What RaptorUK and other sophisticated MQL4 traders have in common

I can only speak for myself.

You assume too much . . . I don't have a profitable EA, I am not a "sophisticated trader", I have learned a little mql4 coding and come here to try and help others learn. Perhaps I don't have a profitable EA because I have a very critical definition of what I call profitable . . .

I don't know if I should be responding with a heading like that :).

However I do agree with RaptorUK coin_toss arguments. Here & Here.

I have a taught question for RaptorUK since he's already been called out :P

Do you believe that the R:R and WR could be manipulated to a traders advantage?

I believe you've hinted that we can force the WinRate to favorable values like 90.

Do you think it'll be possible to force the Average_Win / (Reward) higher by increasing the lotsizes?

In that way, pushing the RR:WR above the coin_toss when ever it falls below it.

1. Being sophisticated does not usually translate in to being successful.
2. There is something to be said for the 'keep it simple stupid' KISS approach.
3. There is only one thing that makes you successful (correction).......
There is only one thing that MEANS you made a successful trade.
4. You bought it cheaper than you sold it. Period.
5. Let the market come to you.

If you like to trade with the larger trend, wait on the pullback. So that you have a framework around which to set a stop loss for good r:r.
(Think of pushing a child on a swing at the appropriate time.)

If you like to trade a range. Wait on the swing high or low for the same r:r reasons.
(Once again think of the child on the swing.)

Look at a range as a sideways trend.

1. look for the high probability places on the chart.
2. Only there can you implement good Reward to Risk.

Does your indicator get you in at some other place on the chart hoping to have 'guessed' the direction to come?
Or, does it get you in at one of the places mentioned above because of WHERE it is RIGHT NOW?

Does your signal come from CURRENT price action. or...
Math on PAST price action?

All that matters is where price is right now. It's either time to get in or time to wait.
Review your strategies and see how many of these rules they are breaking.

I don't care how much equity, leverage or time you have. If you don't buy it cheaper than you sell it. You CAN NOT succeed.
So make sure your strategy doesn't get too 'sophisitcated' and lose it's focus on what it's really trying to accomplish.

I can tell you... It's easier said than done...PipPip..Jimdandy.

p.s. I wrote a few pages about this very subject a few days ago over in my blog. Called the Holy Grail. You may find it interesting.

I've read a few books on trading systems and algorithmic trading.. the one message that keeps appearing, and agrees with my own experience, is:

The actual system can be extremely simple and still earn money. The crucial factor seems to be precise entry and exit strategies.

As for 'days where nothing works', they should be predicted by your backtesting. You should know the (expected) maximum drawdown, and longest drawdown duration before forward testing.

Jim, can you post your blog link? I'd love to read

I don't think there is any such thing as a favourable or unfavourable WinRate (WR) just as there is no such thing as a favourable or unfavourable Risk:Reward (R:R) . . . one has no meaning without the other. Why is a WR of 90% better than a WR of 50% ? we can't answer unless we know the accompanying R:R in each case.

Can we force the WR to a desired figure ? lets take my Coin Toss example, yes we can get any WR we desire, just set the corresponding R:R by setting the SL & TP accordingly and the WR will end up where you want it . . . assuming you execute a large enough number of trades. Even with a WR of 99% we will still not have a profitable strategy . . . remember, it's a Coin Toss.

Yes, of course . . . but that will also make the Average loss higher . . . R:R is a ratio and is independent of position size . . . unless you have some way of knowing in advance which trades will be winners and which losers, then you can have a bigger position size on the winners

If a Strategy is a Coin Toss then the WR is inextricably linked to the R:R and visa versa . . . . the question I still have bouncing round my head is what is a good benchmark to use so that we can decide that a Strategy is better than a Coin Toss ?

If you look at my data here you will see that my table contains actual measured WR and theoretical calculated WR, there is a difference even bearing in mind the number of trades, for example: measured WR 37.79% vs calculated WR 37.5% number of trades 111,418. So even with over 100k trades we can still get an inaccuracy of 0.77%. My average error was 0.3% and my max error was 6% followed by 2.5% then 1.66%, so most data points have an error of less than 1%.

For my Coin Toss system it's simple to determine the R:R, the EA has fixed SL and TP, my tests were all run with 0.0 spread so it's very simple to calculate the R:R. In many Strategies there may not be a fixed SL & TP, is many others there is no SL and maybe not even a TP, so in these cases how can we determine the R:R ? I used to think that the Average Loss:Average Win would give us our R:R and in many situations it will, but not all . . . in the situation where a set SL & TP are used it will give us the R:R and account for the spread at the same time. In situations where a set SL isn't used the Average Loss doesn't give the true picture about the risk, why ? risk is not the same as loss, loss is determined by realized risk, risk can be greater than the realized risk but it is still risk even if it is not realized. The same is not true for reward and win, win does equal realised reward. This may seem counter intuitive but think about it for a few minutes . . . risk is a potential measure not an actual one, reward is an actual measure not a potential one.

So what do we do where there is no set SL ? in this case I think we have to use the average MAE instead of the average loss. So now we can calculate the achieved R:R and use this to establish what WR we need to break even and compare this to the WR being achieved, what difference do we need to have some confidence in our Strategy ? I don't have a concrete answer but the greater the difference the more confidence we can have, I would also say that for a low value WR ( < 50% ) we should be looking for a difference of maybe 10%, obviously for a BE WR of 90% this would require a WR of 100% and perhaps in this situation we should be aiming for a WR at least 10% lower than the required BE WR and then use this strategy in reverse.

Note: Please bear in mind this is all in my opinion, I'm happy to be contradicted by anyone as long as they can backup their contradiction with a reasoned argument.

Advertising is not allowed on this Forum, please don't encourage it, PM Jim and I'm sure he will reply.

Thank you all for the replies. @JimDandy: I like the swing approach and will implement it.

@RaptorUK:

I agree that by increasing the lot_size, there's a potential risk of increasing the average_loss also. I was just hoping that should I somehow increase the win_rate to say 90% by having Take_Profit=6 and Stop_Loss=60 (for example). Then somehow, if I toke a loss, or kept taking a loss, the increased lots on future trades would stand a better chance of success. But of course, thats probably the definition of gamblers fallacy. Funny thing is, I've probably already performed this test sometime in the past and trashed it :) cycles:cycles.

Anyways, reason I'm bringing this up_now is because I'm wanting to generate a thread or case_study about evaluating edge/advantage for systems. A recent system of mines makes me want to believe that a trader can influence his RR:WR. I have some theories which I'm studying with no_stoploss systems. Believing that what applies to no_stoploss could also apply to stoploss systems is why I'm asking these questions. Of course I don't fully understand the answers, so apologize if I sound elusive.

As far as floating losses, I've always considered them as realized losses. One problem which always jumps at me when attempting to use floating losses in the RR:WR equations is the fact that the average_loss becomes huge for no_stop systems. I think most negative_grid systems would appear (unjustly) negative with such approach. Therefore my compromise was looking at the account (profit / drawdown) in $$. Instead of pips, $ is much more flexible to variable lots tests. This ratio should be calculated on on-going basis.

When observing our recent FastWayToMillions, it reminded me about something I discovered and neglected sometime ago. At the time, I wrote this comment here. I'm really trying to figure-out how to apply this concept to stop-loss systems.

And this gives the real story and explains why a signal that appears to be working well suddenly goes bang . . . if the real R:R had been considered instead of the Ave Loss:Ave Win the real potential performance would have been apparent.

Agree with the Bang! part :). Have you ever built a draw-down per order analyzer? I mean like something which could be used outside a back-test like upon strategy_tester report || live reports.