If your using EFS, enter the highest return you've had...

[chrisf95003]

Thats cool. Thanks for posting that. What if the system is in a trade, things are still going good with the trade, but the EOD is coming up. When does the system say its time to get no matter what? I am assuming this is a day trading system.

Thanks,
Chris

[BakedWafer]

3:30 is what I consider end of day (EOD). The system is told to close all positions at EOD and what until 8:00 AM.

+ 2,500 for last 2 days.

I expect hiccups soon.

[Luk]

I realize that not many people would want to share their profitable system, but out of curiosity can someone post an example of their failed or not effective system? I'm totaly green with this efs/system stuff and are curious as to how this whole thing is done and how it works. What indicators or conditions are used as buy and sell triggers, etc? thanks.

[soylent]

Luk --

That's an easy one...

Just look under your (right click on a chart) Formulas -> Backtesting.

Select any one of the sample backtesting scripts and run it.

To run a back test just right click and select 'tools->Backtesting'. Select the same formula to backtest.

-c

[derekg]

LOL!

In a subversive way, that was a really clever response, soylent.

[Luk]

thanks but it's not the backtesting I'm interested in. I'm interested in a system it self. I asked for an example of a INEFFECTIVE formula/system because I know nobody wants to share their effective ones, I'm not stupid. I'm still curious how these systmes look in general. Are they including different indicators that create signals, or are they just based on one indicator. Or maybe they include different conditions such as voulme requirnments combined with certain price conditions. I don't know. From the backtesting results some posted I'm not sure if one can get these results from simple system based on one indicator.

[gspiker]

I think that was the point of the post. The examples that are provided are not effective systems.

But unlike an earlier reply to that post, I don't think that the message was subversive. After all they were put up as "how to code backtests" not of "how to code winning systems".

In fact, just about any system that uses a single standard indicator, in only one time frame (which is what most, if not all of the example ones do) will be relatively ineffective unless you put solid money management rules in place (which none of the examples do).

To make an effective (ie: reasonably profitable) system you need to pick some fairly reasonable indicators that can be used in conjunction with each other, or one reasonable indicator that can be used in multiple timeframes effectively - and then implement sound money management rules that match your goals and the limitations of the rules of the system.

To make an ineffctive system (ie: overall not profitable or not profitable enough to make money) just ignore the above rules.

Simple as very hard pie.

Garth

[KCOMOLLO]

Brad,
Great thread. Can you explain something you said earlier regarding your system. You indicated that you use MARKET THISBAR. Is that correct? If so are you computing this on every tick and then executing intrabar?

I would have to think that the only way to get accurate test results (difficult in Esignal) is CLOSE THISBAR at the very least and more likely OPEN NEXTBAR.

Also, how often do you find your limit orders not filled in realtime?

Thanks and again GREAT THREAD!

Keth

[Doji3333]

Using Strategy functions...

First off... thanks for the compliment - but there are lots of other people here to that make this thread so interesting. So, thanks to them too.

Next.... the use of Strategy functions is critical to the outcome of your strategy's backtest.. I've learned three simple rules that I apply to all my systems...

1. If you are entering on a limit price, you MUST verify the price has actually BREACHED the price you are trying to enter at before you execute the trade. Otherwise, your system will be skewed. Limit orders do not always fill when the market price touches your price level. It has to BREACH it.

When using limit orders for a backtesting system, use Strategy.LIMIT, Strategy.THISBAR - and verify the price bar is within the range of your LIMIT price.

2. If you are using a system that waits for a bar to finish forming, then executes trades, use Strategy.MARKET, Strategy.THISBAR. Now, there is a trick to accomplish this.. you have to check for the entry conditions starting at -1 bar (not 0 bar). This way, you are always checking the finished -1 bar and entering on the open of the 0 bar. If you check the 0 bar for entry conditions and enter on the 0 bar - then your screwed.

3. If you are using setComputeOnClose(true), then you should be using Strategy.MARKET, Strategy.NEXTBAR. Because the setComputeOnClose() function causes the EFS to run at the END of any current bar, you have to use NEXTBAR for entries. If you are not using setComputeOnClose() - then choose #1 or #2.

Now, there may be other instances when you have to alter these conditions, but for the most part, these are generally accepted rules.

Hope this helps..

B

[KCOMOLLO]

Thanks for the great explanation. Makes perfect sense. I had coded something with MARKET THISBAR and it doubled its 10k account in 10 days with one contract on the YM with slippage of 4 ticks and normal commish. I thought I had found the Holy Grail he he. After realizing my mistake I redid it as you describe. Not as pretty.

Appreciate the feedback.

Cheers,
Keth

[Gavishti]

I think Garth has an excellent point:

posted by gspker
The examples that are provided are not effective systems.

Perhaps that is why no one has responded to my query.

posted by gavishTi
I am curious as to how many of these "systems" trade with real money and for how long they have been traded.

[gspiker]

Gavishti,

The quote of mine is taken a bit out of context, as it was from another thread talking about the backtest efs scripts provided as examples by eSignal.


I think nobody has responded because these are mostly new systems with little to no real trades under their belt yet. I personally would wait for a few month of RT trading, even with an intraday system, before I would make any statements on how well a system was doing.

And I understand why you ask. It's very easy to deign systems that give great results in backtesting that do poorly in RT trading. After you have enough experience in backtesting and RT trading you learn what to avoid doing with the backtest, but it can take a while to pass that learning curve.

Garth

[Gavishti]

Garth,

My apologies for quoting you out of context. It was not intentional. I have a great deal of respect for your opinions.

I agree completely with what you said here in your last post.

posted by Garth
It's very easy to design systems that give great results in backtesting that do poorly in RT trading.

That was the whole point of my question. Which is pretty much what spike500 said in an earlier post. Many of us have developed fantastic systems that fail miserably when put into real trading. Before getting overly excited about these great equity curves perhaps it would be appropriate to take a real hard look at the validity of the tests. It may be painful, but far less painful than trading them with real money. I believe it was spike500 that said it took him five years to find a truly workable system.


gavishTi

[gspiker]

Gavishti,

No offense taken. I just wanted people to be clear on the context of the quote least they think I was calling **THEIR** systems unprofitable...which I wouldn't really have a clue on.

I agree with you, but this thread is interesting and does serve a purpose so I am glad it was started. I also think it would be fun to hear back from the posters here after they have traded the systems described here for 4-6 months time and see if they performed as expected, and if not lessons learned as to why.

In the tail end of the internet/nasdaq boom I had found a great system for determining when one of those volatile internet stocks was about to go non-linear on their price. It didn't signal every just move, but when it signaled it was very accurate. I backtested the system with what little data there was (remember, in market terms this was a short craze) and then because I didn't have a lot fo data watched in in realtime for a while. I think I got to trade it all of 9 trades or so before the bubble burst and my money machine never signaled again. You can bet I am keeping this around for the next bubble (maybe biotech?) to see if it will work on any high volatility stocks, but the point is it went from off-the-hook profits to not a single signal in almost no time. Granted I knew the system would be of little value once the bubble burst...but I was hoping for a bit more of a run ;-)

Errors in coding, changing market conditions, having the correct/realistic idea of slippage and not allowing for traders emotions are all well known reasons that good backtests turn into bad RT systems. But not all systems that backtest well will suffer from these things...so it will be interesting to hear back and see what has happened.

Garth

[gspiker]

Gavishti made an interesting point in his post, when pointed out spike500's statement that it took him five years to find a truly workable system.

What is interesting, and may make a nice addition to this thread, is what do you do once you find a system that seems to work for you.

In my case, I found a system and have been making evolutionary changes to it for years. There is a whole process I run through of backtesting small changes, then running both versions of the system in RT (trading only the old one, but watching the new) and then finally adopting the changes if they seem to work better than the old. My system of today bears little resemblance to what I started out with, other than the underlying concept (theory) on how to build a good system. But to me it is still the same system.

So how much tinkering do you do with a working system (if any), and how quickly do you adopt changes? And after months/years of tinkering are you seeing much better results than you saw with the initial system? Where there specific problems you were able to solve and how did you solve them? I would like to hear how others approach this.

Garth