Fixes
I think one of our largest challenges to consistent profits is the vast difference between the profits from the same broker started at slightly different times. The 1 trade per currency is likely a decent EA decision for stability and performance, but it makes examining test results very hard. 5 people will get completely different results. 1 person will have amazing profit, 2 will have great profit, 1 will break even and 2 will loose a lot of money.

In order to get a deeper understanding of market signals and effectiveness, we should probably trade on every valid signal. This would probably not be a system to use live, but at this point in time I don't know how to explain how strange the results are for different people aside from "you have different entry times."

I think this is one of the "deeper questions" we will face with this EA. I am going to run a test where max trades = 100 and lot size = smallest possible with money management off and see if the EA explodes, makes billions, or collapses. Hendrick restricted the trades for a reason, but I don't know what that reason is or how often we get trade signals, how accurate each signal is, etc. We are looking at 3 or 5 puzzle pieces, when the picture was made of thousands. We can't see the whole picture yet, and I think we need to. This would also help tell us how effective the consecutive signals fundamentally is. Dagnar had great success in backtest but there is more involved than that.

I do think the consecutive signals is a great idea - but one of my "bug fix" goals is to make Phoenix more reliable between users. I don't like one account going broke while another makes amazing returns on the same data. I consider that a type of bug.

The primary goal for me right now is to merge in bug fixes. The main one is Mode 3 working properly but we may need settings to go with it. I will review and release the code, probably with consecutive signals. An alert for currencies without settings may be good. We will very quickly need to address how Phoenix interprets settings files, which versions, currencies, timeframes, etc. Maybe next week?

Summary for intended release: make max trades a parameter that can be edited. Incorporate Mode 3 changes. Default to Mode 1(?) since it is longest tested. Consecutive trade signals as a feature that can be turned off/on.

Major changes to consider for development
Intelligently processing trades that miss their countertrend peak. I need to examine a lot of data, but it appears from the "extended trades not profitable" discussion that missing the target leaves a trade hanging for a long period of time. Can we detect when that miss occurs? For example: if we receive an incoming sell signal on a buy trade is that an indication we should prepare for exit with a trailing stop? I consider this "tuning" for another version, but I wanted to write it out.

The AccountEquity vs AccountMargin discussion didn't solve the real problem we face with money management: limiting overall risk without crippling the ability to make money when we are doing well. Money management in phoenix should ideally be reactive to two main situations: 1) hitting stoploss without countertrend occuring 2) consecutive losses. I know that wasn't very clear and I will try to think this through more. One concept is to reduce lot size until Win:Loss for the past time period X increases to a certain ratio.

Another idea is to have lot sizes placed based on (Equity - current exposure )> RiskPercent*Equity as a scaling logrithmic curve. (larger trades when exposure low, smaller trades when we have a lot of loosing trades open with big SL values). This idea has a fundamental appeal to me, but I don't know how effective it is. I did the math on how "max risk" is interpreted and what valid values would be. The Risk parameter is not human readable, and doesn't react well.

Separating Buy/Sell code seems valid, interesting, and a direction to explore. I don't now how two 80% accurate trading systems can combine to a 95% accurate system though. The math from http://en.wikipedia.org/wiki/Bayes'_theorem would be much different but I'm too tired at the moment and I lack some information directly on hand. Essentially by determining the probability distribution is significantly against a specific condition, the negative evidence stacks strongly. On the other hand, a Buy only EA may pay good swap.