NT failure killing me

[ctrlbrk]

zstheorist,

You might also take a look at the lookback period "min bars required" in a backtest. I've also been told that under Tools -> Options -> Data where it says "Chart lookback" that these numbers apply to strategies started in the strategies tab.

I get different results when I load a strategy on a chart and run it real-time than I do when I load it from the strategies tab (no charts).

I also get different results by changing the "min bars required" on backtesting. For instance, I can change it from "20 bars" to "40 bars", on a 233 tick backtest, and it will make substantial changes to the results over a multiple-week period, even though 20 bars is only a matter of minutes at the very beginning of the first day. Very interesting. I already try to program around limitations of any calculations in my strategy using the CurrentBar return; routine...

Anyway, good luck!

Mike

[zstheorist]

thank you

ctrlbrk, thank you for the suggestion. This suggestion has been raised before for others when problems arose. I am 100% certain it is not an issue here, but it is always a good issue to check. It is particularly relevant if you start a strategy from the Strategies window instead of a chart. In either case, I am WELL beyond that threshold. As I said in my previous message, I have explicitly identified the origin of my problem.

By the way, I have been using NT for almost 2 years now - including live paper trades at IB. I've run for cash for nearly 1 year after a long test period. In general the program serves my needs and has been mostly stable over that time period. There are just a few of these subtle issues - some of which "will be fixed in NT7" whenever that comes about.

Ultimately what makes for useful software is not the program with the most bells and whistles, but stability and predictablility. It is the predictability that I am having a problem with here. A design choice of NT creates the situation where a backtest of my strategy cannot replicate the real-time strategy - by construction. I can live with that now that I have identified the timing issue. It is that unpredictability when I swap the order of execution that is getting me here. I don't really want to get more technical, because I do not have the NT source, and would have to start guessing as to why it is failing. I've simply identified how to make it succeed - the converse, not the inverse.

(I should mention I have been programming and debugging everything from complex scientific simulation software to compilers for a few decades now, and I do try to be precise when I type my observations about software. If I do not always succeed, I am sorry.)

Thank you again for helpful suggestions. NT really is the most powerful commercial program for it's price that I've found and used. Some day I may have to write my own program (probably on a dedicated linux server on a more stable line to the CME), but I have a real job as well, and have been willing to sacrifice significant gains for the convenience so far.

[thrunner]

.......
Today I missed another $1375 as it failed to execute. This is maddening! I have been running this strategy for a year with these parameters in this manner, and now 2 days in a row it fails to execute!
........
Identical code is run on SPY/SPY and ES/SPY. Both execute based on the 2nd contract (SPY) only, and open a market order on the first contract. The SPY/SPY confirms absolutely that the 2nd SPY triggers. Again, this has worked for a year, it just no longer does.

After reading most of this thread I see no real code, no pictures, no explanation of what has changed except something about missing a lot of $ and "has worked for a year" and a lot of finger pointing at NT.

I am not sure that is the way to solve your problem and frankly I don't think anybody can learn anything from your lesson.

Has this strategy ever produced any PnL in automated live trading and what has changed?

[zstheorist]

yes

thrunner,
yes, the strategy has been quite successful "for real" since I went live in June, and perhaps will be in the future. This is my last post on the topic unless I am able to disentangle more about where the logical execution asymmetry is coming from.

I have posted enough code to completely identify the problem - no more. The rest is what the constants are, Print statements, and a symmetric piece of code that shorts. This is not a complicated strategy.

I was never asked for more information by the NT staff than I have posted. Pictures would not help except to show that indeed 1 strategy went when the other did not, but I make them every day to fix a record of what happened in case trouble arises. E.g., sometimes AT&T decides to randomly cut the network, and I can see right where it happened. Or, if a rare case shows up live that is not caught in backtesting, I can then figure out what happened (like here). All relevant information content in the pictures has been posted.

Based on what I have posted, I could track down what is happening if I had the source. However, since it is closed, I will have to wait until someone at NT actually investigates the logic flow of the program, and can understand the asymmetrical evaluation I am seeing. I have no illusions this is easy to do - which is why I paid them for the program instead of writing one myself.

I am being pragmatic here: things are working at the moment, and that is all that matters to my bottom line. It just means I have to start things up in a particular order to get them to run as expected. It also means that if I decide to start a new strategy, I have to be careful where to place it in the sequence of the 4-5 I run each day.