 

All backtesting platforms make assumptions , for example the bouncing tick issue. Which is assuming which happened first the high or low of a bar when shorter timeframe data is not available. This is a big issue on daily bar systems.

There also other issues, for example on limit orders if we assume you have to trade 1 tick though the limit or just touch it. Believe it or not in the S&P500 there is a big effect in the results, as much as 30% between these two cases. In TradersStudio I made assumptions for worst case in most places, so that the back tested results are more realistic. In TradersStudio you can require you trade though the limit price to get filled.
In addition often times trades in a system happen or are missed by a small amount,internal rounding can greatly effect results and different platforms handle it differently.

The key is that the strategy needs to be robust, if these difference make the system trade-able or not trade-able then you have a problem.

 

That clarification helps on how the backtesting platforms work. So in essence I guess you’re suggesting that TradersStudio typically assumes the worst case scenario and as such if I code a system that works well in it, it shouldn’t get tripped up on, say, the bouncing tick issue?

 

Another thing is slippage - that is tricky. Whatever you assume, sometimes it sucks (I got 19 ticks in CL once in real - rare, but that may change your logic. If it happens on a stop it is just a loss, if it is on a market entry.... your whole stop logic may change - if it is based on real entry price).

For that you need not bars but full oder book and possibly a size modifier (assume an order 5x as large as mine, get me the worst filles) - something that is more dynamic as "assume 1 tick". Plus as I said - full replay.

 

Backtesting is one of the most misunderstood and dangerous practices that is the main drive behind the massive failures of newcomers to trading especially those trading futures who do not understand how to properly adjust contracts. I know of no trading platform that properly assists traders in adjusting data the correct way so that backtesting results are realistic. If one exists please let me know.