I downloaded the OpeningRange.efs and thought it was very nice but it doesn't come with any backtesting functionality. I'm not much of a coder and eSignal programming is not very user-friendly so how about making available an OpeningRange.efs file that includes backtesting functionality so we can change the parameters and backtest on different time frames, markets etc. I notice just by looking at the past couple of months that it works very well with a 20 point stop and limit and I see a few other index futures where it could work well. I would of course post all my results.

Hi,

Yes the EFS that is in the file share just shades the opening range in. eSignal is very bad for back testing (I had to learn this the hard way). I have a script I use. It is over 800 lines of code and mushed up with calls to external efs references and obselete code. Its something of a frankenstein that I use to test most of the things I write. One of the things on my to do list is to clean it up and make it available for download.

eSignal does have some free EFS's online that have strategy examples included in them. I'm actually having Bearishtrader convert the range rules to Trade Station. That should be easier.

You're right, EFS is a total nightmare to backtest. The examples and tutorials they provide are not very enlightening and given what you say about 800 lines of code I think I'll give up and try it in MetaStock.

This is just a cursory test of the Maxspan method over the last 4 months(1 minute data) using a Tradestation script (KALOpnRang(tick)v2.4) kindly provided by Sunrise. Slippage and commission are not included.

Maxspan returned 1.01% and did not beat a buy and hold return of 5.22% over the last four months, although it only spent 5.47% of the time in the market.

Max. Shares/Contracts Held 3 3 3
Total Shares/Contracts Held 147 75 72
Account Size Required $1,380.00 $1,380.00 $955.00
Total Slippage $0 $0 $0
Total Commission $0 $0 $0
Return on Initial Capital 1.01%
Annual Rate of Return 3.50%
Buy & Hold Return 5.22%
Return on Account 73.55%
Avg. Monthly Return $253.75
Std. Deviation of Monthly Return $472.09

I have just investigated the @YM continous contract, 1 yr- 1min chart and the KALopenrg script. It appeared that if you believe past performances, don't trade Maxspan in March, May, June, August, at least that is what the past 2006 performances tells you:

Over the last 6 months: 11.23% (annual rate of return)
Over the last 12 months: 1.46% (annual rate of return)

Period Avg. Net Profit % Gain %Profitable

January $1,225.16 1.23% 81.58%
February $26.84 0.03% 62.96%
March ($2,243.16) (2.21%) 65.38%
April $622.56 0.64% 76.47%
May ($1,504.84) (1.53%) 72.97%
June ($1,252.76) (1.29%) 70.00%
July $1,235.16 1.29% 81.58%
August ($1,016.72) (1.05%) 75.61%
September $875.16 0.91% 78.95%
October $977.24 1.01% 77.50%
November $1,326.72 1.36% 82.50%
December $1,135.68 1.15% 71.79%

The new report has added commission of 4.48 and a slippage of 5.
The entire report is here: http://www.divshare.com/download/153293-7c8

Kebi,

These results do not match the back testing results I've obtained on my own. That is another reason we have not publicly released this TS script to the users. I spent months and verified my results with charts on each day. I'm confident the results I posted are correct.

For example for August there was a profit of $1,546 but you've listed -1,016.72

There are many nuances with back testing that must be learned. Fill prices on open, close, trail stop problems.

I use back testing to see if a setup is GENERALLY profitable over all. Don't read too much in to it. If a back test shows we lost last Wednesday, that doesn't mean we'll lose next Wednesday. Whatever economic or market events that caused a system to lose one month may not be present when that month rolls around again.

Now if we tested a setup that lost on Monday 90% of the time, yes I'd make the decision to drop that day off the auto trader.

We'll continue to build our track record on the setups we trade. That is the ultimate profitability.

CajunSniper

Thank you CS. Could you post some results, specifically of August 06 for example. The analysis I ran showed at least 5 loses > 600 in August06, the last two shown here:

BTW, thank you for all your hard work on Maxspan. This posting is in no way criticising your work or Puretick. I think everyone wants MS to work and anything that may help to that end is appreciated by all.

203 Buy 08/22/06 09:27 BBrk $11,579 3 ($708.96) (0.40%) $45.00 6.12% $4 $5
Sell 08/22/06 12:06 Sell $11,533 ($690.00) ($4,425.16) ($690.00) 0.00% (93.88%) $4 $5


214 Sell Short 08/30/06 08:55 SBrk $11,548 3 ($633.96) (0.36%) $75.00 10.87% $4 $5
Buy to Cover 08/30/06 09:41 Cover $11,589 ($615.00) ($4,201.32) ($615.00) 0.00% (89.13%) $4 $5

Sure I will PM you in the room in a sec.

Thanks CS, got your PM and data
We may be working on different YM data from different feed. The entry time and price is clearly different. Going by Yahoo ^DJI was 11344.4 open 11426.1 Hi and 11339.8 Close on 8/22/2006 so it looks like your data is more accurate. I will double check TS @YM data if there is a third party data somewhere : http://www.britefutures.com/quotes/futc ... &year=2007

CS's selected data:

24 67 EntryShort Short Entry 8/22/2006 1:30 PM 11350 3 $118
1% $1547 15% $420 4% ($225) (2%)
68 ExitShort Short TGT 1 8/22/2006 1:30 PM 11340 2
69 ExitShort *Short Stop* 8/22/2006 2:02 PM 11344 1

30 85 EntryShort Short Entry 8/30/2006 10:02 AM 11374 3 ($612)
(5%) $1970 20% $90 1% ($525) (4%)
86 ExitShort *Short Stop* 8/30/2006 10:42 AM 11414 3

I have lots of questions about backtesting in general and Maxspan and the method you use to come up with the parameters in particular.

First let me say that I feel out of place in the Geeky Traders section, as I am a computer illiterate. I know nothing about programming or back testing.

I am assuming that once a backtest program is set up, it is relatively easy to change the criteria and run different scenarios.

My first question is why 8 minute candles? I am new to trading futures but this is the first time I have heard of using 8-minute candles. Why not 3 or 5 that are available on most software?

How many different scenarios did you run? What data did you use? How easy is it to break down the results? That is can you give the number of trades that hit the first target and then stopped out? Can you break out the results when the trade was with or against the filters?

Lots more questions but if it’s a pain in the *ss let me know and I’ll go away.

While I'm at it I see in one of the videos you mention 90% reliability.

What was your maximim drawdown in your backtest. How many losing days in a row?

I may have missed it but I assume that if target or stop are not hit, you would exit at the end of the day. How many times did this happen?

Hi boib22, I can answer a few questions. First, please refer to the Jan 5 2007 blog entry on Opening range rules: http://www.puretick.com/blog/index.php?paged=3 and the video therein if you haven't already done so. CS did referred to using two 3 minute bars instead of 8 minute bars...
As to backtesting, apparently it is harder under Esignal as described above, but easier under TS (Tradestation). There is a TS KALOpnrang script by Sunrise, however I think it is slightly less robust and tested than CS's Esignal script. KAL script currently performs at about 75% profitability in backtests. As shown in numbers above and the chart below, on 8/22/06, the KAL script may have incorrectly entered a buy that resulted in a max loss ($690) while CS's script was profitable at +$118 with a short entry 3 hours later.

BTW, it looks like CS is using an unadjusted continuous YM historical data while the previous analysis I posted was using the TS 8.1 default back adjustment method. Any TS KALOpnrang backtesting from now on should probably be done using the Unadjusted @YM data (@YM=11INN) available with TS 8.2: https://www.tradestation.com/support/wh ... _0820.aspx





The CS maximum drawdown for a 6 month analysis was ($885) with 3 contracts:

Run-up/Drawdown

Run-Up
Drawdown

Max Value $1770 ($885)
Max Value Date 8/4/2006 11/27/2006
Avg Value $451 ($371)
Max Value (%) 17% (6%)
Max Value (%) Date 8/4/2006 7/21/2006
1 Std. Deviation $335 $161
Avg + 1 Std. Deviation $786 ($210)
Avg - 1 Std. Deviation $116 ($532)

kebi1

So how tough is it to change parameters and run these back tests?

I was wondering what difference it would make if one were to tighten up the stops and run out the profit trarget.

How difficult would it be to scale out stops at 20 , 30 ,& 40 rather than take the big hits on days the system doesn't work.

It seems to me that with a 40 point stop and a 10 point target for the first 2/3s you need the 90% reliability that CS claims to have gotten.

The reason I asked about 8 min candles is that I wondered if the system wasn't fitted to the data to optimise it.

Looks like I should invest in some sort of backtesting software.

Thanks for the response. I looked at your data upthread that showed a less than 2% gain. Thats not much of an edge.

It is very easy to vary the parameters in TS and run the backtest. If you have a fast PC and internet, it shouldn't take more than 15 to 30 seconds. I would imagine it is similar in Esignal, although setting up the backtest seemed more difficult.
Tightening the stops will not make it more profitable, I think CS and Anek has done quite a bit of that. Anek was talking about increasing stops to a gut wrenching and John Kennedy Toole ( http://en.wikipedia.org/wiki/John_Kennedy_Toole and Friday's room transcript) suicide inducing 50 to 60 points
If you can tell a 20 point trending day from a 40 point trending day then you could scale out stops - but if you can tell that ahead of time - you will also be very rich
I won't be too quick to dismiss CS's analysis of 90% profitability - but I still need to see confirmation using TS KALOpnrang script.

The unadjusted Tradestation 8.2 YM continuous contract (1 minute chart using the TS command sequence: View, data window, show all, save) historical data 2 years compressed as a winrar file is available here: http://www.divshare.com/download/163590-641

Could anyone with Esignal do the same with their YM continuous 1min data please (unadjusted and using exchange time, if possible)?
The way to save ascii data in Esignal is here: http://www.esignalcentral.com/exchange/ ... entral.asp

Esignal continous contract can be unadjusted here: http://www.esignalcentral.com/exchange/ ... t_news.asp

After compression with zip or rar, the file size should be 2 to 3 MB and one could upload to something like http://www.divshare.com/upload for free.

Thank you in advance.