Automatic Pattern Search

Price Action

Is the creation of a 'pin bar' or 'inside bar' a price pattern as far as APS goes? If so, can APS be used to find these daily or intraday?

Many thanks, S
 
Is the creation of a 'pin bar' or 'inside bar' a price pattern as far as APS goes? If so, can APS be used to find these daily or intraday?

Many thanks, S


Yes, I think so, actually in their website example they have an inside bar break-out pattern:

Tradingpatterns.com Michael Harris article

I get tons of patterns including inside bars, outside bars, key reversals, double bottoms and many other interesting formations. Of course the problem is whether they will continue being profitable in the future. If you use price patterns I think you must always use combinations and some type of verification to increase the probability of success.
 
Patterns is not a panacea although I prefer them over indicators and it is difficult to select which ones will stay profitable but I agree that APS can find many but you must study their performance carefully before using them.

Alex
 
Patterns is not a panacea although I prefer them over indicators and it is difficult to select which ones will stay profitable but I agree that APS can find many but you must study their performance carefully before using them.

Alex

Have you figure out a way to estimate the probability of future performance of a pattern other than forward testing it?

I have noticed when I get several signals from different patterns quite often this is a good indication of a profitable trade but I cannot think of an easy way to backtest this.
 
Have you figure out a way to estimate the probability of future performance of a pattern other than forward testing it?

I have noticed when I get several signals from different patterns quite often this is a good indication of a profitable trade but I cannot think of an easy way to backtest this.

In his first (and second) book, Michael Harris covered his "P indicator", which was not followed upon on his latest one.
This indicator is based on just that, the number of patterns, and used to take a trade on the market direction where the indicator points, long or short.
He backtested it on Tradestation.
Is this what you are after?

Eduardo.:)
 
Have you figure out a way to estimate the probability of future performance of a pattern other than forward testing it?

I have noticed when I get several signals from different patterns quite often this is a good indication of a profitable trade but I cannot think of an easy way to backtest this.

You can backtest this fairly easily. If you have many different patterns create a counter of position signals that arrive at the same time, for example Nsignals. Then use that in the following sense:

If Nsignals > Nsignals_min
then open position

In other words, you can find the best value (if it exists) for the counter. If what you say holds, the equity increase should be monotonic with increasing Nsignals.

Actually, I am trying to do this backtest for several currency pairs, futures and stocks. I have more work to do than I can handle:cheesy:
 
In his first (and second) book, Michael Harris covered his "P indicator", which was not followed upon on his latest one.
This indicator is based on just that, the number of patterns, and used to take a trade on the market direction where the indicator points, long or short.
He backtested it on Tradestation.
Is this what you are after?

Eduardo.:)


Eduardo thank you very much. This sounds very interesting. Very different kind of an indicator. Have you tried it? I wonder, how difficult it is to program something like that?

Is there a link to any information on this indicator besides the books?
 
"Have you tried it? I wonder, how difficult it is to program something like that? " - No I haven't as I use other systems to assess the trend; but there is a study on its performance in M.Harris first book using TradeStation.
I suppose that the only way to program this indicator is the hard way, running explorations for every trading day and calculating it from these results.
"Is there a link to any information on this indicator besides the books?" - I haven't found any; in fact M.Harris is strangely silent about it in his latest book.
You could try to contact M.Harris directly and raise the subject with him, or speak with their sales dpt. and see what they say.
I feel that there is a lot of potential for such an indicator, but quite frankly, with the latest developments in my personal life, I haven't got the time to do such studies myself.
Maybe some of you would like to explore the matter further along?

Eduardo.:)
 
Eduardo, Bill,

I struggled with replicating the p indicator for a while (nice name by the way, it denotes probability p). It is true that for some reason Michael Harris is secretive about it after his second book. I concluded though that you can simulate the performance of the indicator if you follow these steps:

(1) Use APS to find patterns for a specific instrument. You need a good number of them.

(2) Monitor the signals generated by the patterns

(3) If you get more than one signal then P = sum(hist. trades x hist. win rate)/sum(hist. trades) over Ns, the number of signals.

(4) If you get both long and short signals, then the win rate of the short ones is a negative number

(5) You then look at the P value AND the number of signals (Ns). For example, you may decide based on historical testing that it is best to trade when P > Pmin AND Ns > 3.

This has worked for me and it still works well but you must do your homework and set your parameters to the right values.
 
Eduardo, Bill,

I struggled with replicating the p indicator for a while (nice name by the way, it denotes probability p). It is true that for some reason Michael Harris is secretive about it after his second book. I concluded though that you can simulate the performance of the indicator if you follow these steps:

(1) Use APS to find patterns for a specific instrument. You need a good number of them.

(2) Monitor the signals generated by the patterns

(3) If you get more than one signal then P = sum(hist. trades x hist. win rate)/sum(hist. trades) over Ns, the number of signals.

(4) If you get both long and short signals, then the win rate of the short ones is a negative number

(5) You then look at the P value AND the number of signals (Ns). For example, you may decide based on historical testing that it is best to trade when P > Pmin AND Ns > 3.

This has worked for me and it still works well but you must do your homework and set your parameters to the right values.

Thanks guys,

The company responed with this link.

Very interesting indicator.
 
Eduardo, Bill,

I struggled with replicating the p indicator for a while (nice name by the way, it denotes probability p). It is true that for some reason Michael Harris is secretive about it after his second book. I concluded though that you can simulate the performance of the indicator if you follow these steps:

(1) Use APS to find patterns for a specific instrument. You need a good number of them.

(2) Monitor the signals generated by the patterns

(3) If you get more than one signal then P = sum(hist. trades x hist. win rate)/sum(hist. trades) over Ns, the number of signals.

(4) If you get both long and short signals, then the win rate of the short ones is a negative number

(5) You then look at the P value AND the number of signals (Ns). For example, you may decide based on historical testing that it is best to trade when P > Pmin AND Ns > 3.

This has worked for me and it still works well but you must do your homework and set your parameters to the right values.


Thanks Ronblack, good effort.
Much appreciated; keep up the good work!

Eduardo.:)
 
I'm backtesting something along the lines of the p indicator and I am getting some very good results. There seems to be a lot of potential in this concept.
 
Does anyone know of a way to run 2 instances of APS on a dual CPU machine? Could a virtual desktop do that? Has anyone tried it?
 
If anyone would like to sell his APS license please send private message.

You again? Really, you should be banished from this forum!
By the way, I could sell you my copy if you want - but don't blame me if it doesn't work!
I think it has already been explained why, so why don't you stop posting here? This is for serious people.

Eduardo.
 
jimbotrader,

As Edurado said, it has been already explained to you that trying to get someone to sell licensed software is not a good idea. I think that if you have determined that you like APS you should spend the money and purchase a license. My opinion is that this software is very cheap for what it does. If you know what you are doing with patterns you can probably recover the cost after a few trades. No software can help you if you do not know how to use it properly. Do not think that any software can make money for you if you do not understand how to fit within a larger scheme that involves a complete trading strategy.
 
They will also include code generation for MetaTrader/Ninja/Amibroker. Good work. I just checked the website and they have examples of the new codes:

APS Codes
 
FYI, A new version with the p Indicator was released today. I can't find any information how to upgrade to that. It appears this new version is not compatible with older ones.
 
Top