I am actually wrong in this argument, but for none of the reasons TSGannGalt was so poorly trying to articulate.
Adding the extra layer will do nothing for profits -- it will only waste development and testing resources.
Why? Well, you see, whether TSGannGalt NETS contracts or I take every trade, we end up entering the IDENTICAL number of trades, and during the period he is NET flat, I am HEDGED flat. Whether he sells one long contract he holds when he gets a short signal, or I take the short, we have identical exposure and the same number of trades.
Lets show a contrived example for the sake of pedagogy. I will use TSGannGalt's L, S, F system to mark signals. Assume the time-series and signals as follow,
800 L F
801 L F
802 L F
803 L S -- Here, TSGannGalt will be flat
802 L S -- Here as well
801 L F
802 L F
802 F F
Now, assume my system, where I don't net out trades. I go LONG @ 800 and sell that at 802. I also go SHORT @ 803 and cover at 801. A total of 4 points profit with 4 trades.
TSGannGalt, on the other hand, will go LONG @ 800, sell out at 803, re-enter at 801 and sell out at 802. He gets a total of 4 points with 4 trades.
Now, this is a rather contrived example, so let us now generalize it. I was speculating that such a system as I was proposing might be useful in the case where a 15-minute counter-trend shouldn't net out a daily trend, as it doesn't invalidate the daily trend necessarily.
Let's assume I am trading M contracts for the long signal, and N contracts for the short signal.
In my case, I will be long M contracts from 800 to 802, and short N contracts from 803 to 801. A total of 2M + 2N profit.
In TSGannGalt's system, he will be long M contracts from 800 to 803, short/long M-N contracts (still long if M > N, short if N > M) from 803 to 801, and then long M contracts from 801 to 802.
In the case where M and N both equal 1, the systems are the same. Expanding, we can see that when M=N, the systems are identical. His net is identical to my hedge.
What about if M != N? In my case, the profits are the same: 2M + 2N. In TSGannGalt's case, you get 3M - 2(M-N) + M ... which is 2M + 2N. Again, not a surprise. Netting and hedging are IDENTICAL.
We can generalize this to any situation. Instead of constants, we can use A, B, C, ..., Z to represent price moves. You can see that they are the same.
So, all in all, TSGannGalt was right (though, quite unable to explain why).
You hear that, TSGannGalt -- you can gloat all you want now. You are indeed correct, sir. There is no benefit in developing such a system.
BUUUUUUUUUUTTTTTTTTTTTTT (oh come now, TSGannGalt, you didn't think you could get me that easily, did you), I also talked about having one system invalidate the signals of the other, so let's look at that.
Now assume that I am trading a M contract trend following system and a N contract counter-trend following system, on the same time-frame -- meaning that 'counter-trend' will now invalidate trend-following.
Examine the situation now.
800 L F
801 L F
802 L F
803 L S -- Here, TSGannGalt will be flat
802 L S -- Here as well
801 L F
802 L F
802 F F
TSGannGalt gets the same results as before: 3M - 2(M-N) + M.
I, on the other hand, get 3M + 2N + 1M. Wait, what? That's right -- if instead of simply NETTING, you can recognize when one signal INVALIDATES the other, you can INCREASE PROFIT. By identifying that on the same time-frame, my counter-trend invalidates my trend, I can not only go short, but I can also take my profit from the long, meaning that I am now bias short to take advantage of the counter-trend.
So the benefit lies not in where they DON'T INVALIDATE each other -- because in that situating, NETTING is identical to HEDGING, but in where they DO invalidate each other.
