Thread: The difference between a MA and a non lag MA?

Originally Posted by ;

Disagree with this. The nearer price approximates an entirely random walk (from the gaussian, IID feel ) the greater chance you have at a non-zero expectancy. In moving averages and real market behavior, it's the fat tails and non IID (serial correlation) behavior that are killers.

From'random walk' I mean a coin-toss, i.e. exactly 50/50 chance of uptick/downtick at each point. That is the exact same as assuming a perfectly'fair' wheel, the game can't be beaten, because there aren't any inefficiencies.

Originally Posted by ;

From'random walk' I mean that a coin-toss, i.e. just 50/50 likelihood of uptick/downtick at each point. That is the exact same as assuming that a perfectly'fair' wheel, the game can't be defeated, because there are no inefficiencies.

A perfectly gaussian generated walk with zero bias matches the above criterion, just the magnitudes are not binary (although the states can be made equivalent). And I know it goes against conventional wisdom, but as soon as you see your conclusion is not true (with your own eyes) it will change your entire way of considering market signals. It was among the deeper insights I had years back (and that I know I am not the first to watch it).

Originally Posted by ;

A perfectly gaussian generated walk with zero bias meets the above criterion, only the magnitudes are not binary (even though the states may be created equivalent). And I know it goes against conventional wisdom, but once you see your decision isn't true (with your own eyes) it will change your whole way of thinking about market signals. It was one of the more profound insights I had years ago (and I know I'm not the first to observe it).

Should you want me to see'it', you will want to explain it in simpler terms.

Are you saying that if I have tossed (state ) 50 consecutive heads, my next toss is more inclined to be a tail? If your argument relies on such a theory ('Gambler's fallacy'), then it is an argument I have had with a number of other folk: let's just agree to disagree.

[EDIT] The (hypothetical) market equivalent is that, no matter how many pips price has moved in a specific direction, there is always just a 50% likelihood of the move continuing, and a 50% likelihood of a reversion toward an arbitrary prior mean. If that were the case, then there would be no exploitable inefficiency, and expectancy of any future move (up/down) would constantly net out to zero. I may be incorrect [I do not understand what you mean by the magnitudes are not binary (though the states can be created equivalent)], but I think the distribution you are talking about earning some sort of assumption that a reversion to a mean is obviously more likely than a move continuation. That may be the some sort of'random walk' in strict math terms (I'm not a mathematician), but when there is a directional bias at any given point, then it is not the type of'random walk' which I'm talking about.

Originally Posted by ;

If you want me to see'it', then you'll want to explain it in simpler terms.

Are you saying that if I've tossed (say) 50 sequential heads, my next throw is more likely to be a tail? If your argument is based on such a theory ('Gambler's fallacy'), then it's an argument I've had with a number of other folk: let's just agree to disagree.

Not today. Seeing Gambler's fallacy, a random computer algorithm could easily generate a billion large nations in a row (and I've seen it done) by chance, but have you ever noticed an unbiased coin come up heads per billion times in a row?

The simple theory I am speaking about will be based upon properties of gaussian behavior and related likelihoods. Notice, to be clear, I never said markets were gaussian lies the rub.

I will simply leave it as food for thought and allow the grenades be thrown for a while. We could agree to disagree; no harm .

Originally Posted by ;

Not today....

So that your argument is based around the assumption that perfect random number generation is hopeless?

I am talking hypothetically; I concur that markets are not perfectly gaussian, and that's precisely why I believe that exploitable inefficiencies do exist.

I think that the point of an MA would be to eliminate noise by averaging it out.

You Would like an MA to lag. Imagine a hundred copies of a photograph. Suppose that every copy was overlayed with a guassian sound filter. Each picture is an difficult-to-recognize and obscured variant of the picture. You'll get a consequent imagine that is much less obscured than some of its composites if you average each pixel between the images.

A nonlagging MA defeats the whole point of averaging... the price itself could be an perfect non-lagging MA.

MA kills greater frequency. It destroys periodic within the signal. Do you really want that? You would like to kill noise but not the signal. This is the reason MA are not good.

Originally Posted by ;

I think the point of an MA is to eliminate noise by averaging out it.

You Would like an MA to lag. Imagine a hundred copies of a picture. Suppose that a guassian noise filter has overlayed each copy. Each picture is an obscured and difficult-to-recognize version of the same picture. If you average each pixel between different images, you'll get a imagine that is much less obscured than any of its composites.

A nonlagging MA defeats the entire point of averaging... the price itself would be an perfect non-lagging MA.

Agree with you; the primary idea behind any averaging procedure (MA, Heikin-i, Renko, etc) is to smooth out the false signals caused by'noise'. Having a MA, smaller interval less smoothness, and less lag. A 1 period MA is the equivalent of price .

Lag results in later entries, forfeiting pips; however, such a delay might cause a false sign being avoided, especially when price is moving sideways.

For better or worse, the'non-linear' MAs (Hull etc) try to decrease lag while still keeping some smoothness. That gives the best of both worldsnonetheless, if those MAs actually delivered profit in spades I'd expect them to become recognized as some sort of Holy Grail. Yet they're not.

Hanover

Know that this is beyond the thread scope but interested on your views about the usefulness of std deviation analysis in FX analysis. Seems though to be a reasonable extension to discussion on MAs. Also has some intuitive appeal to a newbie like myself and tools like Bollinger Bands are.

Originally Posted by ;

So your argument is based around the assumption that perfect random number creation is hopeless?

I am speaking hypothetically; I concur that markets are not perfectly gaussian, and that's precisely why I believe that exploitable inefficiencies do exist.

Pls Mr Hanover you're more educated here and even so on Currency Market. I only want help with this question:

Is it feasible to incorporate the bollinger bands levels in a code, pls can you kindly give me a hint or refer me to somewhere I might get a sample. Thank you in advance.

Regards
.