In the V7 version of Pathfinder, I think Reiner introduced a bug that wasn’t there before on this paragraph:
IF LONGONMARKET THEN
posProfit = (((close – positionprice) * pointvalue) * countofposition) / pipsize
ELSIF SHORTONMARKET THEN
posProfit = (((positionprice – close) * pointvalue) * countofposition) / pipsize
ENDIF
For short position, posProfit will be inverted, because countofposition is negative for short.
In my opinion, enough to use only “posProfit = (((close – positionprice) * pointvalue) * countofposition) / pipsize” for both trade direction
but backtest are worst with this change, which is counter intuitive. I have the same behaviour on one of my strategies, working better too with the “bug”. So I’m a bit puzzle, or I miss somehting.
For me it is most often like that. First I get a great looking result. Then I find that there something is wrong. After correction the result is worse.
Guys do you have any statistics of the general pathfinders trading system? To me honestly they look all pretty bad.. I am wondering wether it is just a lack of volatility issue or a plain overfitting issues. What are your thoughts? Many thanks Francesco
As Reiner himself wrote before : “Pathfinder’s backtests look brilliant but never forget they are all highly optimized and of course they are no money machines.”
I may add the following : 2017 is a difficult year to trade the DAX, because the general volatility is low and movements often occur erratically. Volatility indices such as VDAX have low average values in 2107 as not seen in the last 5 years .
To distinguish between overfitting and low volatility, you can try to re-optimize the pathfinder systems for the year 2017 only. When you get much better results for this single year by re-optimization, it is clearly a matter of overfitting if the systems did not work with the original parameters. When good results are not possible even by re-optimization for 2017, the reason may be low or bad volatility.
And finally, in physics, there is a saying : “Give me 5 points, and I’ll fit you an elephant.”
For trading systems, the same holds : “Give me 5 parameters, and I make you every trading system work.” But only in an idealized past…
Just started DAX, DOW und HS V7 Pathfinder this weekend in prod, and bamm .. this morning HS traded with a 1621€ profit. That is a very warm comeback present!
But I think the HS has quite a high leverage, since he made this profit just in 5 hours. I don’t complain 😉 .. just thinking about reducing the positionSize for this one.
wp01Participant
Master
@Pfeiler
You missed the other HS than? Both the limit reached today in a spike. Profit € 1.430 and + € 3.257. Total € 4.687 profit. Nice and needed this month after a sloppy period…..
Yup, sorely needed indeed.
Hey… there ARE more guys using the System 😉
Dow & ftse each +3 @ 21h.
Dax & HSI flat.
Anyone other Posis?
Hello,
What do you think about this…
I have tried the pathfinder dax v3 (who have been optimized and presented on april 2016).
Since this period, the result is not good.
it seem overfitted?!!
Hello, What do you think about this… I have tried the pathfinder dax v3 (who have been optimized and presented on april 2016). Since this period, the result is not good. it seem overfitted?!!
The same happened with most other pathfinder versions, too. Surely a result of overfitting, and in addition, of low DAX volatility in 2017. However, when only little or no losses are generated in periods of low volatility or sidewards movement, one may hope that the systems will work better when volatility increases again.
Wow i just stumbled across this thread , i have been working on a similar system for a couple years myself . Same principles but different code , i think the original code is exactly what ive been looking for tbh . Will take a while but i will read this entire thread and get back with any ideas/findings once im done . Ive applied this too many different instruments with stellar results in testing . This particluar test starts of slightly leveraged but is underlevereaged for the majority of time , Been working on a dynamic postion size in PRT but cant find a away to reference equity number in code , a way to make position size a variable
sort of like ‘n’ as position size
n = equity/close
Any clues on how to achieve this in PRT
Hello, What do you think about this… I have tried the pathfinder dax v3 (who have been optimized and presented on april 2016). Since this period, the result is not good. it seem overfitted?!!
Hi
Maybe optimized moving averages are not good. I personaly think that standard moving averages should be used (MA 10/20 / 50 / 100 /200).
Some principles are very good breakout and seasonality, time limitation of trades. I will try to use other moving averages, which is is key for non overfitted live results
Regards
Hi Pfeiler,
have just seen your mail. Pathfinder for SAF is working well in 2017 so far. Enclosed the requested V7 (trade only long positions).
regards,
Reiner
Hello Brisvegas
Try this piece of Code
// smart position sizing
Capital = 10000
Risk = 0.01
StopCoefficient = 75
equity = Capital + StrategyProfit
maxrisk = round(equity * Risk)
PositionSize = abs(round((maxrisk / StopCoefficient) / PointValue))
//ONCE PositionSize = 1
It is part of an old Pathfinder version and leadds to huge gains but also intense draw downs. Maybe you can use and optimise it.
Hi guys,
has anybody the V7 ftse and also NOT the opening of 6Pos @335.9 on Dec 4th? I am confused…
