Bob and Shelley Again

[phil_trade]

Well, can the 'garbage' could not be sifted to make everyone more efficient as a trader :

of course yes...but may be not in this thread

[SteveHopwood]

Bloody hell, but it is scary the tricks memory can play.

I settled down to code the candle length atr thingy, to find I had already done it - and forgotten.

So I settled down to update the user guide, to find I had already done it - and forgotten.

V 2o is in post 1. From the updated user guide (formatting gets lost when I do this):
ATR candle length filter: I have used ATR to detect the average pips movement of the chart's candles, and abort a potential trade if the previous candle has spiked – as the result of a news-release, for example:
ClAtrTimeFrame: the time frame to use to calculate ATR. The zero default means the current chart time frame.
ClAtrPeriod: the ATR Period input. A zero value disables this feature.
ClAtrMultiplier: multiplies the value of ATR by this input to produce a maximum candle length for the candle that just closed. If the candle has moved by more than this number of pips, then BASA will not trade.

You will see that BASA sleeps for 5 seconds before leaping into life. My platforms tend to freeze if I have to do a restart whilst the markets are open. Tommaso mentioned that Empty4 has a lot of stuff to do in the first few seconds after startup, and I am making this worse by having about 20 instances of BASA trying to leap into life. The sleep is an attempt to overcome this.

Coders, sleep code in init(); candle length check in IsTradingAllowed()

[fxozgirl]

Bloody hell, but it is scary the tricks memory can play.

I settled down to code the candle length atr thingy, to find I had already done it - and forgotten.

So I settled down to update the user guide, to find I had already done it - and forgotten.

V 2o is in post 1. From the updated user guide (formatting gets lost when I do this):
ATR candle length filter: I have used ATR to detect the average pips movement of the chart's candles, and abort a potential trade if the previous candle has spiked – as the result of a news-release, for example:
ClAtrTimeFrame: the time frame to use to calculate ATR. The zero default means the current chart time frame.
ClAtrPeriod: the ATR Period input. A zero value disables this feature.
ClAtrMultiplier: multiplies the value of ATR by this input to produce a maximum candle length for the candle that just closed. If the candle has moved by more than this number of pips, then BASA will not trade.

You will see that BASA sleeps for 5 seconds before leaping into life. My platforms tend to freeze if I have to do a restart whilst the markets are open. Tommaso mentioned that Empty4 has a lot of stuff to do in the first few seconds after startup, and I am making this worse by having about 20 instances of BASA trying to leap into life. The sleep is an attempt to overcome this.

Coders, sleep code in init(); candle length check in IsTradingAllowed()

Thanks Steve...I'll put it on some demo's and give it a whirl

[garyfritz]

Also, I'm sure (hope ) Gary will jump in here with his mathematical genious and share some more insights he may have gleaned from these results

The only things that come to mind are:

* Your ATR filter is knocking out about as many winning trades as losers, resulting in almost identical win %. That's not ideal but it IS helping overall. You might want to consider looking for some slightly different filter approach that might zero in on losers better? But with 80% wins you're already doing pretty darn great.

* All other things being equal (and they pretty much were), making $X profit in Y trades is much better than making $X profit in 2*Y trades. The larger average trade size means you're much less sensitive to slight shifts in spread, slippage, etc. You also pay less in commission/spread & so on. For that reason I'd say #1 would be my choice.

* You did a great job of back-checking this. This is MUCH MUCH better than watching it for a week or two and making decisions based on what you see in realtime. More data gives you more certainty that the change you're making actually holds up over time. Still, the different tests only change the result by a few dozen trades. That's a smaller sample than I like to use. But given that it has to be based on the live trades in your account, this may be the best you can do.

Gary

[spotdespot]

Very interesting test Shelley - thanks for doing that. I actually spent quite a lot of time at the weekend going though the manual trading system I am working on for the same reason with very similar logic.

My method was much less scientific in terms of results, as my manual trading isn't as rigidly rule based - especially regarding exits BUT I actually came to the same conclusion that 2 * ATR was the best filter for me. The only difference was that I used ATR(50) rather than ATR(14). I don't think this matters much when trading BASA on H4 but on H1 ATR(14) will cause you to miss many good trades imo as, for example, when Frankfurt opens the previous 14 hours is comprised of the second half of the NY session (no London overlap) and "Asia" (NZ/Oz/Tokyo) with the potential issue being obvious. Maybe something to bear in mind for those BASA H1 traders anyway.

One other thing I was looking at was (as my manual system uses MA's to a degree) is the distance from an MA that the trade was taken. I didn't get too far with this yet but I will try to look at it more this week. Essentially my theory that I am working on is that trades that are initiated too far away from the norm (easiest way to measure is too far from an MA) are more likely to fail. If I discover anything concrete and useful I will update here.

Cheers,
Dave.

[fchaman]

There must be a way to avoid such a big drawdown....almost -150pips!

gbpnzd.GIF

[RxCape]

........ Essentially my theory that I am working on is that trades that are initiated too far away from the norm (easiest way to measure is too far from an MA) are more likely to fail.

Cheers,
Dave.

Dave: I have made a similar observation. In my empirical observation, trades taken 250+ pips away from the 240 LWMA are less likely to move into profit. I have recently been trying to identify this low potential area with divergence. I do not have this phenomenon documented to the point that I could publish it here, but it appears to be a valid observation. LAst week I actually took a few counter BASA trend trades based upon the 250+ pips and divergence theory. Worked out well. I closed them at +50 pips each because they were countertrend. I will keep at it and if anything valuable comes out of it, I will publish my observations.

Tom

[SteveHopwood]

........ Essentially my theory that I am working on is that trades that are initiated too far away from the norm (easiest way to measure is too far from an MA) are more likely to fail.

Cheers,
Dave.

Dave: I have made a similar observation. In my empirical observation, trades taken 250+ pips away from the 240 LWMA are less likely to move into profit. I have recently been trying to identify this low potential area with divergence. I do not have this phenomenon documented to the point that I could publish it here, but it appears to be a valid observation. LAst week I actually took a few counter BASA trend trades based upon the 250+ pips and divergence theory. Worked out well. I closed them at +50 pips each because they were countertrend. I will keep at it and if anything valuable comes out of it, I will publish my observations.

Tom

Please do Tom. This kind of contribution is why the likes of Shelley post their systems here.

It would take about 5 minutes to code a max distance from the 240 filter.

[fxozgirl]

Andy (MrLong) shared an ADR Bands indicator he had a few months ago, I quite like the concept...basically similar to what Dave & Tom have discussed above, the main reason I haven't investigated further with this indicator is it seems to have strange behaviour (actually it's probably not strange at all, it's probably just my ignorance )

I have attached a chart below showing the ADR bands attached and you will see there's a nice smooth curve and then a sharp vertical drop to another completely different level. If I can figure out why that happens then this indicator could be a good filter to prevent trading well outside the usual range.

Any input would be appreciated...perhaps Andy can enlighten me/us??

adrbands.png

[BonioM]

Bloody hell, but it is scary the tricks memory can play.

I settled down to code the candle length atr thingy, to find I had already done it - and forgotten.

So I settled down to update the user guide, to find I had already done it - and forgotten.

V 2o is in post 1. From the updated user guide (formatting gets lost when I do this):
ATR candle length filter: I have used ATR to detect the average pips movement of the chart's candles, and abort a potential trade if the previous candle has spiked – as the result of a news-release, for example:
ClAtrTimeFrame: the time frame to use to calculate ATR. The zero default means the current chart time frame.
ClAtrPeriod: the ATR Period input. A zero value disables this feature.
ClAtrMultiplier: multiplies the value of ATR by this input to produce a maximum candle length for the candle that just closed. If the candle has moved by more than this number of pips, then BASA will not trade.

You will see that BASA sleeps for 5 seconds before leaping into life. My platforms tend to freeze if I have to do a restart whilst the markets are open. Tommaso mentioned that Empty4 has a lot of stuff to do in the first few seconds after startup, and I am making this worse by having about 20 instances of BASA trying to leap into life. The sleep is an attempt to overcome this.

Coders, sleep code in init(); candle length check in IsTradingAllowed()

Steve, I think bar length condition checking should be changed (based on fxozgirl advice).
In version v2o there is:

Code: Select all

if (ClAtrPeriod > 0)
   {
      if (iClose(NULL, TradingTimeFrame, 1) - iLow(NULL, TradingTimeFrame, 1) > ClAtrVal)      return(false);
   }//if (ClAtrPeriod > 0)

it seems that we've to use Open/Close values, then the code would be:

Code: Select all

if (ClAtrPeriod > 0)
   {
      if (MathAbs(iClose(NULL, TradingTimeFrame, 1) - iOpen(NULL, TradingTimeFrame, 1)) > ClAtrVal) return(false);
   }//if (ClAtrPeriod > 0)

Please correct me if I'm wrong.
Thanks a lot for your great effort that you put in BASA development.

Marcin