Balls to the Wall, a 1 minute 1 trade EA

[vivaldi]

Very interesting ... a few months ago I made an EA for this, and in backtesting over a decent period of time (2004 onwards) it had some periods where it profited, but longer & more frequent periods where it took heavy losses. I experimented with covering BE, trailing stops, messing with the TP and SL levels, and reversing the signals - nothing improved its basic performance or lack thereof.

I can dig it up if anyone's interested. Not wanting to be a wet blanket or anything as who knows, maybe I coded it wrong or perhaps the market's changed and this approach is now valid.

What exactly were the rules for your EA ? The strong point of this EA is its Risk Reward, which is 3:1. A hitrate greater than 33 % is already enough to make this method profitable.

Is there some wisdom here behind the rules for entry ? Yes and no.

The initial reasons for entry are valid ones : the bias in the morning. This is however a rather weak measure : the low of the previous hour. (Why not the high by the way ?).

Further indicators used are the stochastic and the macd readings.
But in a timeframe of M1 these reasons for entry evaporate rather quickly. Or is it noise to start with ? So, it looks like it's most of the time more a question of luck than of wisdom if a trade works out.
But, as long as the hitratio is high enough, this method should work. We will see.

[garyfritz]

slipshod, I'd love to see your results. Might save me the trouble of coding it in TS.

I looked at this a bit more...

The strategy is based on MACD and Stoch.

M1 MACD is very similar to M1 TMA Slope. Try putting them both on the same chart and compare them.

Meanwhile the color-changing stuff is basically detecting MACD (/Slope) direction. When positive and increasing it's light green, when positive and decreasing it's dark green, etc. So this strategy is doing something very similar to the Slope Direction idea I raised HERE. I haven't gotten around to carefully testing my ideas, but spotdespot in that thread seems to be doing well with his variant of it.

Meanwhile, Paulus' strategy takes one trade after the London open and then waits for tomorrow. I think that's because it has a predictable direction at that time. I suspect that you could extend this strategy to trade almost any time, by using something else to determine direction. E.g. if the M15 Slope is negative and increasing, it's usually a good time to go long, and vice versa. So we may be able to extend this past the one-a-day signal that Paulus created.

[SteveHopwood]

Steve:

Since we have the option of using slope, can you make disabling the manual bias-line requirement an option? The slope should provide the bias we require, I think. Glenn

Okies, V 1d in post 1, The input DisableLineUse allows you to rely on Slope alone to provide the trading direction.

I have also changed the LondonOpenHour to MarketOpenHour. I am experimenting by settup up 3 charts:

• - EG to draw the line at 8
  - EU to draw the line at 13
  - AJ to draw the line at 20

Watch your crim's leverage folks. At 5% risk on my Cowboy IBFX mini with 1:100 leverage, a trade immediately sent the account to within 9 points of a margin call, so I have cut RiskPercent to 2.5. Event his left my Margin Level at 198 (margin call at 100).

[garyfritz]

?? That's odd. I routinely risk 2.5% with my 50:1 account, and it generally uses only about 65% of my margin. I would have thought 5% risk with 100:1 would have been the same??

[vivaldi]

Hallo Steve,

The estimated lot size = 0 on my screen. Also the estimated stop loss = 0 and take profit = 0. (DisableLineUse=true, that is why I see them now outside market open hours. If I set DisableLineUse=false, these estimates disappear.)

However, the estimated lot size should be something, at least 0.01. Empty4 runs in demo, with a $5000 balance. RiskPercent=5, running version 1d.
Do I have something wrong with my settings ?
Are there other people, who have the same problem ?

[SteveHopwood]

Hallo Steve,

The estimated lot size = 0 on my screen. Also the estimated stop loss = 0 and take profit = 0. (DisableLineUse=true, that is why I see them now outside market open hours. If I set DisableLineUse=false, these estimates disappear.)

However, the estimated lot size should be something, at least 0.01. Empty4 runs in demo, with a $5000 balance. RiskPercent=5, running version 1d.
Do I have something wrong with my settings ?

Yes.

I reply in such monosyllabic mode to your posts because you are doing something wrong at your end, and I neither know nor care what it is.

Fact is, it is something that you are doing wrong. Please sort this out for yourself. Previous posts of yours in other threads, have persuaded me that you have something useful to offer in my forum, otherwise your name would not sport the 'Battle-scared vets' colour. Kindly demonstrate that you deserve this by doing some thinking for yourself.

[erickva]

?? That's odd. I routinely risk 2.5% with my 50:1 account, and it generally uses only about 65% of my margin. I would have thought 5% risk with 100:1 would have been the same??

Hi Gary,

I think it would depend on how big your SL is.
With this EA SL can be as small as 7 pips, this will skyrocket the position size, even though you have a fixed/controlled risk.

Cheers,
Erick

[slipshod]

What exactly were the rules for your EA ? The strong point of this EA is its Risk Reward, which is 3:1. A hitrate greater than 33 % is already enough to make this method profitable.

The rules were exactly what Paulus described on the 1st page of his FF thread. Break-even for this strategy is actually closer to 25% hitrate, though you have to add more to allow for spread & slippage etc. Its only a "strong point" if the strategy can actually achieve that hitrate. In my testing it doesn't, not in the long run.

slipshod, I'd love to see your results. Might save me the trouble of coding it in TS.

Sure ok, I'll go find it & send it in a PM.

[SteveHopwood]

The rules were exactly what Paulus described on the 1st page of his FF thread. Break-even for this strategy is actually closer to 25% hitrate, though you have to add more to allow for spread & slippage etc. Its only a "strong point" if the strategy can actually achieve that hitrate. In my testing it doesn't, not in the long run.

It is worthwhile remembering that when Paulus first posted his thread, he only had 5 weeks of trading to demonstrate.

So, whether the system has legs of not has yet to be seen but I can state this with absolute certainty: backtesting Crapql4 in CrapStrategyTester will tell you nothing other than the code is working correctly. That apart, forget it.

Moving forward, there is one huge advantage that we have here that Paulus clearly lacks in his forum.

Us.

If it either works or does not work, then we will prove/disprove it quite quickly. If it works, but in a different form, then someone here will find out how.

[slipshod]

Yes exactly Steve, I was going to point that out as well. Paulus never claimed this was a solid reliable & heavily tested system, it was just an idea he put forward to see where it went.

I look at backtesting a different way. If an EA backtests profitably in Empty4 (provided its not cheating by using MTF candle values) there's a chance it may also do so in real trading. If it chops along sideways then there's hope it may have value. If, however, it drains your account in backtesting, no matter what settings you use, then imho its not worth wasting any more time on.

Regarding this strategy, my backtesting (fwiw) showed that sometimes it worked well for months, even a year, at a time. Then it fell in a heap with multiple days & weeks of the strategy failing to turn a profit, eliminating all the gains and then some. Anyway, who knows perhaps I coded it poorly & that's to blame, not the strategy - I'm sure Gary will tell me soon enough if that's the case

Edit: I attached my EA here fwiw - I won't be supporting or modifying it in any way, anyone's welcome to use it for whatever they like.

Edit #2: This uses the LibOrderReliable4 library, which can be downloaded from the Coder's forum. I also set a fixed stop level & it doesn't check for a prior higher-high (for shorts) & lower-low (for longs), which is a feature that wouldn't be too hard for someone to add if they wanted.