Bug cleaning with 95% possible with the help of the IB Demo TWS. Logic errors correction, I mean right sequence of steps including event processing from TWS, is possible only with real money. It doesnât take much money; it takes a lot of time. Why I am talking that I am not sure 100% for my ATS. Because of, for example, if I have a binary tree with only 10 levels, this event tree potentially can produce 2^10 = 1024 outcomes. To test all this variety in real trading need more real trades. I monitor my system doing my full time job which is not connected with the stock market. As I said the ATS call me to my cell phone about all transactions and system orders. If I see that something wrong, I can call home and ask my wife to correct situation, or just get access to my home computer through the remote desktop. And of course at any situation I am sure, my broker has stop order. ---------------------------
8. Comparing the automated trading system calls with the real trading results In April the ATS provided trading calls for +18.8 ES points or $940.00 per 1 ES contract (see attachment). Real ATS trading made +20.5 points or $1025.00 per 1 ES contract. Difference is due to the discovered error in the coding; it is a very rare case when positive. ---------------------
In May the ATS provided trading calls for +4.45 ES points or $222.50 per 1 ES contract (see attachment). Real ATS trading made +4.70 points or $235.00 per 1 ES contract. Difference is due to the IB two times has executed stop orders better than expected. ---------------------------------
I remember once that CME sent out bad prices for 15-20 minutes long time ago. So remember to add an alert / feedback so that the system will filter out those prices, and you will be notified if abnormal prices are flooding through.
Lawrence, thank you for the advice, I really appreciate it. But to add an alert for bad prices need to set up some criteria. You said a; could you say b? --------------------------
There are many ways to determine if the prices are correct. The most common way is to buffer the last N trades together with all bid and ask data, then test if the newly arrived tick make any sense at all comparing to whatever is stored in the buffer. This method works 90% of the time, except stocks that are having trading halts with news announcements. Then the reopen is pretty much unpredictable. Since you are not dealing with stocks, or stock options, this method should work fine.