NON-INDICATOR (indicator-less) Trading

Jahajee,

I've been working on the same exact strategy for a couple of weeks. In backtesting it does not seem to be particularly profitable. I attach my performance report for ESZ8 as well as backtest of the other contracts I have available.

As you can see, ESZ8 did okay, but the win/loss ratio is poor and 12 bucks a trade is nothing to write home about. In addition, only ESU8 is also profitable.

Would you consider posting your performance report for comparison?

Thanks,

Big Cabrito
 

Attachments

Quote from big_cabrito:

Jahajee,

I've been working on the same exact strategy for a couple of weeks. In backtesting it does not seem to be particularly profitable. I attach my performance report for ESZ8 as well as backtest of the other contracts I have available.

As you can see, ESZ8 did okay, but the win/loss ratio is poor and 12 bucks a trade is nothing to write home about. In addition, only ESU8 is also profitable.

Would you consider posting your performance report for comparison?

Thanks,

Big Cabrito
More...


Thanks for the info. I haven't done anything as yet - caught up in trading today -- actually mostly watching with a few trades taken .
I take your word for it: if it's that bad then I would not waste time on it.

Some have suggested (adv minus decl issue) may be a better parameter.
 
Quote from Jahajee:

Thanks for the info. I haven't done anything as yet - caught up in trading today -- actually mostly watching with a few trades taken .
I take your word for it: if it's that bad then I would not waste time on it.

Some have suggested (adv minus decl issue) may be a better parameter.
More...

LOL! Don't let me stop you! It's quite likely your strategy would be very different. I was hoping to compare your performance to mine to get an idea why mine doesn't perform as well as I would hope. To my understanding this strategy should be similar to Price Action trading a la AHG, and so I don't understand the discrepancy between reported discretionary results and the backtested system.

FWIW I can f**k up just about any strategy. I understand you successfully automated SPM while my version of SPM performs about the same as the one I posted.

Happy Trading,

Big Cabrito
 
Quote from big_cabrito:

LOL! Don't let me stop you! It's quite likely your strategy would be very different. I was hoping to compare your performance to mine to get an idea why mine doesn't perform as well as I would hope. To my understanding this strategy should be similar to Price Action trading a la AHG, and so I don't understand the discrepancy between reported discretionary results and the backtested system.

FWIW I can f**k up just about any strategy. I understand you successfully automated SPM while my version of SPM performs about the same as the one I posted.

Happy Trading,

Big Cabrito
More...


I will look at the system weekend. I think the markets will be very volatile during this option expiration week (today being the calm before the storm??) so I may be trading for most of the day during next 4 days.

It is possible with a filter like momentum or using Keltner we may improve the system.
 
Quote from Jahajee:

ok, remove the lines and trade with HHs and LLs
Are thes indicators also?
More...

Yes, of course they are.

Nobody is objecting to the mechanics of what you're talking about - the objection is to your incorrect characterization that it is "indicator-less".
 
If I might make a suggestion on improvement...

#1 - Formalize the definition of "trend" for this approach (at a minimum entry signal, exit signal, minimum move)

#2 - Figure out what timeframe is currently providing moves consistent with your definition of trend (ok, this is not exactly "currently" as you don't have a time machine, there's nothing for it but to look at the recent past)

#3 - Use bars that are appropriate for the timeframe determined in step #2. EG, having a definition of "trend" that implies 10% moves and using 1min bars in a market moving at 0.5% a day will result in death by a thousand cuts.

You can also flip this around - start with a timeframe you want to trade - determine the "current" trendiness of that timeframe - adjust your entry/exit/target levels to reflect the "actual" state of trendiness rather than what "should" be happening.

It is important to remember that "trend" is a mathematical construction - it is an observation laid on top of historical data and is *always* rearward-looking.
 
Quote from Jahajee:

This is a work in progress .. system to scalp
NQ 1minute

Using basic Keltner double band with momentum extremes ....

mB, kB are buys, mS and kS are sells, X are exits
More...

Thanks, I will take a look at it in the morning. Let's see what we can do with this!

Big C
 
Quote from Random.Capital:

If I might make a suggestion on improvement...

#1 - Formalize the definition of "trend" for this approach (at a minimum entry signal, exit signal, minimum move)

#2 - Figure out what timeframe is currently providing moves consistent with your definition of trend (ok, this is not exactly "currently" as you don't have a time machine, there's nothing for it but to look at the recent past)

#3 - Use bars that are appropriate for the timeframe determined in step #2. EG, having a definition of "trend" that implies 10% moves and using 1min bars in a market moving at 0.5% a day will result in death by a thousand cuts.

You can also flip this around - start with a timeframe you want to trade - determine the "current" trendiness of that timeframe - adjust your entry/exit/target levels to reflect the "actual" state of trendiness rather than what "should" be happening.

It is important to remember that "trend" is a mathematical construction - it is an observation laid on top of historical data and is *always* rearward-looking.
More...

#1 - Trend is defined as HH - HL - HH - HL (or reverse). Entry signal is 3BR after the last HL (not ideal, but the best I could come up with so far). Exit signal is failure to make HH/HL or predefined stopLoss/profitTargets (again, tests better).

#2, 3 - Timeframe was 1-minute bars. I also tried longer timeframes but in general, in backtesting the shorter the timeframe the better -- it "kind of" simulates real life better.

I generally program the strategy and then optimize on stopLoss/profitTarget so by definition it adjusts to the "actual state of trendiness". Of course this is all in hindsight, but as you point out, what other choice is there?

The point of the exercise is to (a) test out a proposed strategy to determine if it has any merit and (b) if it does, try to get it to demonstrate consistent profitability on historical data. My experience is that many things "kind of" work and in fact I'm running one live automated strategy now that is "kind of" successful, and have another candidate that I'm still testing that looks pretty good too. I think this confirms the assertion you will often hear that it's not the system, it's the trader -- in the case of automation, I believe it's the money management that makes it or breaks it.

I also trade discretionary and would never take a trade using a method that I hadn't backtested and proven to be on balance profitable.

HTH,

Big C
 
You must log in or register to reply here.
Back
Top