Thanks for not reacting adversely to my post above …
The main point, I think, is that that when you add other indicators to try to “confirm the signals” of some indicator(s) you’re already using, you’re doing two main things: first, you’re reducing the number of trading opportunities that the system produces even if you’re also increasing their win-rate; secondly, you’re increasing the degree of backfitting (to whatever sample you’ve tested on to achieve validity/profitability from the increased-indicator combination), which tends overall to reduce prediction-value of the system, moving forwards.
It doesn’t follow that the highest win-rate systems are the best (in fact that’s understating it: the highest win-rate systems are usually [U]not[/U] the best). It’s better to make an average of 8 pips profit, an average of 6 times per day, than it is to make an average of 30 pips profit, an average of once per day. (That’s just an example plucked out of the air to illustrate the point: you can, of course, change all the numbers in it to make it more appropriate to your own style of trading and parameters). Not only better for the bottom line, but also often better for the smoothness of the equity-curve - a very commonly overlooked parameter.
When you backtest, to determine which indicator-combination “works best” for the sample data on which you’re testing, the greater the number of indicators you’ve used, the higher is the chance that your findings will be curve-fitted to that specific sample and therefore the lower is the chance that it will continue to be profitable.
Possibly a horse-racing analogy will clarify this last point (which is probably my main point, anyway). Imagine that you’re selecting the horses to bet on at the race-meetings at Ascot. You start by noticing that horses who won their last race tend to do very well, so you use that as an indicator.
It produces some winners and not terribly many losers, but unfortunately they mostly pay bad odds, so the system isn’t quite profitable.
You decide to add another indicator, and looking through the previous year’s Ascot racing results you discover that the horses who won their previous race [I]and are running again within a week[/I] of that previous win contain an even higher proportion of winners. So you add this indicator as well, reducing the number of horses you can back, and find that it’s nearly profitable that way.
Wanting to improve it further, you also notice that the group of “horses fitting the above categories whose names also start with the letter K and whose brothers have recently won a race at Epsom specifically on a Wednesday before 4.00 in the afternoon” contain absolutely no losers at all and that all those horses won their races at odds of 3/1 or better.
Now perhaps you have the “perfect system”, because you’ve confirmed the signals of your first indicator or two by applying another indicator or two, and historically there are no losing trades and a good R:R as well?
The reality, of course, is that you’ve done no such thing. All you’ve done is “backfitted” a system to the sample data from which you started.
Horses having won their last race is a reasonable criterion (because they’re probably in good form against their peers). Horses having run within a week is a reasonable criterion (because they’re probably race-fit). Horses whose names start with a K and whose brothers won at Epsom early on a Wednesday afternoon is nonsense - it has nothing to do with their [B]future[/B] winning chances at all, although it [I]apparently[/I] (key word) had everything to do with their [B]previous[/B] winning chances.
Causation and correlation are two different things.
As you see, I’ve chosen an [U]obviously[/U] stupid example, to help to make the point clearly.
In trading, many people do this with indicators [I][U]without[/U] knowing which ones are stupid[/I]. Because, in trading, knowing which indicators are silly takes loads of experience, judgment, understanding, and all the things that so very few aspiring traders really have. And there’s an enormous amount of “information” out there, especially online, written by people who also don’t know, so it’s [B]terribly[/B] easy to misjudge.
My point is that adding indicators can reduce the chances of dealing with causation rather than correlation.
All that said, it remains true that there can be and are, also, valid examples of doing this and making better systems out of it. As a counterpoint to the warnings above, I’ll mention a friend of mine (very successful trader) who trades a system in which the entries are suggested by a “short, fast Ichimoku” indicator. She has successfully “confirmed” its signals by adding a “long, slow MACD” as a directional bias. She takes the long Ichimoku signals as entries only when the MACD line is above both its signal-line and its midline (and [I]vice versa[/I] for short trades). It worked. It improved her method and her income without unduly reducing her trading opportunities. The reality is that by adding a very slow indicator to a very fast one, what she was [B][U]really[/U][/B] doing (and of course she’s the first to acknowledge this) was effectively adding an additional [U]time-frame[/U]. What the indicator happened to be wasn’t very important at all ([U]within reason[/U], provided it’s not a completely stupid one - and neither Ichimoku nor MACD is inherently stupid). But this takes some experience to appreciate - and, unfortunately, the places one looks (especially online!) to try to harvest the fruits of some of this experience are also full of nonsense about homeopathy and astrology … oops, sorry, I mean Elliott and Fibonacci, and the neophyte has [B]no practicable way of distinguishing[/B] between that and “the real thing”, because nobody was born knowing how to do this stuff.
So what I’m really saying (with absolutely [B][U]NO[/U][/B] disrespect or condescension implied at all, I hasten to add, because we have [I]all[/I] “been there; done that”!!) is that (a) it’s actually a pretty dodgy principle to start with, and (b) you’re probably not at this stage going to know how to do it productively anyway.
Just my perspective. :8:
Edited to mention …
There [B]isn’t[/B] an “exchange” for spot forex. Nor is NinjaTrader necessarily needed for that anyway. You’re posting total nonsense again, Mr. “Verified Analyst”. Still, it may help Grozny to appreciate how much misinformation there is, around, from apparently “good sources”, which is actually part of what I was explaining above - so maybe I should thank you for adducing some evidence for my point. :rolleyes: