After simulation & Testing - How do you start trading a

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
Post Reply
Austrian
Roundtable Knight
Roundtable Knight
Posts: 223
Joined: Sat Oct 25, 2003 12:35 pm
Location: Austria, Vienna

After simulation & Testing - How do you start trading a

Post by Austrian » Thu Oct 14, 2004 3:20 am

After simulation & Testing - How do you start trading a system??
Let us assume the testing and simulation is done and you decide to trade the system?

How would you start?

Taking only new signals, even knowing the system is long term and recent simulated trades are still open?

Knowing the profitability of the system calculated per year depended strongly on trades initiated the year before, but these trades from 2003 are still not closed because the trends in this markets are still valid but you won´t get new signals??

Helmut

TK
Roundtable Knight
Roundtable Knight
Posts: 167
Joined: Tue Apr 15, 2003 5:45 pm

Post by TK » Thu Oct 14, 2004 3:50 am

Tom Basso, a Market Wizard, offered the following:
When starting up again you should put on your positions across the whole portfolio to achieve that nice diversification and balance across the portfolio. We have modeled taking new signals only or putting the portfolio on all at once and putting it on all at once consistently beats the new signal approach.
Source: http://www.pricegroup.com/Newsletter/030102.html

Austrian
Roundtable Knight
Roundtable Knight
Posts: 223
Joined: Sat Oct 25, 2003 12:35 pm
Location: Austria, Vienna

Post by Austrian » Thu Oct 14, 2004 3:57 am

TK - Thank you

Has someone tested Tom Basso's findings??

Best regards,

Helmut

Austrian
Roundtable Knight
Roundtable Knight
Posts: 223
Joined: Sat Oct 25, 2003 12:35 pm
Location: Austria, Vienna

Post by Austrian » Thu Oct 14, 2004 7:08 am

TK wrote:Tom Basso, a Market Wizard, offered the following:
When starting up again you should put on your positions across the whole portfolio to achieve that nice diversification and balance across the portfolio. We have modeled taking new signals only or putting the portfolio on all at once and putting it on all at once consistently beats the new signal approach.
Source: http://www.pricegroup.com/Newsletter/030102.html

TK excuse me but I didn´t find the article quoted in your web link :oops:

Could you please show where you got it :P

Rgds, Helmut

TK
Roundtable Knight
Roundtable Knight
Posts: 167
Joined: Tue Apr 15, 2003 5:45 pm

Post by TK » Thu Oct 14, 2004 7:43 am

Austrian wrote:TK excuse me but I didn´t find the article quoted in your web link :oops:
I have never said it is an article. However, the quote is still there.

Try this:

1. Follow the link.
2. Press Ctrl + F.
3. Type "When starting up again".
4. Press Enter.

Can you see it now?

Austrian
Roundtable Knight
Roundtable Knight
Posts: 223
Joined: Sat Oct 25, 2003 12:35 pm
Location: Austria, Vienna

Post by Austrian » Thu Oct 14, 2004 7:53 am

TK wrote:Can you see it now?
Yes, Thanks - Rgds Helmut

TC
Roundtable Knight
Roundtable Knight
Posts: 101
Joined: Thu Jul 08, 2004 2:12 pm
Location: Toronto, Canada

Post by TC » Thu Oct 14, 2004 9:31 am

Austrian wrote:Has someone tested Tom Basso's findings??[/b]

Best regards,

Helmut
When initiating trading in a long term system I think it is better on balance to enter all trades that your system indicates are open positions than wait to enter each market on a new signal.

Immediate entry gives you two advantages:

1) You are fully diversified to the maximum extent allowed by your system from the outset

2) You start making money earlier even though a higher proportion of the initial trades will turn into losers.


As average trade length decreases you may want to review these advantages as they may not hold up as well on shorter time frames.

Good trading

Tom

Chris67
Roundtable Knight
Roundtable Knight
Posts: 1046
Joined: Tue Dec 16, 2003 2:12 pm
Location: London

Post by Chris67 » Thu Oct 14, 2004 10:18 am

If I can chuck my 2 cents in .. in a million years I would never start by putting all current positions my system holds .. sounds like utter suicide especially given in todays world all markets are fairly well correlated .. a good example would be .. my system has me long oil .. from 26 usd per barrel .. Would I come in at 54 and buy it .. after a 100% move .. I know some would argue you should but in all my experience these days it would kill you ...

Chris

Forum Mgmnt
Roundtable Knight
Roundtable Knight
Posts: 1842
Joined: Tue Apr 15, 2003 11:02 am
Contact:

Post by Forum Mgmnt » Thu Oct 14, 2004 11:34 am

First, let me say this is one of the more interesting questions I've read here in quite some time.

Second, I agree with Chris. The issue is one of stops. You really have to be willing to let the positions go all the way back to the stops for the initial entry or any ongoing trailing stops.

If you use a different stop mechanism, you're really changing the system.

For example, if you have a breakout system with a 2ATR stop and then put on a position after a long move, you have to use the normal trailing stop exit, you can't use the 2ATR stop since the system uses that only at the breakout point.

If you have a system that doesn't have any trailing stops (e.g. a moving average crossover or other indicator-based system), you might find you'll need to allow for losing 20% to 30% of your position for a single trade. If the market has moved considerably, this starts to become pretty likely.

I've tested both and while putting on new positions will get you an identical equity curve as historical testing from the starting point forward, you run a much higher risk of having a drawdown of your starting capital because of the need for much larger stops. The risk adjusted returns are lower for entering mid-stream from my testing. That's why VeriTrader doesn't do this, it only enters new trades when it starts out.

It is also one of the reasons that the Trading Recipes feature "worst case analysis" (where the system tests with all possible starting dates) is not as important as many people think. If you only take new trades your starting date becomes much less important.

If you are trading a new fund, starting off with a large drawdown is a disaster. If you are trading your own money this can be very destabilizing.

To explore these concepts, try testing systems that enter 5 ATR to 15 ATR after a normal entry would have occurred, you'll find that it doesn't work very well.

- Forum Mgmnt

P.S. Rich Dennis very strongly argued for the opposite of my advice above during the first Turtle class. If you are shooting for absolute returns he's right, if you care about risk-adjusted returns do some research and I'm confident you'll side with my approach.

Asterix
Senior Member
Senior Member
Posts: 44
Joined: Mon Apr 05, 2004 11:16 pm
Location: San Diego

Post by Asterix » Thu Oct 14, 2004 12:59 pm

Forum Mgmnt,

If I understand you right, you are saying that you should only paper trade thru an already open position until it is closed out, and you should only enter a position at the true entry point. Is that a correct interpretation?

On this same topic, has anyone attempted to use the equity curve to determine when to start trading a system? Should you start when the equity curve is in a steep rise or wait until after a significant drawdown has been passed? Or is there a better way to approach this?

I've never tested these ideas, but I think knowing when to start (assuming there is some way to know) will make or break you, especially if you are starting with limited resources.

Forum Mgmnt
Roundtable Knight
Roundtable Knight
Posts: 1842
Joined: Tue Apr 15, 2003 11:02 am
Contact:

Post by Forum Mgmnt » Thu Oct 14, 2004 1:37 pm

If I understand you right, you are saying that you should only paper trade thru an already open position until it is closed out, and you should only enter a position at the true entry point. Is that a correct interpretation?


Yes, your interpretation is correct. That's what I recommend unless you have tested that something else works better for your specific system.

The main thing is to understand the implications of whatever you decide through your own testing and analysis rather than taking other people's word for it.

- Forum Mgmnt

TC
Roundtable Knight
Roundtable Knight
Posts: 101
Joined: Thu Jul 08, 2004 2:12 pm
Location: Toronto, Canada

Post by TC » Thu Oct 14, 2004 2:27 pm

Forum Mgmnt wrote:
........ The risk adjusted returns are lower for entering mid-stream from my testing.

........... Rich Dennis very strongly argued for the opposite of my advice above during the first Turtle class. If you are shooting for absolute returns he's right, if you care about risk-adjusted returns do some research and I'm confident you'll side with my approach.
If you trade a very long time frame (I trade weekly bars) it will probably take many months before you have a properly balanced and diversified portfolio. Therefore, whilst the enter-only-on-a-new-signal approach may improve risk-adjusted returns the flip side is you are also trading a random and sub-optimal portfolio you have never backtested.

Random because you have no idea which markets you will enter or when and sub-optimal because your testing and optimisation has been conducted on the basis of trading the complete portfolio of markets.

Because I trade long term, and am not concerned with a drawdown of my initial capital, when I started trading my system I elected to enter all open position trades at the same time with, as c.f. rightly pointed out, the stops that were indicated by the system at that stage of the trade.

As to Chris's comment about Oil, I would certainly have some doubts about profitably exiting a Long position entered at $54. However, I would resist the temptation to opt out of certain markets as this goes against a fundamental tenet of system trading which is to take EVERY trade. My guess is there is a very high probability of that trade losing money on my system given where I have my stops but I would still take the trade for all of the reasons outlined above.

Essentially you have to choose between the lower absolute return & commensurately reduced risk from entering markets one by one at random over a period of many months and the higher absolute return but also greater likelihood that entering all trades at once will result in a larger-than-normal proportion of losers on those inital trades.

As always in this game there are pros and cons for every decision we have to make.

Asterix
Senior Member
Senior Member
Posts: 44
Joined: Mon Apr 05, 2004 11:16 pm
Location: San Diego

Post by Asterix » Thu Oct 14, 2004 3:13 pm

TC has a point. At the start, each market will be added incrementally as the entry points are triggered so you won't have a fully optimal portfolio until some time in the future.

Forum Mgmnt, since Veritrader has this constraint built in, I would assume that it builds the portfolio incrementally. Is that the case? If so, have you looked at the effect on the equity curve while the portfolio is still sub-optimal?

Forum Mgmnt
Roundtable Knight
Roundtable Knight
Posts: 1842
Joined: Tue Apr 15, 2003 11:02 am
Contact:

Post by Forum Mgmnt » Thu Oct 14, 2004 3:35 pm

I've tested both approaches side by side. My testing indicates that the approach I outline is better for my needs which is targeted at fund performance that investors want.

I'll add the ability to configure whether or not VeriTrader just jumps in right away or takes only new signals to 2.0. It will be a simple change. That way our customers can compare it themselves.

This will immediately create the need for doing start date analysis ala Trading Recipe's "Worst Case Scenario" feature, so I'll have to add that too I guess because this approach is much more sensitive to starting date.
If so, have you looked at the effect on the equity curve while the portfolio is still sub-optimal?
All the systems I trade have periods where you are out of the market. You are not getting diversification at that point either. You generally need diversification to counter large adverse moves. You aren't exposed to them if you put on new trades only so the need for diversification is less.

It is very possible that with very long-term systems and large portfolios that there is enough benefit from the diversification to counter the increased risk of a drawdown. That's why I suggest testing for yourself rather than guessing or relying on Tom Basso or me.

- Forum Mgmnt

choppystride
Full Member
Full Member
Posts: 22
Joined: Thu Jul 17, 2003 4:54 am

Post by choppystride » Thu Oct 14, 2004 4:15 pm

TC wrote:
Forum Mgmnt wrote:
........ The risk adjusted returns are lower for entering mid-stream from my testing.

........... Rich Dennis very strongly argued for the opposite of my advice above during the first Turtle class. If you are shooting for absolute returns he's right, if you care about risk-adjusted returns do some research and I'm confident you'll side with my approach.
If you trade a very long time frame (I trade weekly bars) it will probably take many months before you have a properly balanced and diversified portfolio. Therefore, whilst the enter-only-on-a-new-signal approach may improve risk-adjusted returns the flip side is you are also trading a random and sub-optimal portfolio you have never backtested.

Random because you have no idea which markets you will enter or when and sub-optimal because your testing and optimisation has been conducted on the basis of trading the complete portfolio of markets.

Because I trade long term, and am not concerned with a drawdown of my initial capital, when I started trading my system I elected to enter all open position trades at the same time with, as c.f. rightly pointed out, the stops that were indicated by the system at that stage of the trade.

As to Chris's comment about Oil, I would certainly have some doubts about profitably exiting a Long position entered at $54. However, I would resist the temptation to opt out of certain markets as this goes against a fundamental tenet of system trading which is to take EVERY trade. My guess is there is a very high probability of that trade losing money on my system given where I have my stops but I would still take the trade for all of the reasons outlined above.

Essentially you have to choose between the lower absolute return & commensurately reduced risk from entering markets one by one at random over a period of many months and the higher absolute return but also greater likelihood that entering all trades at once will result in a larger-than-normal proportion of losers on those inital trades.

As always in this game there are pros and cons for every decision we have to make.
Hmmm....interesting discussion...but I'm a bit confused here....

I thought one of the goals of a system is to exhibit performance consistency regardless of its test period (provided it's long enough and has covered various types of markets). This implies that its validity should be insensitive to its initial conditions (i.e. start date).

Therefore, at any arbitrary point in time, a trader should be able to begin trading the system from scratch - that is, by initially taking new signals only - and expect its performance to conform to its backtest results going forward.

Because, after all, don't all backtests begin at some arbitrary point in time by taking only new signals and ignoring possible open trades?

If there's such a need for continuity that a trader must force himself to take all open trades at the beginning, doesn't this imply that the system is highly dependent on its initial conditions?

Consider this hypothetical (though perhaps unlikely) scenario: let's say that your backtest began on 1980-01-01 and accordingly, you're supposed to be long 2 Silver today. But what if you began your backtest on 1979-12-01 instead? In the second case, the system says that you should be short 2 Silver today. Which position would you take?

Asterix
Senior Member
Senior Member
Posts: 44
Joined: Mon Apr 05, 2004 11:16 pm
Location: San Diego

Post by Asterix » Thu Oct 14, 2004 5:09 pm

choppystride wrote:
I thought one of the goals of a system is to exhibit performance consistency regardless of its test period (provided it's long enough and has covered various types of markets). This implies that its validity should be insensitive to its initial conditions (i.e. start date).
It depends on what you mean by consistency. I would expect a system to be consistent if you get about the same distribution of closed out trade results (wins & losses) in two different tests of different time periods. But that doesn't imply that they will occur in the same order or that the corresponding equity curves will look similar.

From my perspective, the main point of this topic is that there may be some ways to determine when and how to begin trading a system as opposed to just jumping into a market randomly. The last thing anyone would want is to start trading a good system and immediately go into a steep drawdown. This point goes a little beyond what c.f. stated regarding waiting until you get a true entry signal to enter a specific market. I'm wondering whether or not you can use the equity curve (or some other parameter) as a means to determine when to actually enter positions.

For example, one approach might be to paper trade until the equity curve breaks a 55 day high and then open real positions. I haven't actually tested this idea, (which I will when I get some time) but I'm suggesting it here as food for further thought in this discussion.

Regarding your scenario about getting different signals depending on the start date. I guess that might be possible depending on the type of rules you use although I doubt if a LTTF system would do that. However, as I stated earlier, the consistency would show in the overall distribution of closed out trades, not the order in which they occur.

Mathemagician
Full Member
Full Member
Posts: 20
Joined: Thu Jun 28, 2007 1:46 pm
Contact:

Post by Mathemagician » Thu May 15, 2008 9:55 pm

Chris67 wrote:If I can chuck my 2 cents in .. in a million years I would never start by putting all current positions my system holds .. sounds like utter suicide especially given in todays world all markets are fairly well correlated .. a good example would be .. my system has me long oil .. from 26 usd per barrel .. Would I come in at 54 and buy it .. after a 100% move .. I know some would argue you should but in all my experience these days it would kill you ...

Chris
Hindsight is 20/20...

M

Post Reply