Live Testing reports - page 39
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Phoenix 5: "optimize for market" Phoenix 6: "optimize forever"
Huuuuuuuuuuuuuu!!!!!!!!!!!!!!!!!
The news of the day !!!
Darak : you are the king of Phoenix!
Did you miss me Bertbin? I've gone looking but haven't really seen any new optimizations. I'm trying to wrap up this silly EA still. Funkiest channel breakout I've ever heard of...
Did you miss me Bertbin? I've gone looking but haven't really seen any new optimizations. I'm trying to wrap up this silly EA still. Funkiest channel breakout I've ever heard of...
Yes... I found something about trend on Forex factory last week.... but I lost it !!!!!! I will seach again.. How about you ???
Backtesting 6 pairs
Glad to hear that Daraknor is feeling better and back in fighting form. Keep up the good work.
Attached are reports for demo accounts with Alpari and Interbank, one with Default settings in Phoenix 572, the other "optimized" over the last 6 months in 2006. Neither has performed satisfactorily. Either I am doing something wrong or the settings that I find via optimization do not hold after only a few weeks. Even the default settings don't show much promise thus far.
Using maxtrades (5) and high ConsecSignals (15-20) in order to push activity.
Any comments or suggestions. Incidentally the two GBPUSD trades did not yield much at all.
I've been doing a lot of testing lately, and was wondering if there was a way to calculate the effects of the other pairs on one another. Since we can't backtest multiple charts/pairs at a time; is there some kind of equation that would be able to give some kind of result like that?
Backtesting all 6 pairs(w/ defaults, but best performing modes) accumulates $98,798 for January from a $10,000 deposit. I want to know what the effects of loss/gain on the combined pairs would be. Would the account accumulate less gain because of combined drawdown, or would it double/triple/quadruple gains because of the compounded gains throughout the month & money management?
For the optimized settings: did you notice how GBPUSD and GBPJPY both continued to generate the same signal, even through the signal was bad? Could you look at the history and find out if the currency reversed soon after that?
If the signal itself was bad (many consecutive bad trades in the wrong direction) I think the settings discovered by the testing were 'not good' for the current market. Did you do the optimization on the entire 6 months, or did you do it on 3 months with 1 month after and 2 month before used for comparison?
A fundamental belief of mine, relevant to ALL EA, is that we need to get solid performance on data we didn't optimize against. Some of the settings will be good, a lot of them will be awful. When I went through and described this in the optimization thread, I noticed that they came in clumps. Have you tried some of the clumps that I selected?
Since we can't backtest multiple charts/pairs at a time; is there some kind of equation that would be able to give some kind of result like that?
I'm not aware of a general formula, but I'll see what I can do to describe it.
#1 Optimized Backtesting will not telll you a thing about performance. I need to say that up front. I recommend doing optimized backtesting for a few months of data, and then a few weeks or a month of the most recent market data as a test to see if you found a good setting to be used outside the backtested period.
#2 If the money management is based on balance, all effects immediately compound. If the money management is based on FreeMargin, your second trade is slightly smaller than the first, same with third, fourth, fifth trades if they all start within a few minutes of each other. If the trades are winning well, then you see the compounding effect as the trades close. Generally you get the same compounding effect as balance based calculations, but it is dampened slightly.
Balance based trading can literally kill the account when you have multiple currencies. If you have 10% of a the balance committed to a trade, and your SL is configured to actually loose 10% of the balance (SL matches margin requirements) then having 10 trades open means you have $0 if they all loose. The same settings on a free margin system would loose 10%, 9%, 8.1%, etc... determined by (balance after loss in percent)^(number of trades) = 35% in this case. (0.9)^10 We have a FreeMargin based calculation, but we don't include SL in the formula for Phoenix5. Phoenix6 will have a Real % at risk.
GBUSD performance
For the optimized settings: did you notice how GBPUSD and GBPJPY both continued to generate the same signal, even through the signal was bad? Could you look at the history and find out if the currency reversed soon after that?
If the signal itself was bad (many consecutive bad trades in the wrong direction) I think the settings discovered by the testing were 'not good' for the current market. Did you do the optimization on the entire 6 months, or did you do it on 3 months with 1 month after and 2 month before used for comparison?
A fundamental belief of mine, relevant to ALL EA, is that we need to get solid performance on data we didn't optimize against. Some of the settings will be good, a lot of them will be awful. When I went through and described this in the optimization thread, I noticed that they came in clumps. Have you tried some of the clumps that I selected?Thanks for your reply Daraknor. I attach the charts for GBPUSD and GBPJPY running 5.7.2. I should point out that the GBPJPY and EURJPY settings were essentially the same as the original, as only the TS, SL, and TP were adjusted. Signals are all default for these.
I hope that you will find the answer to your question here. I attach the GBPUSD with original settings for comparison. You will see that the two are essentially the same. I would expect the EA to enter sells when the pair shows an unequivocal movement downward, and buys when the movement is upward. Phoenix appears not to behave in this way at all, and appears to be missing many opportunities to harvest pips.
The optimization was a straight 6 months, and the forward testing was intended to confirm the results of optimization. Neither optimized nor default settings have shown positive performance in the past two weeks. I have not particularly been looking for clumps, as I optimization is very time consuming and I did not have time to go back and review all the data.
I look forward to your comments and suggestions.
The philosophy of Phoenix is to Buy when the price goes down, and sell when the price goes up. CounterTrending. The key thing to note is that the the price curves up and down, and we are looking for a certain angle in the curve, approximated by a change in price and a change in momentum. In other words, the first and second derivatives need to decline drastically.
In the first chart, GBPUSD, you notice that there is a cluster of buy trades immediately after a short dip. This is good but not great (I'd need to see previous data) (5win)
It is followed by 3 trades. The first one is very bad (steep slope). The second and third sell are good.(3loss)
Next buy signal is good.(1win)
Next group of 2 buy signal are good. (2 win)
Next 4 sell signals split up. 1&2 I can't tell. 3&4 sell are good. (4win)
Next 4 are buy buy sell sell. I can't tell. (2win 2 loss)
Next is a sell, right before the gap. This is a good sell signal, but the gap destroyed it.(1loss)
Next 4 sells, 2 I'm not sure on, second 2 are great. (4win)
Next 6 buy signals. 1 I wouldn't do, 2&3 no. 4&5 definitely yes. 6 not sure (6loss)
Next 3 sells. 1&2 yes, 3 not sure. (3 win)
Next 1 buy. I don't know(1win)
Last 4 sells. First one I would do, 2 don't know, 3&4 don't know (2 win 2??)
It appears this currency isn't faring well with the slope detection, meaning the signal itself needs to be examined. GBP is a weird currency, USDJPY has more smooth curves and stable by the numbers.
I just checked my track record. 21 4 9 I think, with 21 predictions on yes/no being accurate, 4 predictions inaccurate, and 9 I don't know. (if i was just controlling yes/no on authorizing trades)
If you can see what I see in the chart maybe we're on to something.
Looking at the second and third chart, I don't know what to say at all. I expect to see little or medium sized humps on the chart for prices, but looking at the volume, I don't think the signals are playing in the same ballpark as the TP/SL levels. The signals might be playing short term, and the TP/SL could be set to long term trading. This is one of those situations where profit was maximized, and signal accuracy may have just come from curve fitting news events. I can't really tell at a glance, but it looks like the curves are simply on a different scale than the TP/SL. I'm pretty sure this situation would have been caught by trying the same settings on nonoptimized data, but I don't know.