Deadline June

To do's today: more on Gap Filler

*backtest Gap Filler using 1 tick time series for the entries
*optimise Gap Filler for time series on the Friday closing bar (to see if a robust-looking increase/decrease occurs, for all ye curve-fitting spotters)
*create portfolio strategy for all pairs - call it Polyfiller
*import 4th Q tick data from disk trading
*import 1min data from 2000 on for FXCM
*run out-of-sample test unoptimised on non-disktrading pairs 2008-2010


By the way if you had the choice between historical 1 min bar forex data based on the bid or based on the ask, which would you take?

That list button on the wysiwyg editor doesn't do anything, does it?
 
To do's today again: more on Gap Filler

* optimise Gap Filler for time series on the Friday closing bar? I decided to scrap this. This is a red herring and a possible source of curve fitting. The important numbers are the last tick on Friday and the first tick on Sunday.

* optimise the NinjaScript for speed if using 1 tick bars

* backtest Gap Filler using 1 tick time bars

* try with filter to prevent trading small gaps

* try with filter to prevent trading against the trend

* try stop distance = 40% * Fridays true range (only tested using 15min ATR length 2 to 10).

*create portfolio strategy for all pairs - call it Polyfiller

*import 4th Q tick data from disk trading

*import 1min data from 2000 on for FXCM

*run out-of-sample test unoptimised on non-disktrading pairs 2008-2010
 
Last edited:
Mar

meanreversion said:
There's a problem with causality here.

First of all, there are an infinite number of patterns. You don't need to look very hard to find one. I've seen an entire presentation from a trader who spends his entire day looking for patterns, and then assumes they'll repeat.

For example, over the last 10 years, the FTSE has rallied on the second Tuesday of March 85% of the time (I just made that up, but it might be true).

So, next time it's the second Tuesday of March, he'll buy the at-the-money binary at 55% at the start of the day (or whatever the offer is), GENUINELY believing that it is "undervalued". Of course it is not undervalued, as the seller can quite easily hedge it and perhaps lock in the 5% from the true mid, which is 50%.

Read "Fooled by randomness" for a more mathematical explanation of this. You're seeing a pattern, but true randomness itself does not look random, it will have patterns within it that look non-random.

Now, brettus offers another explanation, namely that these gaps are the result of stop hunting and other such activity. I have no idea if this is true, but even if it is, it doesn't tell you where to stop loss or take profit (you'll get this from optimizing, probably). I would imagine the person with the real edge here is the trader in Aus/NZ who can see where the stops are, and just jam them before Tokyo gets in.
Click to expand...


Jacob, I'm sorry that I didn't immediately say at this point in the conversation yesterday that I am aware of all that you said above. I have read all Taleb's books except the arcance options ones and I swear by him, but your reference to him doesn't apply here.

I don't believe my systems will work. I build them on a test period, test them on an out-of-sample period and if the pattern continues, I trade it. End of story. I don't have any beliefs about it. Of course I pray it will work ;)

So sorry for wasting your time and mine. Communication was never my strong point. Or no, actually timing was never my strong point. I never have the ideal response in the nick of time.

That's why I'm not a discretionary trader. I have to delegate the timing to the computers.

While I'm at it, you still haven't revealed what MAR is (see post #740)
 
Technical barriers

NinjaTrader can't backtest this system against 1 minute bars and use the first Open of Sunday evening. In backtesting, when the computer looks at the Sunday Open, it is looking at a OHLC bar, so it knows those 4 points O, H, L & C simultaneously, unlike in real-time trading.

Therefore it doesn't know what would have happened if it placed a market order just after the first tick of the bar. It might have been up, or it might have been down, and unfortunately there's no option to tell NinjaTrader "hey look don't worry about it, just give me the Opening price".

So it has to be tested on tick data. I guess it's logical but it means I can't test it on my IB historical data.

Well, I can test it on 1 minute bars, but the market orders that the historical backtest simulation places are all executed at the Open of the next bar, i.e. 1 minute too late. This is what I've done so far and it appears to be profitable still.

I'm coding it to run on tick data now, but of course the other downside is that backtests will take possibly hours to run, looping the system over all those ticks one-by-one.
 
Re: Mar

Adamus said:
Jacob, I'm sorry that I didn't immediately say at this point in the conversation yesterday that I am aware of all that you said above. I have read all Taleb's books except the arcance options ones and I swear by him, but your reference to him doesn't apply here.

I don't believe my systems will work. I build them on a test period, test them on an out-of-sample period and if the pattern continues, I trade it. End of story. I don't have any beliefs about it. Of course I pray it will work ;)

So sorry for wasting your time and mine. Communication was never my strong point. Or no, actually timing was never my strong point. I never have the ideal response in the nick of time.

That's why I'm not a discretionary trader. I have to delegate the timing to the computers.

While I'm at it, you still haven't revealed what MAR is (see post #740)
Click to expand...

MAR? Managed Account Report. Return/drawdown, I thought we had done this to death?

http://www.commodity-trading-solutions.com/mar-ratio.html

On the other point, just so I'm clear ...

You don't believe your systems will work? Presumably you mean you don't have any beliefs about your system .. ?

How on earth will you be able to persist in trading your system when it hits serious drawdown? I'm struggling to take you seriously here. You seem to be making a virtue of the fact that you have no knowledge of markets or how their participants behave, but hey, it doesn't matter as long as the backtest looks good on out of sample data?

Try to imagine pitching your idea to an investor. Or let's say it's your wife. You'd say "I have no idea why this should work, nor do I care, but I've looked at historical data and it did in the past, so it will going forward" .. do you think this will fly?

I'm a bit lost for words.
 
In addition, you say you are not "discretionary".

Uh huh. So who chose the markets that you trade, who chose the system, who chose the parameters?

Guess what, you did. So as mechanical as your system ends up, there is a huge element of discretionary input. You really are deluding yourself in thinking there is some kind of completely computer driven model solution to trading.

You say you're a fan of Taleb? Did you not read the section where he lampoons backtesting?
 
Yes, but then automation does not exist. Every automated machine was conceived by a human, with a set of parameters and so on. So I would argue that his trading can indeed be defined "automated" and "mechanical" (and mine, too). Or we'd have to say that nothing is mechanical/automated, not even the machines in a factory.
 
Let's just stick with trading only, I'm not sure the analogy is helpful.

A discretionary trader will use rules, a mechanical trader uses discretion.

Adam presents his style as merely identifying a pattern which is repeatable in out of sample, then applying to the market.

This is not enough. If you have no belief behind your system, you'll abandon it once it (inevitably) hits drawdown.
 
yes meanreversion

as well i will say that to know about what your belefs about markets are, you need exposure to them and to watch them in real life, you cant get that by looking at colums of numbers from the past.

most people i think in the beginning start trading without any beliefs or mind training and thats the part under 0 in the J curve... but then they read books and get help from older traders and end up down a few dead ends before finally making beliefs about markets that have started out in their OWN mind that agree with what they see in the markets in real life, that then you can use these to make a trading strategy whatever it is.

you cant get to the upwards bit of the j curve without going through the bottom bit of it... and you cant even get to the bottom bit if you arent involved in the real life markets. it won t come from looking at old statistics.

so adam is not wanting to go through the bottom bit of the j curve and find his beliefs that he can build a system on, but he won't be able to build a system to trade until he has some beliefs to attach to it. it goes round and round, like pinning the tail on the donkey when no one has decided which end is the head and which end in the ass! LOL

sorry Adam im not trying to be nasty on your blog, i dont mind if you delete this one as well.
 
Just some thoughts, with no intention to get into an argument.

Adamus and I have been trading since the 1990s (discretionary and automated) so it's not like we haven't observed the markets, because it would be impossible not to peak at them even by mistake.

Another thought: are we sure that looking at the live markets is much different than looking at them once that action has become recorded history? I think you can learn about how the markets move even if you look at their recorded history, rather than just live (maybe even more). Just a thought, because you might have meant something different.
 
Philosophy

meanreversion said:
On the other point, just so I'm clear ...

You don't believe your systems will work? Presumably you mean you don't have any beliefs about your system .. ?

How on earth will you be able to persist in trading your system when it hits serious drawdown? I'm struggling to take you seriously here. You seem to be making a virtue of the fact that you have no knowledge of markets or how their participants behave, but hey, it doesn't matter as long as the backtest looks good on out of sample data?

Try to imagine pitching your idea to an investor. Or let's say it's your wife. You'd say "I have no idea why this should work, nor do I care, but I've looked at historical data and it did in the past, so it will going forward" .. do you think this will fly?

I'm a bit lost for words.
Click to expand...

You've done quite well on my journal for someone who's lost for words. Look, if you are going to lambast me, it's only fair you give me a definition of what you are talking about beforehand. I'm not into philosophy as far as trading goes and when you start talking about beliefs I just turn off.

If that is what you mean by belief, then of course I have my beliefs for developing and implementing and trading a system, but can't we call them my reasons?

(Just adding this here afterwards, when most of my best thoughts occur to me) How about this: I believe I have an edge. I don't believe I'll win without fail because my funds and my time to do research are both limited and I have to take risks.

As far as my ability to stick with a trading system is concerned, then without the rationale that you require I have been able enough to stick through two serious drawdowns now without chucking in the systems I'm using.

I think what's going on here is that you're concerned over my actions because you've had relative long exposure to the story through reading this journal, and as that has involved you, you are trying to transfer your mental framework onto me, the actor of this story, with all the subconcious requirements that you have built in - e.g. a reliance on stats from the investment world like CAGR and Sharpe Ratio - but your mental framework doesn't fit my activities and is partly obscure and irrelevant to me anyway.

Please don't let my attitude put your nose out of joint. It's most likely just an artifact of communication on the internet - a bit like shooting fish in a barrel.
 
Last edited:
DashRiprock said:
yes meanreversion

as well i will say that to know about what your belefs about markets are, you need exposure to them and to watch them in real life, you cant get that by looking at colums of numbers from the past.

most people i think in the beginning start trading without any beliefs or mind training and thats the part under 0 in the J curve... but then they read books and get help from older traders and end up down a few dead ends before finally making beliefs about markets that have started out in their OWN mind that agree with what they see in the markets in real life, that then you can use these to make a trading strategy whatever it is.

you cant get to the upwards bit of the j curve without going through the bottom bit of it... and you cant even get to the bottom bit if you arent involved in the real life markets. it won t come from looking at old statistics.

so adam is not wanting to go through the bottom bit of the j curve and find his beliefs that he can build a system on, but he won't be able to build a system to trade until he has some beliefs to attach to it. it goes round and round, like pinning the tail on the donkey when no one has decided which end is the head and which end in the ass! LOL

sorry Adam im not trying to be nasty on your blog, i dont mind if you delete this one as well.
Click to expand...

Dash

what on earth is the J curve?

I know, I should google it.

Oh I see it's just a J-shaped curve.

My equity is more like a W at the moment though.

I like your analogy about a donkey. It will however go down on your permanent record.
 
You must log in or register to reply here.

Similar threads

S
Conducting and Evaluating a Forward Test
Replies
4
Views
5K
selfmade
S
L
My remarkable record
2
Replies
27
Views
4K
Tenapenny
Tenapenny
M
marsman's Project X
5 6 7
Replies
125
Views
25K
itspossible
itspossible
J
  • Locked
Joey's MP Journal
13 14 15
Replies
299
Views
36K
Joey25
J
danfreek
danfreek's Blog
2
Replies
27
Views
4K
danfreek
danfreek
Top