Thread: Statistically testing every indicator with my Genetic Algoritm

Originally Posted by ;

quote I'm also using ATR to determine whether RR is okay. That way I never use mended SL/Tp, those need to be dynamic as situation develops.

What ATR setting would you use, since its really hard to quantify that which is optimal.I use the previous 48 hours average.

Originally Posted by ;

quote What ATR setting would you use, since its really hard to measure that which is optimal.I use the past 48 hours ordinary.

Since I trade largely on greater timeframes I use past 12-18 days average (past 72-110 bars for 4H) determined by the pair

Hi Proximus,

I wonder whether it's really a good method to test an indior by basing both entrances and exits it. If the indior triggers infrequently, such as RSI 90 and 10, the system will be holding transactions and the results will be functionality based on what the markets failed.

Also I am interested in if you rearrange the time show inside your own clusters. Are the period of the time show inside your random clusters exactly the same?

Originally Posted by ;

Hello I wonder whether it's a really good method to check an indior by basing both entries and exits on it. If the indior triggers for example RSI 90 and 10, the system will probably be holding transactions and the results will probably only be performance based on what the markets did. Also I'm curious about if you rearrange the time series inside your own clusters. Are the period of the time series inside your clusters the same?

It was only an approach, nobody said that its the best one, of course for any indior there should be a neutral zone besides BUY/SELL zone where you could exit the transaction.

For RSI you could set it to 50, since 30 is BUY and 70 is SELL.However I tested this technique in which you buy hold type, according to my theory that if the indior is accurate, then it signals the very best place to get in and the very best place to get out.

So its logical to buy at buy signal depart buy at market sign and vice versa, cos if the indior is really accurate it should provide the signs only when there's a true tendency shift, not at fakes.

As for intra-trade drawdown, that is due to fakeouts.Its not a healthful method from MM standpoint, but it engulfs all noise and you only trade pure BUY SELL signals.

The strategy was meant to measure the truth of this indiors not to supply a small risk trading platform.
The period of the cluster has a random seed, however its largely based on volatility cycles * with a random amount to encompass an error margin.
Formerly I used only random seeds but it turn out that it was innefective therefore I introduced a volatility based cluster.

Hey, why are my 3 favorite indiors mentioned previously or not listed? Just wondering... Something for you to think about, sir.

How did you backtest your algorithm? Was it on anonymous information? Or was it you used in the devellopement?

I believe one should have two TPs. One is static, one is lively, whichever hits first.

In my opinion lively TP are something like this:

For Long entries such as:
EntryPrice (Sqrt(BarsSinceEntry) * (ATR (Stdev(ATR) * x))
Here the x would become your TP multiplier (in fact that would be the Z-Score) which should be between 1 and 2 2.58).

This way if there is a spike in you favor, the trade would choose the profit quicker, if the market is slow than it'd hit you inactive TP eventually.

Just a thought...