How important is the filter?

Forums › ProRealTime English forum › General trading discussions › How important is the filter?

Viewing 11 posts - 1 through 11 (of 11 total)
  • 07/29/2018 at 12:47 PM #76999
    YachtTradaren

    Hello! As the topic complies – how important is the filtering in strategy developement? of course I understand that it depends on the type of strategy but I mean in generall speaking. From what ive seen, the indiacators such as: bollinger, rsi, MAs etc etc gives about the same win ratio signals. And the no lag “price action” parameters gives sligthly better results. I have an idea that the filtering parameter is absolutley crucial. If you could come up with a filter that is not curve fitted that suits your holding sight it shouldnt really matter (that much) wether you use the MAs/RSI or no lag param. Also this would suggest that a lot of the backtesting should be done with focus on the filtering purely.

    Am I right? Do I state something obvious? because I feel like im not.

    07/29/2018 at 1:10 PM #77001
    Vonasi
    YachtTradaren wrote:

    If you could come up with a filter that is not curve fitted

    Unfortunately every filter is curve fitted. We only have historical data to work with and every strategy we write is based purely on that. Every filter we apply is set to best deal with the past but we have no way to know if in the future markets will move in the same way.

    For something like a long only indices strategy a simple filter to keep us out of the market in major downturns may be a sensible thing as we have less chance of being right and more chance of being wrong during these times – but the more finely you try to tune that filter the better your historical results will look but not necessarily your future results – maybe even possibly the opposite. Something as simple as just staying out of the market if an index price is less than where it was a year ago and maybe also under a 200 day average can be a simple way with minimal curve fitting. You have just drawn a line in the sand and said those are the rules – it may not have been the perfect filter in back testing but we know that a lot of traders pay attention to where price was a year ago and to the 200 day moving average. Sometimes when price is near the one year look back price trading can get a bit choppy so you might want to set a second rule that you do not get back in until price is a certain percentage above your filter level.

    As for the importance of a filter that will depend on the strategy. I will normally write a long only price action strategy without a filter and then at the end add one to keep me out of the markets during major downturns when the strategy may still be profitable but the equity curve is generally more volatile and the biggest draw downs usually occur.

    07/29/2018 at 6:07 PM #77030
    YachtTradaren

    Great to hear from you Vonasi 😀

    Yes I understand that a too optimized filter would totally destroy a system. But the premise of constant profitable trading/investing in the long run, is that parts of the market dynamics are somewhat the same, right?. If the market would be completly random it would be gambling. The question that arises then is – what dynamics are more likely to not change that much?   I think many people think about filters as, for example, a high numb MA that historically fits to the major waves were the alternation between the slopes/price above-below etc etc will bring cuts out of those moves. But the standard use of  MA200 brings like what 25 iterations? I wouldnt be surprised at all if the global markets would go sideways for ten years. However I would be very surprised if the market on intraday basis would have moves of 1000 points up and down up and down. That is not likely to happen as people most likely would be scared and take money of the market and that would lead to a crash. And this theory has led to my conclusion that the short term dynamics are more (but not completly) resistant to change.

    Therefore im thinking if it would be a good idea to use a filter as a strategy within a strategy. Like if you would code a swing type strat with 1-2 parameters and use that as a filter. I just realized now that maybe this is the wrong definition of a filter. For someone with a profitiable swing strategy, couldnt that be used as filter for a intraday strategy? If the profitable swing strategy does not have too many parameters of course?

    These are just some thoughts I have, and Im a newbie so I might be wrong or this could be to complex for me to understand. But I would like to know if people agree with me or not and come with inputs 😀

    07/29/2018 at 7:29 PM #77036
    Vonasi

    I’m afraid that when I look at a chart it all looks pretty random to me and the faster the time frame the more random it looks. Try staring at a chart and predicting whether the next candle will be up or down and then try to guess how much up or down. You will be wrong as much as you are right which is pretty random. It is called Spread Betting for a reason because you are betting which way it will go and whether it will go more than the spread – so it is gambling.

    Maybe you have heard the story about a prize that was being given out in the early days of computers for the best program that could predict the weather. The best program simply said that today’s weather will be the same as yesterday’s weather and it was right more often than it was wrong. It is similar with markets.

    Screenshot_8-2

    …and that image is the FTSE100

    I find that if you find something that works on a slower time frame and then try to adapt it to a faster timeframe then it stops working. The markets get more random and the spread becomes much more important.

    07/30/2018 at 12:54 AM #77049
    Nobody

    If markets truly are random , what are you doing here  ?  Why are we bothering , now i got that out the way  .   Markets are random sometimes and not so random others . Key to this whole game is identifying when markets are one or the other  .   If markets are random how is it that algos with  winrates north of 65% with symmetrical  R/R  that perform over 60/70 years are possible  , Is this just pure luck or is it possible markets are not that random  ?    We spend a lot of time identifying patterns that repeat over and over  , and they do  with better chances of coin tosses . Is this not some sort of predictable behaviour  ?

     

    random

    [ran-duh m]
    adjective
    1. proceeding, made, or occurring without definite aim, reason, or pattern:the random selection of numbers.
    2. Statistics. of or characterizing a process of selection in which each item of a set has an equal probability of being chosen.

    When someone states an opinion as an absolute this is stating their limitations not mine  .  Markets are NOT random  for if you truly believed that you would not be here ???

     

    Filters are the questions we ask of price to build a decision tree / flow chart to mathematize price patterns so we can quantify said patterns . without filters there will be NO algo

    07/30/2018 at 7:57 AM #77052
    Vonasi

    All markets are random because they are effected by random events. Some markets more than others I agree.

    A currency is at the whims of supply and demand and at the whims of the governments on both sides of the market whose currencies are involved in the exchange. Just look at the recent performance of the Turkish Lira – one day it would be recovering slightly and then Turkey’s newest dictator Erdotwat would open his mouth and say he hated interest rates and the lira would collapse. Then the central bank would ignore him and raise rates and it would recover massively. No algo would be able to calculate when and what he was going to say or when or what the central bank would do – so that is a random event and a random market move.

    Commodities are also at the whims of random events. A poor harvest because of bad weather, a mine closed down due to striking workers or a war or the sudden discovery of how to extract cheap oil will all throw random price moves and changes in to the market.

    Indices take a random route upwards because they are based on economic growth of the biggest companies so should naturally go up – but what algo could calculate the date of the dot.com bubble or the 2008 banking melt down, or the great recession – the answer is none and so they were random events that threw the markets into downturns in what are markets that always goes up in the long run. The fact that indices always naturally wants to go up is why you can be right 65% of the time – because that is the natural bias.

    So yes our filters give us an edge in random markets. Only buy an index – never sell it is a filter. Get out of the market if an index falls below a certain level is a filter. These filters are what we use to make decisions and to protect us from the randomness of markets.

    If markets are not random then can you please post on here when the next major downturn is going to start so that we can all put it in our strategies as a filter date and time to stop trading?

    07/30/2018 at 8:21 AM #77053
    jebus89

    im on both teams here tbh!

    I belive the markets are random because random stuff happens, just take 9/11 which im gonna guess 99.99% of the world had no idea would happen that day. Thats a random event happening that caused a huge shift in the overall markets.

     

    But i belive that there are repeatably patterns happening in the trading that goes on in those said random markets. People react to trends and shifting markets in a very similar way across similar markets.

    So they key that im working at is the key to take the repeating patterns trades, while knowing that I would be “safe” even if a completly random event where to happen tomorrow, that would rock the markets.. A filter is what would keep my safe when shit hits the fan, but in the market when trends/patterns occur.

    just my 2 cents.

    07/30/2018 at 8:47 AM #77054
    Vonasi
    jebus89 wrote:

    i belive that there are repeatably patterns happening in the trading that goes on in those said random markets. People react to trends and shifting markets in a very similar way across similar markets.

    Exactly – take any market and look back in history and count the maximum number of months in a row that it went down. Now wait until it hits this number of down months in a row again and I’m pretty certain that you might see a bounce. The emphasis is on the ‘might’ as you might also see a new record number of months down in a row but the probability based on historical data is that you might be somewhere near a market bottom. A random event could be thrown in and ruin the whole probability but it is a pattern and many people will be looking at it and thinking the same thing so the probability of it doing what they all think is higher – if there is no random event or we are not under the influence of an earlier larger random event.

    It is interesting on some markets that if you have two down weeks then there is a higher chance of the next week being an up week but if week 3 is a down week then week four has a very high chance of being a down week. It is as if the end of week 2 is a deciding point and what happens in week 3 makes up peoples minds as to which way it is truly going. Unless a random event over powers peoples thinking.

    The above are patterns that might improve our probability of being right more than wrong in what is still overall a random market.

    07/30/2018 at 8:56 AM #77055
    Nobody

    “”  All markets are random because they are effected by random events. Some markets more than others I agree. “”  Its binary  , you cant have degrees of random  . The occurence of a black swan is random but what happens in between isnt    READ THE DEFINITION OF RANDOM , dont make up your version of it   …    Random is a coin toss  , there is an edge to be found in markets , thats 100% FACT  , just because you cant find it doesnt make it so  . AN EDGE means its  not RANDOM and if you think it is WHY are you here  .   Dont worry i ask WHY am i here as well ( albeit for diff reasons )  . A place to learn code and how to beat the market and find an edge  but the mod thinks its impossible , go figure   …

    random

    [ran-duh m]
    adjective
    1. proceeding, made, or occurring without definite aim, reason, or pattern:the random selection of numbers.
    2. Statistics. of or characterizing a process of selection in which each item of a set has an equal probability of being chosen.

    Taleb has a lot to answer for

     

     

    07/30/2018 at 9:15 AM #77059
    Vonasi

    Brisvegas – first of all I need to point out that your understanding of what a moderator is needs correcting. A moderator is the same as every other forum member on here except for the fact that they have been noted for the amount of time on the forums and the amount of help that they offer others and so have been given some basic cleaning tools to help keep the place tidy and deal with posts that might disrupt the normal flow of discussions on the forums. Moderators are allowed thoughts and opinions and allowed to express them – they just do it while holding on to a broom. 🙂

    You seem a little hung up on the definition of random. Markets are random because they are effected by random events – you do not need a dictionary to understand that. Yes we can find an edge in this world of randomness but we can also lose a lot of money thanks to this randomness.

    …and as for the mod thinking it is impossible you might want to read this post:

    https://www.prorealcode.com/topic/detail-report-results/#post-76946

    ….where you can see every trade on my demo accounts over a 90 day period and then see if you still think that I think that it is impossible.

    As a moderator I would suggest that at this point we return to the subject regarding filters that the OP first asked about rather than get hung up on our own personal definition of a random market….. 🙂

    07/30/2018 at 9:20 AM #77060
    Nobody

    You cant have it both ways    , im done here

  • Author
    Posts
Viewing 11 posts - 1 through 11 (of 11 total)

Find exclusive trading pro-tools on ProRealCode Marketplace

Similar topics:
adaptive kalman filter – can u convert pleases
adding a filter to a code
multiple filter criteria not working
triple moving average filter
usage of multiple filter criteria
what indicator/code to use to filter charts with many osc

Create your free account now and post your request to benefit from the help of the community
Register or Login