Strategy performance dies overnight - EMA implicated?

[Bushman]

I have been working on 2 complimentary forex strategies, and had reached a point where performance for each system was encouraging as follows:

1) System 1 - 140 trades in 120 days at 85% win rate, net 18 pips per trade.

2) System 2 - 160 trades in 120 days at 75% win rate, net 11 pips per trades.

System 2 was generally active when system 1 was out of the market, making them a nice team when worked together. I had started to trade them and the performance was holding at stated levels.

Both these systems turned sour overnight last week. Thanks to Microsoft, my system did an automatic security update in the middle of the night (a "feature" I have now turned off). Upon restarting my PC and eSignal, the performance for both systems was obliterated to the extent that they had both become net losing strategies.

While developing systems, it has become my habit to regularly reboot my PC and/or restart eSignal, as I have previously observed variation in system performance across such events. Both of these systems have remained solid in their performance across restarts, with only minor variations observed in results.

Neither system was particularly sensitive to different settings in the various parameters that are used. (ie. you could make meaningful adjustments to parameters without destroying performance).

The only “oddness” I have been able to detect while trying to determine the cause of this deterioration relates to EMA. . My systems are sensitive to the degree of change in the EMA from one bar to the next. The values returned by the builtinEMA formula that comes standard with eSignal vary greatly with the values returned by the EMA study within EFS, even though both use 36 periods to calculate their results. The EMA study within EFS uses a type of “EXPONENTIAL”.

e.g.
builtinEMA returns values of:

1.29625, 1.29623

for consecutive bars, while the EMA study within EFS returns values of:

1.29657, 1.29634

for the same two bars. As you can see there is a large difference in both the values and variation in values returned by the two EMA result sets.

Is this expected behaviour?
Has anyone else experienced similar catastrophe with systems they have developed?
Anyone got any ideas?
HELP!

Phil.

PS. Yes, I have restored the system to it's pre MS update state. No joy

[ACM]

Phil
You may need to post the ema studies you are referring to. I just ran a comparison between the builtinMA.efs (set to Exponential), the EMA in Basic Studies and the ema.efs that is in the Library folder and they are all returning exactly the same values (both historical and in real time)
Alex

[Bushman]

Alex,

Thanks for your prompt reply. I am currently at work in my "day" job. When I get home this evening, I plan to install eSignal on another computer and backtest the strategy on it.

Notwithstanding the result of this test, I will post the code that produces the strange EMA results. I am very interested to understand what is going on, in light of your finding exact consistency among EMA variants.

I must be doing something wrong. When I start with this assumption, I am usually right.

Phil.

[Bushman]

As usual, I was right to assume I had done something wrong. I discovered why I was getting the strange EMA results... and it was pilot error.

Sadly, this has not provided an answer to the degradation in the performance of my systems, as the EMA code was the same prior to and after the disaster. I have recovered one of them to an acceptable level, but am completely mystified as to why they have gone "off the air".

The performance is consistent when eSignal and the formulae are applied to an alternate PC.

I'd still be interested to hear if anyone else has experienced similar behaviour.

Cheers,
Phil.