20EMA Strategy Filter???

[cd173]

Morning Everyone

I'm currently working on a mechanical system and I've hit somewhat of a brick wall.
The rules are as follows:
On a 1 min chart(but works with other time frames) when the 20EMA changes direction set entry to the break of that bar. Stop is set to low of first bar with additional tick. I also use a target of 5 points for DAX and say 20 points for Mini Dow.
Now as it stands with backtesting this system is profitable, even if it only has a 40-45% win probability.

The problem I have is that, although it is profitable, there are too many trades taken. On an average day you may have 3 wins, but 7 losses perhaps.

Does anyone have any ideas on how to reduce the no of losing trades with some sort of filter system?
I'd also be interested to hear if anyone else has a similar system?

I need something to spark the old marbles rolling around upstairs !

All suggestions/criticism welcomed!

Cheers
Chris

 

[TWI]

What do you consider a change of direction in EMA?

 

[cd173]

Hi twalker

If the EMA is pointing downwards and then turns upwards and vice versa. I use a stepped colour line for the 20 ema which changes colour when it changes direction (blue for up, red for down). So entry is on the break of the bar that changes colour........does that make sense???

Any ideas, or is it a dead duck...........?

Cheers
C

 

[TWI]

Since EMA is calculated every minute on a 1 minute chart this could theoretically lead to a change in direction every minute. Need to define what turn up or turn down means. If colour change occurs when the EMA has made a single higher/lower value than the last value, it may be possible to filter out some erroneous signals by considering only changes when EMA has made a number of consecutive higher/lower values.
Make sure you add bro and slippage to backtest results. In these types of very short term systems unless you are realistic about this it is doomed to failure.
No idea if this is a dead duck or not as I have not done much work with MA's

 

[cd173]

Thanks T

Just so I understand, only buy if signal is higher than last ema signal and vice versa for sell?
Something to experiment with. Also might try a points filter, say a break of 3 ticks before entry.
With regards to slippage, I tend to get negative slippage on stops, but positve slippage on target. Overall I tend to use slippage as 1 point, but in some cases I get none at all. I suppose it depends on time frame adopted.

Cheers
C

 

[jmreeve]

My experience of these sort of systems is that if you test them using small time interval bars they show a profit, however, if you run a full level1 simulation then the profit evaporates.

Tradestation is very good at giving false positives with this sort problem.

 

[cd173]

This particular backtest was done manually to check it's validity. It's more of a pain, but you get a better feel for the systems problems. I only trade the strongest hours of the trading day for a particular instrument.

C

 

[StockCoach]

Any system built off of inherently delayed information will always fail. Unless you are dealing with price and volume directly, you are hitting your head against the wall.

 

[minx]

Now as it stands with backtesting this system is profitable, even if it only has a 40-45% win probability.

The problem I have is that, although it is profitable, there are too many trades taken. On an average day you may have 3 wins, but 7 losses perhaps.

Have I missed something here but 3 wins and 7 losers doesnt give a Win rate of 40-45%, its 30%. Anyway, you could add another MA, perhaps use an EMA and a simple MA, just thinking out loud here, or use a 20 and a 25 so they both have to turn down, that might help.....

 

[minx]

With regards to slippage, I tend to get negative slippage on stops, but positve slippage on target.
C

How do you get positive slippage on targets? Do you only place your order once the tgt is reached, surely if the order is in the book then you cant get slippage?
When I backtest stuff I pencil in about 0.5-1.25 ticks slippage, depending on the strategy, but in reality I've found it to be more like zero and sometimes even positive over a month.

 

[StockCoach]

How do you get positive slippage on targets? Do you only place your order once the tgt is reached, surely if the order is in the book then you cant get slippage?
When I backtest stuff I pencil in about 0.5-1.25 ticks slippage, depending on the strategy, but in reality I've found it to be more like zero and sometimes even positive over a month.

Most likely this is because his stops are market orders, and his targets are limit orders. Fairly common approach.

 

[minx]

Most likely this is because his stops are market orders, and his targets are limit orders. Fairly common approach.

Yes but how can a limit order get positive slippage? If you put in an offer of 75 then thats what you get, not 76.....

 

[StockCoach]

http://invest-faq.com/articles/trade-price-impr.html

I thought he was trading stocks, which he isn't apparently. I'm not sure how his scenario is working out. But the time lag for routing orders could be the reason.

 

[cd173]

minx / StockCoach thanks for the feedback.

Just to clarify orders and slippage..I use an API for IB which has a pre-determined target at which to send a market order thru when it is hit. So, as I'm 'usually' selling into a rising market I generally get positive slippage. Not always though as it depends on market conditions.
For the stops I use a limitchase order which 'guns' an order thru after one or two seconds if the first order does not get filled at the set price, hence the negative slippage.

Thanks minx for the suggestions for filters. I only used the '3 wins 7 losers' as an off the head example.
There's something there perhaps, but I can't quite see it!

Cheers
C

 

[commanderco]

cd173

Personally, I think that you are barking up the wrong tree with your system.
However if you want to persist, why dont you just fade the entries.
The most important factor in a system is consistency.
If the system is consistently wrong, then just reverse and fade the signals.
Having said that, I still will not trade moving averages.

 

[alpha_monkey]

Any system built off of inherently delayed information will always fail. Unless you are dealing with price and volume directly, you are hitting your head against the wall.

On first pass, I don't completely agree. I'd argue what you are suggesting only holds in the context of very short term trading systems.

For me, "inherently delayed" data implicitly requires a timeframe to reference to. In other words, I think one would need to specify what the delay is relative to.

If you are talking about systems that attempt to trade off 1min & 4min EMAs, the granularity (in vague terms the number of ticks traded relative to the timeframe ) may be too great and your statement may hold. However, if you are using 1 min data to calculate/update 3 week & 8 week EMAs, I don't think the delay is relevant. Moreover, old school trend followers such as Dunn, JWH, Graham etc illustrate it is possible to build a robust, profitable systems for the long haul with delayed/historic data.

AM