Deadline June

[meanreversion]

Re: causality

I only believe in patterns if i can understand the trading mentality behind them. I could easily make up a pattern that works, only for price to do the exact opposite the next year. I think the problem with systems that most people dont realise is that there is probably 1million systems that have worked backtested.

Precisely. Actually suspect more like 1 billion systems tested.

To believe in a system, there needs to be a rationale. Then it also helps if there is some evidence that the OPPOSITE of the system doesn't work.

 

[meanreversion]

For example, trends exist, we have seen it in the past. BUT that is no guarantee they will continue in the future. So, is there evidence that contrarian trading loses money? And so on.

Simply finding a pattern is meaningless; the optimizing tool helps you do that in seconds.

 

[Adamus]

Re: more causality

For example, trends exist, we have seen it in the past. BUT that is no guarantee they will continue in the future. So, is there evidence that contrarian trading loses money? And so on.

Simply finding a pattern is meaningless; the optimizing tool helps you do that in seconds.

I would say, the optimisation tools help you destroy the pattern that you are looking at.

I agree there must be a reason behind every pattern that works, but I don't need to know it to be able to trade it. Knowing the causality would just be the icing on the cake.

There is another public system I want to program, called One Night Stand. It places orders on Friday night in the direction of the trend before the close and seeks to profit from the movement into the close and over the weekend, exiting on Sunday at the Open. I wonder if you would be satisfied by the 'explanation' given for that (by the author Rensink of a PDF outlining the system)

As you can see from the equity curves displayed for the various currencies, the system has been
working very well since it was released in the late '80's. What is not usually appreciated by those
investigating this system is this--- these returns are accomplished staying in the market less than
15% of the time!

Interestingly, a large number of important government reports come out on Fridays, and have
for many years. That probably helps the persistence of this trade too. I do believe the main
reason One Night Stand still works is for the reasons Joe Krutsinger gave many times from the
platform.

Individuals and banking institutions are reluctant to stay heavily positioned over weekends if
they don't have to. And they don't have to. So, traders like us pick up a risk premium for doing
what others are unwilling to do.

 

[Adamus]

re: Gap Filler

Simulated real-time results:

$AUDCAD	-10.06
$AUDCHF	120.00
$AUDJPY	-6.05
$AUDUSD	90.00
$CADCHF	0.00
$CADJPY	-90.82
$CHFJPY	133.23
$EURAUD	168.15
$EURCAD	-117.74
$EURCHF	15.65
$EURGBP	-24.00
$EURJPY	90.74
$EURUSD	30.00
$GBPAUD	187.95
$GBPCAD	0.00
$GBPCHF	41.76
$GBPJPY	72.66
$GBPUSD	0.00
$USDCAD	-109.68
$USDCHF	15.66
$USDJPY	-234.40
	373.07

Brett, I guess you had no trade?

Examining the results throws light on a few problems with the system.

First, it occurs to me that the size of the gap is inversely proportional to bar period. I started off using 60 min bars, then moved to 15 min bars, then 1 min bars and the turn-over and the number of trades both increased.

Following that to its logical limit, using 1 tick bars would mean that there is a trade in every instrument every weekend.

Hopefully optimisation of the time series length will display a good progression of profitability between the extremes.

What also emerged from the trial is that NinjaTrader goofed again - or rather NinjaTrader and IB failed to co-operate again. The USD/CAD trade was broken off and closed out automatically by NinjaTrader for reasons currently unknown. I think what happened is that IB gave me a partial fill that was less than IB's own 'odd-lot' size limit, and NinjaTrader automatically reacted, as it should, by placing the stop and target orders for the position but since the lot size was below the 'odd-lot' threshold, IB routed the orders differently and NinjaTrader couldn't follow them, and thought they were rejected.

 

[DashRiprock]

morning everyone

I agree with these last few comments very much, about their being having to be a rational behind a strategy. otherwise you can data mine a system just like that!

One think i'll say though is that the reasons don't need to be complicated, but they should at least be sensible. Sensible reasons are like what is in "madness of crowds" and "people don't change" and so on. Bad reasons are like trading from astrology or palm reading

another thing is that you should heavily believe in your rational, because if you start to doubt your strategy trading is a very hard business!

 

[DashRiprock]

adamus have you read the book by Toby Crabel? If you havent I would recommend it.

 

[meanreversion]

Let's take a step back here.

Adam, what metrics do you use to compare systems?

For example, system A and system B. What do you need to know about them (in quantitative, not qualitative terms) in order to decide which one you will trade?

 

[DashRiprock]

Re: more causality

....

I agree there must be a reason behind every pattern that works, but I don't need to know it to be able to trade it. Knowing the causality would just be the icing on the cake.

...

Sorry Adamus but I think you are a bit wrong here.

What I mean is, no, you don't need to know for sure what the real reason behind a move is. I mean, you might trade a breakout system on stocks based on the turtle rules, without realising every stock you traded long blew out their earnings estimates. Trading isn't like maths in school where you get marks for "showing your working" even if you are wrong or lose marks for just being lucky

but the reason I say something is that if you don't have a reason to believe that your strategy can work other than the backtest results, it can get very difficult to carry on trading it when you in a long drawdown.

sorry if im interrupting the thread

 

[Adamus]

Re: Toby Crabel

adamus have you read the book by Toby Crabel? If you havent I would recommend it.

No I haven't, but I heard that it is out of print and very difficult to get hold of. I searched the web for a PDF of it but couldn't find one. You don't have it, do you?

 

[Adamus]

Let's take a step back here.

Adam, what metrics do you use to compare systems?

For example, system A and system B. What do you need to know about them (in quantitative, not qualitative terms) in order to decide which one you will trade?

Net profit, profit per trade, number of trades, drawdown, average profitable trade, average loss, max profitable trade, max loss, MFE, MAE, and a couple of others. Above all I like to see the equity curve.

 

[DashRiprock]

Re: Toby Crabel

No I haven't, but I heard that it is out of print and very difficult to get hold of. I searched the web for a PDF of it but couldn't find one. You don't have it, do you?

no sorry i don't, but I'm pretty sure there will be copies in the internet. The full title is "Day Trading with Short Term Price Patterns and Opening Range Breakout" by Toby Crabel.

I think the specifics might not apply nowadays but things like opening range and narrow range breakouts are still being traded today.

 

[meanreversion]

Ok, all you've done is list a bunch of output parameters without telling me quantitatively what you're looking for.

You say 'above all, the equity curve'. Are you able to plot the equity curve for the entire portfolio?

If system A has a higher net profit than system B, is it better?

 

[DashRiprock]

Re: Toby Crabel

No I haven't, but I heard that it is out of print and very difficult to get hold of. I searched the web for a PDF of it but couldn't find one. You don't have it, do you?

How hard did you look???

I found it on scribd already!

http://www.scribd.com/doc/41329694/Crabel-Toby-Day-Trading-With-Short-Term-Price-Patterns-2

anyway here is a link so happy days :cheesy:

 

[brettus]

Re: Gap Filler

Simulated real-time results:

$AUDCAD	-10.06
$AUDCHF	120.00
$AUDJPY	-6.05
$AUDUSD	90.00
$CADCHF	0.00
$CADJPY	-90.82
$CHFJPY	133.23
$EURAUD	168.15
$EURCAD	-117.74
$EURCHF	15.65
$EURGBP	-24.00
$EURJPY	90.74
$EURUSD	30.00
$GBPAUD	187.95
$GBPCAD	0.00
$GBPCHF	41.76
$GBPJPY	72.66
$GBPUSD	0.00
$USDCAD	-109.68
$USDCHF	15.66
$USDJPY	-234.40
	373.07

Brett, I guess you had no trade?

Examining the results throws light on a few problems with the system.

First, it occurs to me that the size of the gap is inversely proportional to bar period. I started off using 60 min bars, then moved to 15 min bars, then 1 min bars and the turn-over and the number of trades both increased.

Following that to its logical limit, using 1 tick bars would mean that there is a trade in every instrument every weekend.

Hopefully optimisation of the time series length will display a good progression of profitability between the extremes.

What also emerged from the trial is that NinjaTrader goofed again - or rather NinjaTrader and IB failed to co-operate again. The USD/CAD trade was broken off and closed out automatically by NinjaTrader for reasons currently unknown. I think what happened is that IB gave me a partial fill that was less than IB's own 'odd-lot' size limit, and NinjaTrader automatically reacted, as it should, by placing the stop and target orders for the position but since the lot size was below the 'odd-lot' threshold, IB routed the orders differently and NinjaTrader couldn't follow them, and thought they were rejected.

Yea it was a no trade for me. usd was stronger at 9pm, but at 10pm, usd was weaker, so where at 9 i would have had potential buy if it was slightly lower. You had a sell. Given this information, you still would have won, and so would i if i bought. But i think the 9pm trade has backtested better.

 

[Adamus]

Re: causality again

Ok, all you've done is list a bunch of output parameters without telling me quantitatively what you're looking for.

You say 'above all, the equity curve'. Are you able to plot the equity curve for the entire portfolio?

If system A has a higher net profit than system B, is it better?

That's because I can't tell you quantitatively what I'm looking for. You may see just a bunch of output parameters, but what do you do when they are all similar? I weigh them off against each other. I guess I prioritise them with profit per trade and drawdown being the most important, but they all count and if you said I was only allowed one thing, that would be the equity curve.

Yes I can plot the equity curve for a portfolio.

If system A has a higher net profit than system B, then - ceteris paribus - yes it is better. Why do you ask? Don't make the mistake of thinking that this is all I do.

Understand that the non-linear dynamics involved in market action do not necessarily allow you to divine the causality of any particular exploitable phenomenon. You obviously already understand non-linear dynamics, you just need to put 2 and 2 together to realise the validity of this approach. Or look at the profits.

I wouldn't have touched this gap filling system a year ago, but since then my way of looking at things has changed.

I don't need to educate you and you've no doubt heard it all before, but I am always open to being educated myself - I need some follow-up to flip me out of my current non-linear but stable state.

 

[Adamus]

Re: Toby Crabel

How hard did you look???

I found it on scribd already!

http://www.scribd.com/doc/41329694/Crabel-Toby-Day-Trading-With-Short-Term-Price-Patterns-2

anyway here is a link so happy days :cheesy:

Do you see any text beyond page 2? I don't see anything. I googled before. I had set the search options to format: pdf so perhaps that excluded the link you found. However I think it's corrupt anyway.

Sorry forget that! It's just my browser had security settings too high. It's all there. Well, at least the first 5 pages.

Argh, can only read it online. Oh I see I can buy it. What is scribd? I never heard of it before.

 

[DashRiprock]

"That's because I can't tell you quantitatively what I'm looking for"

sorry but if you haven't got a target return and volatility / max drawdown in mind, how do you know when to stop looking?

also the scribd thing, i looked and you can upload something instead. But really, you should be able to find it somewhere. it took about 15 seconds to find that one, there must be more.

 

[meanreversion]

Right, allow me to make a few observations. If you want me to leave this thread, please tell me, but I'd like you to consider the following.

1. You've been working at this some time now, and don't seem to have settled on a system. In my experience, the main reason that people can't find a system after extensive backtesting is because they don't have quantitative metrics to compare systems. At no stage have you mentioned return divided by drawdown, or Sharpe (return divided by vol of returns). I fail to see how you can compare systems without using one of these two metrics.

(You mention profit per trade and drawdown in the same sentence.. this is not the same as CAR/MDD, because you have not included frequency of trading, i.e. you aren't incorporating a time element.)

I don't believe you can confidently choose between systems in a short space of time - you need to fix this.

2. Looking at the equity curve is important to you, BUT your eye will see what it wants to. If you're testing something and are quite keen on it, you'll look at the curve and see how it's always rising. An investor might look at it and see all the dips that your brain skipped over.

3. Higher profit does NOT mean a better system. If system A has a return of 60% a year with a drawdown of 40%, this is WORSE than system B, which returns 20% a year, with a drawdown of 10%. Why? Because all you need to do is trade system B in three times the stake size, and hey presto, it's better than system A. This confirms my belief that you're not completely clear in your own mind on how to compare systems quantitatively.

4. You will not be able to trade a system unless you believe in it, and for you to have belief, you need to have a rationale, or a theory, or something. Simply finding a pattern and assuming it will continue is completely bogus.

DashRiprock has made comments along similar lines. My advice to you now would be

- be clear in your mind precisely how to quantitatively compare systems
- trade your beliefs.. what do you believe about the market and the way it acts, and design a system around this
- please forget this gap system (GBP/AUD gaps - are you serious?)

 

[Rossini]

^^^^^^I think that's good advice.

 

[Yamato]

I want to add this. If you see that one system is better than another based on the metrics he's suggesting, this doesn't mean you should discard the worse system. You should keep it and trade them both. Diversification is also important. If you don't have enough money to trade them both, then of course you trade the best one.