Old Dog with New Tricks | Page 10 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Old Dog's Taming of the Beast? 34 replies

Old Dog Fib Trading - EA Development 18 replies

Old Dog Fib Trading - EURUSD 399 replies

5 Tricks of Trend Trading, by Rob Booker 25 replies

How To Work Around A Market Maker's Tricks 0 replies

Attachments: Old Dog with New Tricks
Exit Attachments

Old Dog with New Tricks

  • #181
  • Jan 17, 2010 8:24am Jan 17, 2010 8:24am
Quoting Mindsphere
I had this running all week on some settings which actually seem quite similar to what I can ascertain yours are, however it didn't take any trades. I think that's because I had WinPercent & MinimumSuccess set to something.

Just trying to extrapolate your settings.. the SL & TP are simple enough from the trade list. The settings which are hard to figure out are the BarStart and Initialrange - because the numbers of trades produced in the optimizer are dependent on each of these. i.e. There are a very large number of permutations which would provide...
Yes. There will be slight differences between brokers regarding the stats generated, but you're very close.
  • #182
  • Jan 17, 2010 8:36am Jan 17, 2010 8:36am
Quoting Ronald Raygun
Yes. There will be slight differences between brokers regarding the stats generated, but you're very close.
I installed FxPro's MT4 demo in order to get as close as possible.. but I can't seem to quite replicate your numbers

I've got a 54.4% counter success ratio though.. That's on the equivalent of 2 months worth of bars on M1, default initial range, and SL & TP at 20.

Is there more to this EA than simply taking whichever trade matches up to the stats?

I mean, if I have this right:

1. The EA backtests the set range of bars to calculate what would be more favourable - a breakout from the daily high/low or a countertrend trade.

2. Depending on what success or win rate you select the EA will trade as long as the win/success rate for one of the types of trades is higher.

3. This is where I'm not clear. Will the EA always take a trade when the high or low of the day is hit, as long as it's after the first 10 (default) bars? Surely this cannot be the case, as your stats show only 4 trades for the entire week. The daily high or low is constantly hit as a day's trading range is established. There is an additional filter here, correct?

Further on this point - your numbers show a 54% win rate for counter-trend trades. Yet your actual trades for this week show 80% winning trades. Was this week an anomaly? Your numbers earlier in the thread (8 wins 6 losses) are closer to what would be expected (a 57% win rate).

I guess I'm just wondering about the predictive value of the previous month (or more) worth of data for predicting future breakout/countertrend trades. If, as I would expect, a backtest over longer periods of time does bring us closer to statistical breakeven numbers - then might it not make more sense to trade the opposite of what the stats tell us? i.e. If over the past month counter-trend trades would have won 60% of the time, should there not be some sort of correction so that it may deviate back to the 50% mark, and hence we would be better off taking breakout trades?

That's a whole lot of questions - but as you said, you've been trading this profitably for over a year and a half. I'm interested to hear your view on this
  • #183
  • Jan 17, 2010 8:38am Jan 17, 2010 8:38am
Quoting Mindsphere
I installed FxPro's MT4 demo in order to get as close as possible.. but I can't seem to quite replicate your numbers

I've got a 54.4% counter success ratio though.. That's on the equivalent of 2 months worth of bars, default initial range, and SL & TP at 20.

Is there more to this EA than simply taking whichever trade matches up to the stats?

I mean, if I have this right:

1. The EA backtests the set range of bars to calculate what would be more favourable - a breakout from the daily high/low or a countertrend trade.

2. Depending on what success...
For #3, the additional filter was trade overlap. The EA was coded to handle one trade at a time. There are some instances where a particular trade ran for a very long time, on the order of a few hours. During which time other signals were triggered.
  • #184
  • Edited 9:58am Jan 17, 2010 9:43am | Edited 9:58am
Hi RR,

1. Does this EA is designed for low SL and TP? If I set SL20/TP20, my breakout percent and counter percent is about 54% and 44% respectively. However, when I increased my SL and TP to SL100/TP20. My breakout percent and counter percent will increase to 85%. The highest SL I will get a better breakout percent and counter percent. The highest breakout percent and counter percent will better?

2. If Breakout Total is shows 2434, does this means that the EA will open 2434 trade?

Thanks so much for your reply.
  • #185
  • Jan 17, 2010 10:47am Jan 17, 2010 10:47am
I think the profitability of this EA could be increased with trailing stops to allow the profits to run, and perhaps if trades were only triggered by patterns with higher probability outcomes such as trend divergence.

Have you seen any characteristic differences in the different live tick data feeds your getting now? anything like one lagging behind a considerable amount at busy times? Ive been trying to notice by watching different demo's but you will have a much better view of the differences.
The greatest force on earth is compound interest.
  • #186
  • Jan 17, 2010 11:28am Jan 17, 2010 11:28am
Quoting alankw88
Hi RR,

1. Does this EA is designed for low SL and TP? If I set SL20/TP20, my breakout percent and counter percent is about 54% and 44% respectively. However, when I increased my SL and TP to SL100/TP20. My breakout percent and counter percent will increase to 85%. The highest SL I will get a better breakout percent and counter percent. The highest breakout percent and counter percent will better?

2. If Breakout Total is shows 2434, does this means that the EA will open 2434 trade?

Thanks so much for your reply.

  1. Higher win percent doesn't always mean a better EA. Look for my thread on "Success Score"
  2. Not simultaneously, but yes. It would have opened that many trades if you placed the EA on the chart however many bars ago you used.

  • #187
  • Jan 17, 2010 11:31am Jan 17, 2010 11:31am
Quoting sebastionay
I think the profitability of this EA could be increased with trailing stops to allow the profits to run, and perhaps if trades were only triggered by patterns with higher probability outcomes such as trend divergence.

Have you seen any characteristic differences in the different live tick data feeds your getting now? anything like one lagging behind a considerable amount at busy times? Ive been trying to notice by watching different demo's but you will have a much better view of the differences.
Put the Hilo Trader Lines version of the EA on your chart. You'll notice that some trades took a long time to hit their TP or SL, during that time, a trailing stop would have ended such a trade in a loss. I realize that a trailing stop can help cut losses, but it can also cut profits. By my research, 40% of winners would have ended up as losers if I used a trailing stop. Perhaps you can show some hard numbers showing otherwise? Trend divergence is moving into the realm of human-trading. I know of no reliable way for an EA to calculate divergence.

As for tick feeds, they are all generally the same from what I saw. You'll probably see some errors in the EA. That's largely because my webhost closed my account. I ahven't seen tick streams from [TD] enabled EAs in 3 weeks. I'm trying to find a service that can host the database and handle the traffic that these EAs throw at it. So far, not much is working out. Major differences tend to come at news time, one broker may spike while another will not. General rule of thumb, a broker's quote is within 5 pips of any other broker's quote at one time--for mainstream pairs. Exotic currencies obviously have a larger difference between brokers.
  • #188
  • Edited 12:44pm Jan 17, 2010 12:20pm | Edited 12:44pm
Quoting Ronald Raygun
Put the Hilo Trader Lines version of the EA on your chart. You'll notice that some trades took a long time to hit their TP or SL, during that time, a trailing stop would have ended such a trade in a loss. I realize that a trailing stop can help cut losses, but it can also cut profits. By my research, 40% of winners would have ended up as losers if I used a trailing stop. Perhaps you can show some hard numbers showing otherwise? Trend divergence is moving into the realm of human-trading. I know of no reliable way for an EA to calculate divergence....


Ronald,

I have read the whole thread and I can't grasp the concept yet. I just put some numbers and these are the results on the below chart. Do you see anything wrong. This week will start testing. I am very interested in this project.

These are my settings

SL : 200
TP : 200
Win Percent : 90
Minimun Confidence : 90

Do you see anything wrong with the settings?

Thanks
Attached Image (click to enlarge)
Click to Enlarge Name: temp.gif Size: 24 KB
  • #189
  • Jan 17, 2010 12:43pm Jan 17, 2010 12:43pm
I think a trailing stop would be fairly detrimental to quick trades, but if this EA could place longer term trades based on the daily and weekly the breakouts are going to be much larger. So then a trailing stop would be a useful way to collect the profits in a manner that lets them run. perhaps even worth using on the 4hr, not sure.


If the EA's could prioritise your trades though. the best RR trades are going to be on the higher TF's and the worst on the lower TF's.

What if collectively the EA's could decide that a 4hr trade is an A trade 1hr a B trade, 1min a C trade.(just an example) to prioritise the trades with a better chance of return for the available account balance. So if not enough trades are being triggered by the higher TF's then the account is kept busy with the 1min trades.

Jacko posted live trades with trailing stops that did really well, ive noticed DANUK used them in his journal and James16 uses trailing stops aswell. they all take very big rewards compared to the risk to trade.
The greatest force on earth is compound interest.
  • #190
  • Jan 17, 2010 1:30pm Jan 17, 2010 1:30pm
Hello Ronald,

Thank you for sharing your system.
I've finished reading this topic. Looks interesting and i'm going to start this week "playing" with your system.

Regards, Peter
  • #191
  • Jan 17, 2010 1:58pm Jan 17, 2010 1:58pm
Quoting sebastionay
I think a trailing stop would be fairly detrimental to quick trades, but if this EA could place longer term trades based on the daily and weekly the breakouts are going to be much larger. So then a trailing stop would be a useful way to collect the profits in a manner that lets them run. perhaps even worth using on the 4hr, not sure.


If the EA's could prioritise your trades though. the best RR trades are going to be on the higher TF's and the worst on the lower TF's.

What if collectively the EA's could decide that a 4hr trade is an A trade...
I'm not sure how you are able to develop a correlation between a timeframe, and the likelihood of profitability. The EA trades based purely on price, not on any indicators which would be affected by time. The only thing that a timeframe change would do to the EA is decrease the Confidence level. The reason why is because the EA uses the past bars on the chart to determine if a SL or a TP is hit.

Jacko's trading strategy focuses largely on the trend, which is why trailing stops work well there. This strategy does not focus on a trend, it might if you use a large enough TP and/or SL...

I think I'll put that in the next version of the EA.

New features thus far:

 

  1. Trailing Stop, more trade mangement
  2. Choose the timeframe to trade off of.

Anything else?

  • #192
  • Jan 17, 2010 1:59pm Jan 17, 2010 1:59pm
Quoting Forexaccount
Ronald,

I have read the whole thread and I can't grasp the concept yet. I just put some numbers and these are the results on the below chart. Do you see anything wrong. This week will start testing. I am very interested in this project.

These are my settings

SL : 200
TP : 200
Win Percent : 90
Minimun Confidence : 90

Do you see anything wrong with the settings?

Thanks
Those seem fine.
  • #193
  • Jan 17, 2010 2:18pm Jan 17, 2010 2:18pm
Quoting Ronald Raygun
I'm not sure how you are able to develop a correlation between a timeframe, and the likelihood of profitability. The EA trades based purely on price, not on any indicators which would be affected by time. The only thing that a timeframe change would do to the EA is decrease the Confidence level. The reason why is because the EA uses the past bars on the chart to determine if a SL or a TP is hit.

Jacko's trading strategy focuses largely on the trend, which is why trailing stops work well there. This strategy does not focus on a trend, it might...
Yeah I was just thinking of some much bigger TP's automated by a TS with 100 pip SL.

Can't think of anything else yet sorry.

The correlation I see is that with higher TF's you tend to have a lower success rate but the R:R is really high. Wheras with scalping you really need a high success rate because the rewards are so comparatively low.

Overall I think the best trades to be placed are in the higher TF but then your rate of compounding is far smaller than the scalper.

I thought it might be possible to get a happy medium using the EA so your compounding at a high frequency still by using scalp trades to keep the amount of trades nice and high whilst always including the best trades from the higher TF's.

Its probably too complicated or wont work but just an idea. All this was based on me thinking the EA is trend following though.

I look forward to seeing how the EA performs on monday.
The greatest force on earth is compound interest.
  • #194
  • Jan 17, 2010 4:02pm Jan 17, 2010 4:02pm
Quoting Ronald Raygun
Those seem fine.
Thanks for your answer. On which pairs have you applied this EA

Thanks
  • #195
  • Jan 17, 2010 6:46pm Jan 17, 2010 6:46pm
Nice numbers coming up again but no trades placed as of yet.
Attached Image (click to enlarge)
Click to Enlarge Name: trade.gif Size: 33 KB
The greatest force on earth is compound interest.
  • #196
  • Jan 17, 2010 9:15pm Jan 17, 2010 9:15pm
Quoting sebastionay
Nice numbers coming up again but no trades placed as of yet.
That's a HUGE spread!!!
  • #197
  • Jan 17, 2010 10:05pm Jan 17, 2010 10:05pm
Quoting Ronald Raygun
New features thus far:

 

  1. Trailing Stop, more trade mangement
  2. Choose the timeframe to trade off of.

Anything else?

Hi RR,

1. The 'Bar Start' setting tells the counter where to begin counting past opportunities. I have been assuming that bar '0' is the current, forming bar - is this correct?

2. The EA will start at any number of bars after 00:00 ('initial range'). This allows us to determine when it will START its count for statistic gathering, and START to place trades each day.

What about adding the ability to set an STOP time? So if one only wants the EA to look at the London session (as an example), one could set the EA's counter to STOP counting new highs and lows that occur after London has closed, and STOP trading at the London close.

Just a thought as I've been playing with it.

VBro
  • #198
  • Jan 17, 2010 10:12pm Jan 17, 2010 10:12pm
Quoting vbro
Hi RR,

1....

  1. Bar Start 0 means all bars.
  2. This is to generate the initial range. The reason behind this 'delayed start' is 1) to give the EA time to add the previous day's trades to the statistics generated, and 2) to give a sufficient hilo range to start trading off of.

Trading Times are a possibility, I just need to figure out how to build that option into the backtesting engine I wrote.

  • #199
  • Edited Jan 18, 2010 12:36am Jan 17, 2010 10:49pm | Edited Jan 18, 2010 12:36am
Quoting Ronald Raygun

  1. Bar Start 0 means all bars.

Just to be sure...

So if one sets Bar Start to 1, the counter would include all bars on the chart except the most recent bar, or all bars on the chart except the first bar?

Thx.

VBro
  • #200
  • Jan 18, 2010 4:35am Jan 18, 2010 4:35am
Quoting Ronald Raygun
For #3, the additional filter was trade overlap. The EA was coded to handle one trade at a time. There are some instances where a particular trade ran for a very long time, on the order of a few hours. During which time other signals were triggered.
I see.. and other than trade overlap nothing else? I still don't see how that could reduce an average of around 400 trades per week on GBPUSD (by the settings I extrapolated) down to only 4?

I mean, the daily range on the cable is somewhere between 150-200 pips on average. The high and low are hit numerous times as the trading range is established. Even with a 20 pip TP/SL, that's around 4-5 trades per day at a minimum. Unless there's something I'm missing here? :/

One thought I had regarding improving the EA is perhaps adding an option to the optimiser to actually optimise itself. As in - for example -provide a range for the TP, SL & InitialRange (not the BarStart number, as this is probably not so useful to optimise). I'm not sure about your PC, but a month worth of backtesting takes not more than 10 seconds for me. Even a very limited range for that optimisation procedure might work well.

E.g. TP 15-30 (5 pip step), SL 15-30 (5 pip step), InitialRange 10-100 (15 bar step) would be 192 permutations, and on 1 month of data probably take between 15mins-30mins (depending on your PC). Surely this would possibly add great benefit to the system?

A further step along - you could even utilise some sort of distributed computing backtest, whereby others could provide collaborative optimisation data who are trading on the same pair straight to a server, thus providing the best results for all. Thought that might result in a massive bandwagon effect which would end up shifting the market. But hey, it's a thought
2 traders viewing now