[BlowFish]

Originally Posted by Tams »

delay?

no worries... try Moving Average with extension

http://www.traderslaboratory.com/for...sion-5642.html

126.83

That won't speed up the crossover it will just tell you when it might happen?

[razzbarry]

The point I was trying to make is demonstrated well by your chart. Going long at the point where the MA line first turns blue will be up the third bar of the trend. That may not be too bad on this trend but look where it gets out and goes short. You would lose 50% of the gain on the long.

Then you get whipsawed on the short with a large loss. Then you get whipsawed more on the next long, short, long. But the end of those trades you would have a large loss on your hands. I have a system that makes 25,000% on a static back test. The system is based on MACD and two EMAs. When we run it with a live data feed using one contract RUT e-,ini, it may make $4000 to $8000 one day and down that much the next. Overall it loses money. So much for back testing, I am not a fan of BT. A lot of this is due to whipsaw when the market turns. But the rest and worst is the amount it loses with price action exactly as your chart shows.

What I have been trying to develop is a set of adaptive trend lines that automatically follow the bottoms on longs and tops on shorts. When the market turns the system changes directions. To see what I mean draw a tend line or your charts you will see it gets you out very close to the turns. At times it will get out as the market turns into a sideways move. Looking at the exit point will show the exit, though early, gives a good return. Once I get those running I am going to try to add the VSA rules and then add the support / resistance lines. That would be more of a pattern recognition method. I have tried this with caned data with AmiBrokers BarReplay and it seems to work.

Barry

[Tams]

MA alone is not going to cut it...
if you are going to use MA, you have to look deeper than the superficial cross over.

this might give you a bit more insight.
http://www.traderslaboratory.com/for...ombo-5851.html

100

[phase21]

Hello Razz,

Originally Posted by razzbarry »

So much for back testing, I am not a fan of BT.

You might want to have a look at : Quantitative Trading Systems

The book is an excellent resource for trading ideas, and especially the methodology for "walk-forward" testing, which has been integrated into AmiBroker.

Regards,

[razzbarry]

Thanks for the suggestion. I have Bandy's book and have read it a number of times.

The problem with back testing and walk forward testing is that it uses static data as opposed to live data. When designing systems for use with live data things happen during bar formation that static testing can't see. Of course you can prevent intra bar errors entering a system by trading on the close, actually the open of the next bar. But with futures the price can move so quickly that waiting to the next bar causes too much gain to be lost. Recently the TF contract has been moving 2 to 6 points in a bar early in the day. Would you really want to wait until the next bar? I don't.

A feature in later AmiBroker is BarReplay. I have designed programs to use bar replay with tick and minute data. Analyzing a program in this fashion allows you to see how it will operate with live data. Using tick data you can see all the false signals that happen mid bar. That is a real eye opener on many of the indicators. They do things that are totally unexpected during bar formation. When using indicators but allow them to trade during a bar rather than at the close opens a whole bucket of worms that are invisible using static data. All the transient signals are invisible but can send a hundred signals, buy and sell, during a bar when the indicators are fluttering, such as MAs crossing up and down when the ticks are up or down a sufficient amount to cause a cross. Most indicators rely on a comparison with the last full bar. When using cross or such, if the system goes long and then reverts to short during a bar the system will act and go long but has no way to reverse itself since the cross function will not give a short signal during the bar since it was sort and still thinks it is.

Only a system that can simulate live data can find errors like this. BarReplay is one tool that will allow analysis of the system with simulated live data. Bandy's book does not cover this type of environment or the testing of it. If you can, set up an Ami data base to capture tick data. Then play that back into your system on whatever time frame and watch your trading system. It will blow your mind.

Barry

[bootstrap]

i don't mean to sound like a broken record here, but what does a MA cross represent? PRICE. but i am going to leave that alone for now.

i am going to be the first to say, if you think you can trade it and it works for you, "GREAT and GOOD for you." i actually know a trader that has MA's as part of what they do.

but think about what you are wanting the cross to tell you. you are wanting buy or sell. and while it can be that easy, what made you take the signal? if you say the cross, then you have merged two parts of a strategy.

SETUP and ENTRY. they are not the same.

either a) the cross is the entry and something occurred beforehand or b) the cross is merely the setup and something else has to occur before you take the signal.

[razzbarry]

Broken record is good and I appreciate what you are saying. I have been looking at this from a specific viewpoint for so long it is healthy to have someone to challenge what I am doing. I will take some time and try to look at it differently. Actually I had already started doing that.

Thanks,
Barry

[phase21]

Razz,

Have you selected the most appropriate interval per bar? If your system is currently based on 10min intervals, have you tried it on say 5 or 3min bars, with/out adjustment to the number of periods in you MA?

Have you looked at the time of day that you are trading - if you look at the volume and price movements on the TF, you'll see that most of th action takes place 8:30-11:00am and 2:00-4:00pm. The rest is just so choppy, unless there's been a major reaction to an announcement.

Have you looked at different data for the bars, eg using a fixed number of ticks per bar will give you whole different picture, than a fixed number of minutes.

Using tick data you can see all the false signals that happen mid bar. That is a real eye opener on many of the indicators. They do things that are totally unexpected during bar formation.

Having traded YM, TF, and ES in real-time, I understand what you mean about intra-bar formations.

There was a time when I thought that reacting to intra-bar price movements was a good thing, but came to realize that it was a waste of time - I'd invariably get it wrong: bad fill, wrong direction, etc.

But, do you really want to react to thousands of (false) signals that occur during a bar, or, do you want to decipher and trade the main direction. Will you MAs give you the main direction? Is there something else that you need to use in conjuction with them?

I agree, Bandy's book doesn't cover real-time.

One of the things he does talk about though, is the difference between signal and "noise" - we want to trade the signal. If your signal is sound, and has statistical significance (which he outlines in QTS), then proceed to walk-forward testing, and then trade it if it still stands up.

Regards,