Automatic Loss Recovery System (ALR) - read this first

[madpipa]

I'm not going to go look for this in charts. I might get a chance to write a system to demonstrate it. I will bet, though, that it happens more often than you suspect.

I have attached 2 screenshots from a backtest of EURUSD from 1/1/14 - 30/6/14. This is using a TP = 60 with pseudo SL = 20. So the ALR starts where the SL would normally be. The first screenshot shows the 6 months compressed with the number of legs beside all those that had 5 or more turns. The maximum was 11 legs in this dataset. The second screenshot shows the 11 legs in detail.

This can & does happen anywhere - more than I thought it would. The other results are:
1 leg - 8 times
2 legs - 4 "
3 legs - 2 "
4 legs - 2 "

This is not using a trade system for entry - just following the direction of the last candle for opening new trades. And it is using lot sizing suggested in Dustin's video. So each new leg's lot size is rounded up so it can reach the TP in profit.

ALR-EURUSD 1.png

ALR-EURUSD 2.png

[Dewey McG]

I have attached 2 screenshots from a backtest of EURUSD from 1/1/14 - 30/6/14. This is using a TP = 60 with pseudo SL = 20. So the ALR starts where the SL would normally be. The first screenshot shows the 6 months compressed with the number of legs beside all those that had 5 or more turns. The maximum was 11 legs in this dataset. The second screenshot shows the 11 legs in detail.

This can & does happen anywhere - more than I thought it would. The other results are:
1 leg - 8 times
2 legs - 4 "
3 legs - 2 "
4 legs - 2 "

This is not using a trade system for entry - just following the direction of the last candle for opening new trades. And it is using lot sizing suggested in Dustin's video. So each new leg's lot size is rounded up so it can reach the TP in profit.

I am trying to recreate the trades you show with 11 turns, but I see a lower TP on my chart before you get to as many turns. It was harder to see exactly where you were on the other trades, but I could not replicate the same results.

In any event, this makes an interesting point. When we do get an EA we will want to test different settings for different pairs. For the visual back testing I did (and visual back testing is not worth very much) I used a strategy on an H4 chart on a different pair with the 150/50 TP and ALR settings and was not able to generate very many turns.

[madpipa]

Yes Dewey - it is very close to be taken out on that candle. But the EA didn't get an exit. I measured it about 59+ pips - very close! I backtest with 2 pip spread; my standard unless looking at very volatile pairs. It doesn't really matter. I have seen many examples like this. If this trade doesn't get to 11 legs or more then another trade will. As I mentioned previously - one backtest had 30+ legs before it got out of trouble. If you use the ALR enough you will get a situation where it starts to dig itself into a hole.

One answer may be to allow XX legs & then fully hedge the positions. This allows you to step back from the charts & try to find a way out of the problem - either manually or automated. Many good traders, including nanningbob, use hedging to "hold" trades until they decide when/how to deal with them. And therein lies my problem. If I was good enough to trade out of a bad situation that the ALR might get into then I would be good enough not to need it in the first place!!

Another answer may be to allow XX legs, say 6-8, then hedge. Then wait for price to move outside the range where it was getting into problems before allowing the EA to re-start trading. This could be easily automated. Advantage: if price was rangebound we can wait for it to move away & then the EA can trade it's way out of trouble. Disadvantage: price could move out of that range with a firm direction, then go straight back into another range. Sounds fairly common these days!

I also tried using larger settings & the EA got into very little trouble. I think I tried TP = 240 & ALR = 80. Problem was it took a long time because the recent volatility is fairly low. So far TP looks best about 2.5 - 3.5 times ALR - generally, over multiple pairs. I just ran EURUSD from 1/1/2013 till today with TP = 150 & ALR = 50. It only made a total of 177 trades. On 2 occasions it reached 10 legs. I started with 0.5 lots. The largest filled order was 0.37. Next largest was 0.29. So the total exposure is actually quite small.

You will need different settings for different pairs. But don't fall into the trap of highly optimising the EA to your data (I have, a long time ago ). It may be valid to block the pairs into 3 or 4 groups by volatility & assign a TP/ALR for each group. Or maybe use a factor of the ADR for ALR & then make TP = ALR * 3 or similar? Or maybe use the CSS indi - if 2 currencies are paralleling or both low volatility then hedge & wait till later to let EA find an answer?

[Dewey McG]

Yes Dewey - it is very close to be taken out on that candle. But the EA didn't get an exit. I measured it about 59+ pips - very close! I backtest with 2 pip spread; my standard unless looking at very volatile pairs. It doesn't really matter. I have seen many examples like this. If this trade doesn't get to 11 legs or more then another trade will. As I mentioned previously - one backtest had 30+ legs before it got out of trouble. If you use the ALR enough you will get a situation where it starts to dig itself into a hole.

One answer may be to allow XX legs & then fully hedge the positions. This allows you to step back from the charts & try to find a way out of the problem - either manually or automated. Many good traders, including nanningbob, use hedging to "hold" trades until they decide when/how to deal with them. And therein lies my problem. If I was good enough to trade out of a bad situation that the ALR might get into then I would be good enough not to need it in the first place!!

Another answer may be to allow XX legs, say 6-8, then hedge. Then wait for price to move outside the range where it was getting into problems before allowing the EA to re-start trading. This could be easily automated. Advantage: if price was rangebound we can wait for it to move away & then the EA can trade it's way out of trouble. Disadvantage: price could move out of that range with a firm direction, then go straight back into another range. Sounds fairly common these days!

I also tried using larger settings & the EA got into very little trouble. I think I tried TP = 240 & ALR = 80. Problem was it took a long time because the recent volatility is fairly low. So far TP looks best about 2.5 - 3.5 times ALR - generally, over multiple pairs. I just ran EURUSD from 1/1/2013 till today with TP = 150 & ALR = 50. It only made a total of 177 trades. On 2 occasions it reached 10 legs. I started with 0.5 lots. The largest filled order was 0.37. Next largest was 0.29. So the total exposure is actually quite small.

You will need different settings for different pairs. But don't fall into the trap of highly optimising the EA to your data (I have, a long time ago ). It may be valid to block the pairs into 3 or 4 groups by volatility & assign a TP/ALR for each group. Or maybe use a factor of the ADR for ALR & then make TP = ALR * 3 or similar? Or maybe use the CSS indi - if 2 currencies are paralleling or both low volatility then hedge & wait till later to let EA find an answer?

I know in one of the videos they said if you do reach your maximum number of turns to manually trade out of the hedge. Joseph says in one of his videos to use ALR to determine size of the zone. I would also like to test whether using rising volume as a filter to enter into trades helps since that would indicate a bigger move is about to happen.Once we have an EA to work with I can start testing these ideas to see if we can make this work.

What is the maximum number of turns you saw?

[alfonsomg]

I've coded this weekend an EA about this strategy. I'm not 100% sure I covered all the system details, I think I did (basically buy or sell at the key levels, recalculating the lot size to escape from a loss).

Testing it in Empty4 EURUSD 2001-2014, with 150-50 TP and SL (I have an option to make it dynamic depending on ATR). Entries with long term trend. I didn't play too much with the testing.

Conclusion: eventually it blows the account. I can upload the code later (I'm sure there is a lot of room for improvement) if anyone is interested. But in my honest (newbie) opinion, this is a martingala, a soft martingala if you want.

EDIT: I just want to say that wen I said "it blows the account" I meant that if the lot size grows too much eventually you won't be able to open the next trade, so if price goes against you you'll have a margin call.

[Dewey McG]

I've coded this weekend an EA about this strategy. I'm not 100% sure I covered all the system details, I think I did (basically buy or sell at the key levels, recalculating the lot size to escape from a loss).

Testing it in Empty4 EURUSD 2001-2014, with 150-50 TP and SL (I have an option to make it dynamic depending on ATR). Entries with long term trend. I didn't play too much with the testing.

Conclusion: eventually it blows the account. I can upload the code later (I'm sure there is a lot of room for improvement) if anyone is interested. But in my honest (newbie) opinion, this is a martingala, a soft martingala if you want.

EDIT: I just want to say that wen I said "it blows the account" I meant that if the lot size grows too much eventually you won't be able to open the next trade, so if price goes against you you'll have a margin call.

There is no way it should blow an account if used properly. You can and should limit the total risk and total number of turns. Whether or not this proves to be a valid strategy, if you blow an account you are not using it prudently.

[Dewey McG]

Here are some thoughts while waiting to see if someone can code an EA for this:

Gary makes a valid argument why this might not work. I am not totally convinced, but his warning should be heeded and explored. If you watch the videos on this from either source it would be easy to get carried away by all the hype, but NOBODY should use this on a live account until all sorts of testing has been done. I cannot begin to count the number of times when due diligence in testing has saved my account when I thought I had a great strategy. (Now if I had only used more about my broker, … but that’s another story.)

I do think there is also enough reason to believe this is worth testing and that with the right strategy or strategies could prove to be very effective. Random entries won’t work. A high win rate on the initial trade will be crucial to the viability of this.

Choosing the ALR zone will be very important for each pair and strategy. Upon more reflection I don’t think this would work well on a scalping strategy as you could get caught in a “porcupine” with sharp spikes up and down triggering too many trades. The other issue with an ALR setting that is too tight will not allow the initial trade enough breathing room to have a high enough win rate.

I would suggest that everyone starts thinking about strategies that might be well suited for this. Set ups that look for big moves come to mind. One example would be breakout strategies, especially ones that use volume (e.g. using the Gadi OBV indicator). I use VSA (Volume Spread Analysis) on my own trading and am thinking Effort to Rise, Effort to Fall, Shakeouts and Reverse Shakeouts would be good candidates since they usually result in a big move in your direction except when they don’t—then they often have a violent move the opposite way instead. Feel free to suggest other strategies that would be worth trying out with this.

Anyways, not much more we can do until we have something to play with.

[madpipa]

Dewey McG » Mon Jun 30, 2014 8:54 pm

What is the maximum number of turns you saw?

The maximum I have counted so far is 33 turns/legs. Probably worse cases out there too! But remember - my EA is always in the market. I am trying to get it into as much trouble as possible. There is no point looking at the "nice" side of the concept without looking for its failings.

I do agree with you - I don't think this approach would work very well in a scalping type environment (despite what Dustin says ). The vibrations of price up & down could make a mess of ALR fairly quickly if you let it. I also believe in the fractal nature of the market. So what can happen in a small timeframe can also happen in a long timeframe.

[Dewey McG]

I also believe in the fractal nature of the market. So what can happen in a small timeframe can also happen in a long timeframe.

True, but there is less "noise" on higher time frames.

Also, I think the issue should not be whether it hits a full stop loss, but how often. Will it be low enough to make the strategy viable? As a marketing tool it is wonderful because the stop losses are rare and by the time the user notices them the refund period is long past.

If nothing else I think it will be fun to test.

[wealthmaster]

Dewey McG » Mon Jun 30, 2014 8:54 pm

What is the maximum number of turns you saw?

The maximum I have counted so far is 33 turns/legs. Probably worse cases out there too! But remember - my EA is always in the market. I am trying to get it into as much trouble as possible. There is no point looking at the "nice" side of the concept without looking for its failings.

I do agree with you - I don't think this approach would work very well in a scalping type environment (despite what Dustin says ). The vibrations of price up & down could make a mess of ALR fairly quickly if you let it. I also believe in the fractal nature of the market. So what can happen in a small timeframe can also happen in a long timeframe.

It does not matter if it goes 1000 turns. We will take our loss at 10th turn max.. ( and that is still worst case scenario and still around X% of our account, not a whole account at stake)

I think you guys haven't really understand the point Dewey is making..

The whole idea is converting your loosing trades into Breakevens or in super worst case scenario (i.e 10th cycle X% of your account)