Out of sample testing of patterns

[Code7]

Quote from goodgoing:

Which random set of all possible random sets? You can find a random set that outperforms any system and one that underperforms it. This is the reason people use a benchmark of some kind, like free risk return, the return of some index or another benchmark portfolio. I understand you are probably a programmer and you have never taken a course in finance but think before you spew crap.

How about you try to read and understand first what I wrote, then reply? In the article, 8 long-only patterns were forward-tested from 05/07/2002 to 08/22/2008. With daily QQQQ data, trades were entered on the open. That's all your possible trades, about 1562.

The job of the patterns was to select out of all possible trades. The patterns are independent, so one day can be selected 8 times at max. They did a poor/meaningless job because a large number of random trade selections averages to the very same performance.

 

[intradaybill]

Quote from Code7:

How about you try to read and understand first what I wrote, then reply? In the article, 8 long-only patterns were forward-tested from 05/07/2002 to 08/22/2008. With daily QQQQ data, trades were entered on the open. That's all your possible trades, about 1562.

The job of the patterns was to select out of all possible trades. The patterns are independent, so one day can be selected 8 times at max. They did a poor/meaningless job because a large number of random trade selections averages to the very same performance.

Do you know what the trading cost of a large number of random trade selections averages is?

This is what you do: there is a biased coin, 60% heads. Someone claims to have a system that averages 60% in 30 say runs. You say that if you toss 1562 times you will get the same result. You forget the tossing cost, which exceeds the profit of the system.

This is how you done that from another thread:

Quote from Code7:

I used daily QQQQ data from TradeStation and this code for my test.

Inputs: Num(1);
Vars: EP(0);

if CurrentBar = Num then
	Buy next bar on Open;

if CurrentBar = Num + 1 then
	EP = Open;

if CurrentBar > Num then
begin
	Sell next bar at EP * 1.07 Limit;
	Sell next bar at EP * 0.93 Stop;
end;

I raised Num from 1 to 2000 by 1 to buy on all days and then looked up the results by sorting the optimization report. So in essence, I just weighted all days equally without any bias whatsoever. Random entries will average to the same winning %.

For how long are you going to try to convince everyone that you do not understand anything about real trading?

Keep on...dig hard...

 

[goodgoing]

Quote from Code7:

They did a poor/meaningless job because a large number of random trade selections averages to the very same performance.

How large? Can you give the number of trades? Or are you just speculating that this is the case?

 

[Code7]

Quote from intradaybill:

Do you know what the trading cost of a large number of random trade selections averages is?

This is what you do: there is a biased coin, 60% heads. Someone claims to have a system that averages 60% in 30 say runs. You say that if you toss 1562 times you will get the same result. You forget the tossing cost, which exceeds the profit of the system.

All 1562 possible trades have the average expectancy of 60.5 % success rate with both stop and target 7 % away from entry. That's a profit factor of about 1.53, assuming avg_win/avg_loss ratio of 1. That's the average expectancy for every single possible trade you goddam idiot.

That also tells me the avg_trade in terms of $ has to be positive before slippage and commission. If you look at trading costs as fixed $ amount per trade per QQQQ, it doesn't even matter how few or many trades you have. Of course, the average expectancy per trade will be lowered by trading costs but that's also the case for Harris' patterns, his avg_trade will be lowered exactly the same way you imbecile. So if my costs per trade exceed my avg_trade, Harris' system will also have negative expectancy.

Quote from goodgoing:

How large? Can you give the number of trades? Or are you just speculating that this is the case?

You can't even read, I wrote large number of selections, not trades. The number of trades per selection doesn't even matter here. The overall expectancy of (many) random trade selections will always average to profit factor 1.53. That's a mathematical fact and doesn't have to be shown empirically.

I'm done with you fools now.

 

[intradaybill]

Quote from Code7:

That's the average expectancy for every single possible trade you goddam idiot.

Goddam idiot? Is that how you talk home to your kids and wife? Do you know anything about statistics other than insulting people? Because the law of large numbers claims that a fair coin frequency will average to 50% heads it does not mean that every random sample of tosses will average to 50% heads.

You are a psycho and you are now on ignore. Get a life, learn how to talk in a civil way and learn statistics.

 

[intradaybill]

Quote from Code7:

All 1562 possible trades have the average expectancy of 60.5 % success rate with both stop and target 7 % away from entry. That's a profit factor of about 1.53, assuming avg_win/avg_loss ratio of 1. That's the average expectancy for every single possible trade you goddam idiot.

Something else psycho. Expectancy is neither a profit factor nor a winning rate. Expectancy is an average expected profit per trade that applies only to the limit of very large samples and it is equal to E = avgwin x win rate - avgloss ( 1 - win rate).

You are talking about all that nonsense because you know nothing about statistics and real trading. Small trade samples cannot be compared to large trade samples. Small samples can only be evaluated based on forward testing, actual or simulated. You cannot compare a system with 30 trades in 6 years to a hypothetical system with 2,000 random trades where you calculate the average expectancy. Out of the many combinations there are many that are losing. I say it again: you cannot average 2000 random trades you idiot and then compare the result to a system that had 30 trades. Anyone can tell you here that there is a hidden commission cost you have not taken into account.

Idiot...

 

[goodgoing]

Quote from Code7:

You can't even read, I wrote large number of selections, not trades. The number of trades per selection doesn't even matter here. The overall expectancy of (many) random trade selections will always average to profit factor 1.53. That's a mathematical fact and doesn't have to be shown empirically.

HI imbecile. Here are two selection of trades for R:R = 1:

(1) L L L L L L L L (L = Loss)

(2) W W W W W W W W W W W W ( W = Win)

The average success rate is 60%

which selection (system) do you prefer? (1) or (2)?

How can you be such an imbecile. It is really horrible to be that way.

 

[Code7]

Quote from intradaybill:

You cannot compare a system with 30 trades in 6 years to a hypothetical system with 2,000 random trades where you calculate the average expectancy. Out of the many combinations there are many that are losing. I say it again: you cannot average 2000 random trades you idiot and then compare the result to a system that had 30 trades.

Just wanted to point out that you LIED again. First you claimed buy-and-hold return was almost 0 while it was in fact 62% during the out-of-sample period in the article you quoted. Now you claim I compared 30 to 2000 trades. That's again plain wrong. Harris' 8 long-only patterns selected a total of 128 trades, 77 were profitable. That's exactly the same performance in terms of success rate and profit factor all possible 1562 trades have on average already.

But you still insist with more and more lies what great achievement it was to select those 128 out of 1562. Even though nothing was added to overall performance through that selection. If one just randomly chose 128 trades without any APS patterns, results on average would be exactly the same. You tried hard to confuse that fact away but it didn't work.

 

[intradaybill]

Quote from Code7:

Just wanted to point out that you LIED again. First you claimed buy-and-hold return was almost 0 while it was in fact 62% during the out-of-sample period in the article you quoted. Now you claim I compared 30 to 2000 trades. That's again plain wrong. Harris' 8 long-only patterns selected a total of 128 trades, 77 were profitable. That's exactly the same performance in terms of success rate and profit factor all possible 1562 trades have on average already.

But you still insist with more and more lies what great achievement it was to select those 128 out of 1562. Even though nothing was added to overall performance through that selection. If one just randomly chose 128 trades without any APS patterns, results on average would be exactly the same. You tried hard to confuse that fact away but it didn't work.

No, I did not try to confuse anything. You are the confused one, a very compulsive and rude individual.

When I said 30 trades I was referring to the 27 trades of the second out_of_sample test.

Regardless, someone else already pointed out the fact that you know no statistics and you are totally confused about what you are doing. You have elected to ignore the issue of commissions for large trade samples that would significantly skew your average results towards 0 or even negative expectancy.

The conclusion is:

(1) You do not understand statistics

(2) You ignore real trading effects such as commissions

(3) You are rude and vulgar

(4) You consiously obscure issues that do not fit your agenda.

Do you understand of course what (1) - (4) mean about you? A very ugly picture.

 

[Code7]

Quote from intradaybill:

You have elected to ignore the issue of commissions for large trade samples that would significantly skew your average results towards 0 or even negative expectancy.

You just lied again.

Quote from Code7:

If you look at trading costs as fixed $ amount per trade per QQQQ, it doesn't even matter how few or many trades you have. Of course, the average expectancy per trade will be lowered by trading costs but that's also the case for Harris' patterns, his avg_trade will be lowered exactly the same way you imbecile. So if my costs per trade exceed my avg_trade, Harris' system will also have negative expectancy.