|
PureBytes Links
Trading Reference Links
|
Well there doesn't seem to be much interest in benchmarking, or the philosophy of optimization .... perhaps a few are just observing, or the forum doesn't know what to make of it all?
Our discussion has moved over a little bit ... it is more to do with style and finding the 'holy grail' of trading, which some say doesn't exist, but, like Santa Claus, why spoil a good fairytale that so many get so much enjoyment from.
This has become slightly more personal but then again I believe in transperance and sharing our trading insights (if RalphVince, Markowitz etc hadn't published where would I be?) and surely our observations must have relevance to a certain number of traders.
I think I am preaching to the converted now.
Right from the start my style was always builtin and I was already biased towards it .... I just didn't become consciously aware of what my style was until I had a fair amount of experience under my belt.... as soon as I became consciously aware of what worked best for me the rest was easy and still is.
We can not change our natural style, only discover and enhance it.
This forum is naturally biased towards algorithmic trading etc and I have deliberately infringed on the OT boundaries from time to time, to stick up for my style and my kind of people (discussions on the Psychology of Trading, Discretionary Trading, Intuition etc) ... continual discussion of code examples etc gets a little boring and somewhat inhuman at times, so I like to inject a little human interest (small personal exchanges with people I like ... Haiku poetry and suchlike).
The interesting thing is that even though I understood myself reasonably well, before I started trading, the journey has still been one of immense self-discovery, which is why I don't confine my discussion to just 'learning to code'.
In order to simplify my perspective, for the benefit of others, I have said that we have our psychological typologies, that we can sort everyone into, generally speaking.
I could use some different classifications but for consistency I have stuck to one simple model i.e. I have told the forum that my primary psychic function is Intuition and my secondary is Logic (of course this is a gross simplification but we have to start somewhere).
I have also told the forum that Intuition has been poorly represented by the scientific community and modern culture (we use the term intuition as if it is an inferior function).
IMO intuition is more like super rationality (the rational process goes on, almost unconsciously, at lightning speed and it uses processes akin to 'fuzzy logic' and more).
I also warned the forum that intuition is not fortune telling or 'reading the tealeaves' and so it is not infallible and like the 'scientific method' the outcomes depend on the practitioners skill and knowledge with the mode (it is also very energy dependent, and self aware/sirected so it has an auto failsafe when the energy pack is drained).
So, when 'visual' traders say they just look at the charts some people assume that the conclusions that we arrive at are all just 'subjective' and hence irrational.... in some cases that may be true but not necessarily in all cases.
Some people think they are using their intuition when they are just jumping to irrational conclusions... they call this their intuition and this is why it gets such a bad name.
In short, science, and our community, are greatly underestimating the capacity of the human mind (far superior to any computer built so far).
I think people also struggle with the possibilty that anyone can be highly creative (artistic) and totally logical (objective) all in the one package but this is possible (we just don't do them both at the same time although we can quite easily switch back and forth between the modes ... but not at the drop of a hat).
So, without analysing it all, our styles are quite similar, except that from what you say, although a few snippets don't allow me a full analysis, you are even more of a 'visual' trader than I am.
I am still developing, and learning, and I have more 'good ideas' than I will ever need ... you are probably the same.
However I will go through your comments with a view to sharing with you a couple of my undeveloped ideas that seem relevant to your interests.... just in case you haven't thought of them (no two individuals are identical).
- I call eSignal eS .. I have got you now on the trading ES part
- I said before in this forum that if we excluded clip trading (arbitrage) and trades that benefit from no 'volatility', like selling options, then we are all 'trend traders' i.e. if we can find the trend (and I tend to the view that the trend doesn't exist ... if I model a trend, for any particular purpose, I always assume that it won't exist for long).
- I also assume that the market doesn't have much of a memory and so nothing lasts (I consider that this models the markets as being dynamic i.e. dynamism is a fundamental quality of the markets)
- however, we have to deal with the reality of different timeframes, so each timeframe will have a memory of different duration (in time but not in bars)
- so generally I consider the 'holographic' like inter-relationship of the various timeframes as being a second fundamental quality of the markets
- I use a wave concept (cycles are a third fundamental quality of the markets) but I don't use Fibonacci or Elliot etc because I don't use a set magnitude and frequency ... uncertainty is a fourth fundamental quality of the markets (so I only have a probalistic expectancy as to what the frequency and magnitude will be).... for that reason Fib retracements are no use to me.
- from our recent discussions on randomness etc I am considering the proposition that a certain amount of chaos needs to be injected into financial models ... this is a new and developing idea for me .. probably the creative mathmeticians. who are around, are right on to this ... I am tentatively proposing that there has been a burgeoning use of computer analysis since around 2000 and that this is producing some chaotic influences not previously seen .... post approx 2000 the markets are different?....(computers are binary ... unsynchronized binary events may be an approximate model of limited and man made chaos ... I am only considering this within the context of the financial markets) ... if I am correct we might be in for a rough ride for a decade or two .. eventually financial theorists, and the lawmakers, will have to catchup and change the model somehow (regulation or freemarket driven changes or what?) ... unfortunately the normative view always lags and opposes the Sigma 3+ position (of course conservatism is necessary to prevent unbounded chaos in our culture ... ironic isn't it that this mechanism is also the cause of some unneccesary pain but that's life).
- in the above scenario 'old world models' will not do as well as current models (due to computationally driven investing/trading 'old' may well only be 10 years ago, or less).
>I never have to worry about if the
> liquidity is there and if the slippage is accurate to my >simulation.
I am still considering the uses that I can find for high liquidity and low liquidity ... every thing has its' pros and cons?
In my 'world view' the market has endemic behaviours and I am interested in finding out everything about them that I can.
If you factor in 'slippage' you might miss the endemic behaviours in lower timeframes (where slippage and commissions are a greater % of the move ... i.e. including commissions and slippage etc might kill some theoretically good trades and you will miss the connections that exist from one timeframe to another ) ... IMO we should do our thinking, and design our prototypes, without including C&S in our models ... only include C & S when we go into production testing (I got the terms prototype and production from bruceR .. I really like them ... I have recently found a new love of words ... very important things because they represent an idea ... Yuki will be pleased to see I am paying more attention to the words I use ... now I only need to learn some proper grammar and I will almost be there).
> fundamental data is of little consequence. Only the immediate >news
> has a non-technical impact, and that shows up within one minute in >the chart anyway.
Fundamental and value investing is a valid style ... look at Buffet, he has proved the success of valuie investing over and over ... of course even Buffet makes mistakes ... it is just not our style.... we can't eat all of the cake on the table so lets' leave fundamental analysis to the righful owners.
Instrument selection is important though .. I have to replace fundamental analysis, as a way to chose the instrument, with something else .. haven't worked hard on that aspect (note that Reefbreak has his algorithm to sort for his underlyings ... I am sure it is not an alogrithm that sorts by fundmentals).
Definitely I am a classical technical analyst in that I believe I can only think fast enough to react to price action (even with a computer doing the thinking) and can not get handle all of the data input and computations required to stay ahead of the news that makes the moves ... however last year when the big crisis was on I experimented with some news following (economists newsletters etc) and I actually did quite well at picking how it would all unfold ... after it was all over I just left it at that because if I try to master several styles I will end up being a 'jack of all trades and master of none' ... I did enough to prove to myself that it can work.
> Because my trading universe is so limited in scope,
I also have a tendency to filter everything down to elemental simplicity .. works for me.
I also believe in detailed analysis of a very small patch of the available trading turf.
Once before in the forum I talked about the fact that I tentatively believe that there is only ONE trade and that we are all chasing with different degrees of efficiency (of course we have different markets and timeframes to chase it in).
I have isolated market behaviour to some simple repetitive behaviours - they are persistent in most timeframes and all instruments I have tested.
I was only working down to a 1min base (with eS == eSignal data) ... I don't like to work lower than that because at tick level etc my computer starts to misbehave ... too much number crunching ... I don't really want to get into buying and managing the fastest computer on the planet so I keep away from that level ... some people keep going to faster and faster (lower) data levels simply because they arent' succesful at one level and think the answer is in getting faster and faster == wrong! ... also I am sure some people don't understand how computer, or software gridlock, kills the creative effort .. hence my dislike for technical software glitches or computer glitches ... I keep it simple to allow the intuitive energies to function.
Re timeframes and persistence patterns:
- intitially I tested down to 1min base with 5 min selected
- I found my endemic patterns (I don't call them fractals because they might not be ... I always own the ideas and choose my own nomenclature, rather than become a Mandelbrot clone for example) ... I found they are persistent from monthly (albeit this is a limited dataset) down to 5 mins (very large datasets) ... they started to breakup at 1 min .. I thought this might be something to do with the 'snapshot' nature of the basetimeframe i.e we should never work at the basetimeframe ... always set the base a little below where you want to work and then compress the underlying to where you want to be .. sure enough when I downloaded 5 sec and tested in 15sec the patterns are back.
Note that I have something funny with my low level eS RT databases so that could have contributed to this effect... timestamp minimum in my fast eS databases is 5 sec ... I haven't followed up to see if it is an eS or an AB thing ... probably eS because AB is good to 1 sec? ... I am subscribed to NinjaTrader data so maybe I didn't read the fineprint and their eS variant is snapshot data ... perhaps I need to go back to the standard eS subscription.
>I was able to
> write my own back tester in AFL that only runs in indicator mode >with
> the equity curve always showing in my chart.
Yes, I also 'backtest' outside of the 'backtester' ... for one thing I don't like AB's BT .. it is not an intuitive model (nothing wrong with it technically but I can't see the wood for the trees when I use it).
It is funny that I left Metastock for AB so that I could backtest with customized stops (that was my only reason at that time) ... but since I have owned AB I have hardly ever backtested ... as my trading philsophy evolved, and I used it to make better and better predictions/systems, I was able to think of much smarter ways to perform my objective testing .. recently in AB I repeated some testing in a week that took me 1 year in Metastock ... all because AB is faster and the logic behind my algorithm was light years ahead of what I did in 2004.
I have talked a bit, in the forum, about BT design and effective metrics (CoreMetrics) ... maybe a few get it ... not sure if there is much point in going further with that type of discussion (StDev1 always dominates any subset of our culture and the dominant culture itself).
Anyway, I am quite well aware that my posts are long, ambiguous and dispersed in time and location .. in fact my communication style is almost symbolic at times ... I have considered all along that I am mainly writing 'for my own' and that they can interpret the code.
> I overlay a lot of
> indicators and stats on my charts as desired to gain greater >insights
> about what is happening.
You haven't mentioned any specific metrics ... perhaps you want to keep them secret or more likely you don't use any, or many ... my secondary psychic function is logic so I drop onto that wavelength as soon as I come out of brainstorming or sometimes interpose both... hence my interest in stats and my sort of compatibility with Howard and Patrick ...except I don't love stats/quants as much as them and my quant is always coloured with intuitive stuff (I make it as easy as eating moms apple pie) .. so far every stat I have used I have boiled down to a readers digest version of the academic function.
You might benefit from following my ruminations on stats e.g. have you thought about how adaptiveStDev (thanks Herman, bruceR, RZ) can work with median, mode, skew etc to model your trade dists, as you walk forward ... all done in arrays with no binning required (assuming RV's math is good and it hasn't failed me anywhere so far)... did you see the link I posted from the German site about how these moments model our system, via the trade series ... I still have some work to do on this but I am slowly moving along with a rebaking of the evaluation pie.
I am just checking that you are not underdone on stats ... they are great as long as they are restated into traderspeak.
I hope you got something out of my ruminations on BiSim and CoreMetrics .. anyway I will probably post a few more bits and pieces on that subject to help the very interested get to the meat in the sandwich .....but once again anyone who is going to get it should have got it by now.
> The recent addition of static arrays was a great help for me in >saving
> the temp results of previous runs for equity curve comparisons.
Yep, that was the idea.
I haven't tried the function out yet ... I don't take the betas ... too much administrative effort for me .. I wait for the upgrades that come with some help ... I think Tomasz underestimates the reliance I (certain types of people) have on help notes e.g. intuitives start with the meaning i.e. first of all, what does this mean? .. what contextual environment does this fit into? .. OK ... now how do I use it?)
Sometime the logicos version of the help manual is very frustrating for me.
We should also thank Tomasz very much for having the perspicacity to grab hold of the idea and implement it ... this level of responsiveness to user discussion is very rare (thanks once again Tomasz).
> My best system is a revision to mean system so far. For this >system,
You might be missing something ... if the trend really does exist then 'reversion to mean' will underperform 'following the mean' ... if it doesn't exist then trading both ways will net out to being equal in the long term.
Going back to the HolyGrail ... I only have one system but I can vary it in several ways ... your methods should have lead you to my system, or very close to it, so we must be almost talking about the same thing.
You seem to have been mesmerized a tiny bit by the dance of the seven veils .. I was hinting at this in the discussion on MA and habituation:
- I don't like it that Howard refers to 'reversion to mean' and only references on side .. I actually think of 'erversion to mean as crossing the mean from above.
- when we look at any chart, without indicatros, it is meaningless
- we have a strong natural desire to see meaning in it .. so strong we can even overlay false, or approximate meanings, and feel very satisfied with our efforts
- MA is very potent ... it 'identifies' the trend (yes! won't be long now before I am rich) and makes the trading universe symmetrical ... how beautiful is that? i.e. it divides the chart into above and below which has a satisfying logic to the rational (with a small r) aspect of our mind (I went on to talk about the powerful reinforcing effect of what feels right == habituation).
No model is forbidden in my approach .... as long as we then go onto analyze the pros and cons of the model which leads to an assessment of whether we can gain an edge out of them .. anyone of them .. and from there we then have to decide if the edge si significant enough to trade i.e. quantify it.
So, MA has a reversion to mean coming from above and below ... how close is this model to the HolyGrail? ... is there an edge to be had from focussing on one direction of the reversion as compared to the other?
Tading with the trend always beats countertrend trading (this is a better name than reversion to mean) .. as long as the trend persists.
Of course we always pay the price when trends change.....and they always do this quickly, in their own timeframe ... we pay the price because we can never achieve theoretically achieve the perfect trade, let alone achieve it in real trading ... when we go into production, with our prototypes, C&S detract from the perfect trade even more.
Re volatility:
- I am still learning about it ...more to do ... very important (thanks to JohnBollinger who first introduced me to the subject via BB's ... been hooked ever since).
> I start with the philosophy, and work towards the optimization from
> there. In the beginning, it helps to do it the other way around,
> until you learn the relationships between parameters. Mostly it is
> staring at charts with indicators overlaid and asking yourself if
> there is anything significant about the patterns you see, then program
> the ideas in and see what it gives you.
I agree.
Just to refine the details:
- which came first the chicken or the egg?
- my trading philosophy has developed in tandem with my struggle to find ways to make a buck.... one seems to shape the other ... I definitely think everything through in detail before moving on to testing for confirmation .. if my idea isn't confirmed I go back to more thinking to find out why it failed etc.
- an analogy is that if we want to hunt the Tiger we could go into our office, read all of the research and run simultions of our plans to formulate a plan of action OR we could go into the field and sit quietly in the hide for extreme periods of time ... I do this frequently, take notes on all of the observed behaviours and then when I go back to the lab I start to hypothesis, based on my notes.
When I reference others trading research/opinion I cross check their observations and theorems with my field observations .. if it doesn't stack up I reject their work, or parts of it.
>It also helps to manually
> trade your ideas some. The perspective of what can and will go >wrong
> in the real world is quite eye opening.
Yes, I am at the point where I might not Backtest much more, ever .. it isn't imperative that I do so anymore ... an exception might be if I develop a MatrixBacktester for fun .. then I would find some uses for it, at least for a while.
I have accepted that my methods, combined with using paper trading as the OOS test, are perfectly adequate for me.
As an aside I am sceptical that there is evne such a thing as OOS when we use historical data ... for one thing we are all too widely read to be naive about any trading idea so in that sense we have all walked over any historical data, we care to get our hands on, thousands of times .. in that sense live trading is the only 'real' OutOfSample data.
Thanks for sharing .. very helpful for me.
I hope you get something specific out of my commentary or that at least it sparks of some fruitful creative thinking for you (some commentators have tried to filter the qualities that make a top trader/investor but of course it is very simple .. they are highly creative which exhibits as a passion for investing/trading).... like passionate golfers .. we will talk to other passionate traders anywhere at anytime.
I have done a fair amount of commentating on TradingPsychology ... it is spread all around the forum .. this might be the last in the series ... I am not certain about that but I am starting to get bored with it and the forum must be getting bored with it .. anyone who was going to get it should have got it by now (and I am very realistic about that) ...tough luck for newcomers ... they miss out.
Still, I never say never.
All the best with your trading efforts..
brian.
--- In amibroker@xxxxxxxxxxxxxxx, Dennis Brown <see3d@xxx> wrote:
>
> Brian,
>
> First, remember, I am only trading ES, the e-mini S&P 500. That
> simplifies my problem immensely. I never have to worry about if the
> liquidity is there and if the slippage is accurate to my simulation.
> They are known quantities. I also only look at 1 minute bars, and do
> not hold overnight. My trades rarely last over an hour, so
> fundamental data is of little consequence. Only the immediate news
> has a non-technical impact, and that shows up within one minute in the
> chart anyway.
>
> Because my trading universe is so limited in scope, I was able to
> write my own back tester in AFL that only runs in indicator mode with
> the equity curve always showing in my chart. I overlay a lot of
> indicators and stats on my charts as desired to gain greater insights
> about what is happening.
>
> So my secret is to simplify the unknowns to the smallest universe
> possible and specialize on just one kind of trade. This gives me a
> fighting chance of actually understanding what I am doing.
>
> The recent addition of static arrays was a great help for me in saving
> the temp results of previous runs for equity curve comparisons. I
> manually control every aspect of parameter changes and selection of
> which curves to save. I also take a lot of screen shots of my charts
> for later comparison purposes.
>
> My best system is a revision to mean system so far. For this system,
> volatility is good and trends are bad. I am currently working on a
> trend following version to see what I can do with that. In that case
> trends are good. I am only talking about the action over one day of
> course.
>
> I don't rely on mathematical straightness, because the ES does not
> give equal opportunity all the time. For a reversion to mean system,
> volatility gives more opportunity. I measure volatility and apply
> that value to modify different parameters on the fly.
>
> I start with the philosophy, and work towards the optimization from
> there. In the beginning, it helps to do it the other way around,
> until you learn the relationships between parameters. Mostly it is
> staring at charts with indicators overlaid and asking yourself if
> there is anything significant about the patterns you see, then program
> the ideas in and see what it gives you. It also helps to manually
> trade your ideas some. The perspective of what can and will go wrong
> in the real world is quite eye opening.
>
> BR,
> Dennis
>
> On Jun 20, 2009, at 12:50 AM, brian_z111 wrote:
>
> > Thanks .. its great to get some feedback about how people are
> > actually going about their evaluation.
> >
> > I notice that you are using visual methods, rather than a metric, to
> > select your top model.
> >
> > For my first optimization I didn't look at any eq curves ... this is
> > not the default in AB's opt?
> >
> > How do I create and plot the relative curves for all of the possible
> > combinations ...d o you limit this to subsets to save time OR
> > perhaps from your later comments, you don't search all candidates
> > but start in a chart and then manually add plots to test the
> > combinations you fancy?
> >
> >> I also block out the
> >> best performing times to concentrate effort on the low performing
> >> times as part of my process.
> >
> > I agree that this is an area worth focusing on.
> >
> > I did notice that AB only optimizes the 'concatenated' data, as a
> > portfolio, (at least as I understand it ... couldn't find any
> > guidance on this in the help manual or Howards' books .. trial and
> > error seems to indicate the opt report is a 'one in all in' approach
> > so I couldn't differentiate under/over performance relative to under/
> > over performing stocks without a special effort on my part).
> >
> >> (last Fall was
> >> a great addition to the max volatility set).
> >
> > Yes. I noticed that if I run the MACrossover opt on monthly data,
> > for the entire 1970-present range, then it still qualifies as a
> > trendfollower (just).. IOW the dip of 2008, wasn't quite big enough
> > to take us out of an uptrend, from the long term perspective, so
> > bull trend following systems still work in that timeframe/timeslice.
> >
> > I didn't report on it but when I ran an MACrossover opt on the
> > random dataset that I produced (designed for and run in EOD data)
> > the spread of the optimized results is pretty tight and low (not
> > significant) ... using ABs' total portfolio approach .. this is
> > reassuring since the concatenated random datasets converge on the
> > mean of zero returns (half the datasets underperform and half over
> > perform with low volatility) ... I wonder if that will change if I
> > introduce some volatility ... I'll have to try it.
> >
> >> I overlay my equity curves on top of
> >> each other and look for the straightest curve relative to the >market
> >> opportunities (volatility).
> >
> > You are not confident of the metrics that measure straightness or
> > you just prefer the eyeball method (I found it sort of ironic, or
> > something, that Howard also likes to eyeball the curves).
> >
> > Volatility could be good or bad ... say it is all over the shop ..
> > don't you want trendiness with volatility?
> >
> > Are you measuring or eyeballing volatility?
> > Do you filter, by volatility, to select the instrument to trade?
> >
> >> When rare
> >> events hurt the performance, I zoom in on those trades and try to
> >> understand what about the underlying algorithm lets it happen.
> >> >Then,
> >> I think about how I can make my algorithm smarter in the general case
> >> to reduce these types of problems.
> >
> > If another line of code removes a significant number of losers then
> > it is likely to be generic going forward?
> >
> >> I do a lot more thinking than testing and optimizing
> >
> > That was my initial reaction to my optimizing experience ... that if
> > we enter some parameters and send the computer off to search then
> > what comes back might not have a lot of meaning, in terms of a
> > trading philosophy, whereas if we are working at developing a
> > trading philosphy first, then later on opt can help develop/test
> > systems derived from the philopsphy.
> >
> > It takes a long time , and a certain skill, to develop a trading
> > philosophy, but anyone can quickly learn to run an opt, with or
> > without applying a lot of thought to what they are doing.
> >
> > --- In amibroker@xxxxxxxxxxxxxxx, Dennis Brown <see3d@> wrote:
> >>
> >> Brian,
> >>
> >> When I optimize (and I am only working with ES), I watch the complete
> >> equity curve over more than a thousand trades under all market
> >> conditions of volatility and trends in both directions (last Fall was
> >> a great addition to the max volatility set). I also block out the
> >> best performing times to concentrate effort on the low performing
> >> times as part of my process. I overlay my equity curves on top of
> >> each other and look for the straightest curve relative to the market
> >> opportunities (volatility). In other words, the slope of the equity
> >> increases with the volatility. If an optimization step generates a
> >> better profit by virtue of clipping off a big drawdown, or other rare
> >> event, then it is over optimization and I reject it as just data
> >> mining. However, if an optimization step results in a steady,
> >> constantly deviating increase in outcome over all market conditions,
> >> then I accept it as a fundamentally good optimization. When rare
> >> events hurt the performance, I zoom in on those trades and try to
> >> understand what about the underlying algorithm lets it happen. Then,
> >> I think about how I can make my algorithm smarter in the general case
> >> to reduce these types of problems. I do a lot more thinking than
> >> testing and optimizing --which I do by hand with parameters, so I
> >> know
> >> what the relationships are intuitively after a while. I do not
> >> consider it cheating to have parameters that adjust themselves to
> >> general market conditions like high or low volatility, etc., just as
> >> long as the algorithms are very general and make logical sense
> >> regardless of the data.
> >>
> >> It is a slow process, but when I am done, I have an algorithm and
> >> settings that are robust to whatever the market throws at me. I
> >> don't like being fooled by randomness!
> >>
> >> BR,
> >> Dennis
> >>
> >> On Jun 19, 2009, at 8:10 PM, brian_z111 wrote:
> >>
> >>> OR
> >>>
> >>> ... is opt correctly flagging something about market behaviour or
> >>> system trading that I don't understand?
> >>>
> >>>
> >>> OR
> >>>
> >>>
> >>> ... all of the above, none of the above, a combination of some of
> >>> the above?
> >>>
> >>>
> >>>
> >>> Also, the MACrossover changed from TrendFollowing to MeanReversion
> >>> when I added 1.5 years to a 30 year lookback and not a 10 year
> >>> lookback ... not a 9 year lookback as per my previous post.
> >>>
> >>>
> >>> --- In amibroker@xxxxxxxxxxxxxxx, "brian_z111" <brian_z111@> wrote:
> >>>>
> >>>> So, in my first ever attempt at optimization I am presented with a
> >>>> conundrum.
> >>>>
> >>>> If I opt MACrossver(C,X), and look down the list of top candidates
> >>>> (using the AB default objective function), I see that historically
> >>>> this system was both a 'trend following system' and a 'reversion to
> >>>> mean' system.
> >>>>
> >>>> If I could travel back in time, armed with this info, should I
> >>>> trade MA crosses as a 'trend follower', 'reversion to mean' or
> >>>> both?
> >>>>
> >>>> OR
> >>>>
> >>>> ...on the other hand am I failing to interpret the results
> >>>> correctly ... is there something about opt that I don't
> >>>> understand ... if I develop my opt skills will this help me solve
> >>>> this riddle?
> >>>>
> >>>> OR
> >>>>
> >>>> ... is optimization itself somehow not providing me with a clear
> >>>> understanding of how I should have followed the markets (for that
> >>>> historical period)?
> >>>>
> >>>> OR
> >>>>
> >>>> ... is it something to do with MACrossovers ... perhaps other
> >>>> systems are more amenable to optimization ... if so how can I
> >>>> filter systems in advance to save me wasting my time opting non-
> >>>> compliant systems?
> >>>>
> >>>> OR
> >>>>
> >>>> .... is the objective function the underlying cause of this
> >>>> dilemna?
> >>>>
> >>>> OR
> >>>>
> >>>> ... I have done something wrong .. failed to optimize correctly or
> >>>> used AB incorrectly?
> >>>>
> >>>> OR
> >>>>
> >>>>
> >>>> ... is it the data... is there something wrong with the Yahoo data
> >>>> I used?
> >>>>
> >>>> --- In amibroker@xxxxxxxxxxxxxxx, "brian_z111" <brian_z111@> wrote:
> >>>>>
> >>>>> What I am headlining here is that looking back,at 9 years of data,
> >>>>> the best MA system was 'trendfollowing' then, only 1.5 years later
> >>>>> the best MA system was turned completely upside down into a
> >>>>> 'reversion to mean' system ??????
> >>>>>
> >>>>> Say, what?
> >>>>>
> >>>>> I don't have an explanation for any of this but it is too early
> >>>>> anyway ... I need to make a lot more observations before it is
> >>>>> time to hypothesis.
> >>>>>
> >>>>> --- In amibroker@xxxxxxxxxxxxxxx, "brian_z111" <brian_z111@>
> >>>>> wrote:
> >>>>>>
> >>>>>> This is the first time I have optimized or used any kind of
> >>>>>> synthetic data in AB ... so far I haven't used any sophisticated
> >>>>>> methods to produce synthtetic data either.
> >>>>>>
> >>>>>> I have only done a small amount of testing but I immediately
> >>>>>> found three anomalies that might be worth further investigation.
> >>>>>> I have aleady reported bcak on two:
> >>>>>>
> >>>>>> - why does an apparently worthless 'system' (plucked out of thin
> >>>>>> air unless my subconscious mind intervened) outperfrom on approx
> >>>>>> 6% of stocks when those stocks are assumed to be correlated to a
> >>>>>> fair extent ... chance? OR some property of the data that
> >>>>>> correlation does not measure ... what property of the data would
> >>>>>> favour that randomly selected 'system'?
> >>>>>>
> >>>>>> Note. if anything I expected the system to test the assumption
> >>>>>> that the MA is the trend and I expected the system to 'fail'.
> >>>>>>
> >>>>>>
> >>>>>> - why does the same system then outperform approx 50% of the time
> >>>>>> when tested over randomly generated price series ... is it a
> >>>>>> coincidence that the outperformance ratio, on random data, is
> >>>>>> close to the expected for randomness? and why didn't the bull
> >>>>>> 'system' outperform only on the random price series that
> >>>>>> outperformed?
> >>>>>>
> >>>>>>
> >>>>>> The third anomaly is:
> >>>>>>
> >>>>>> - I optimized the following on some Yahoo ^DJI data ... 10 years
> >>>>>> EOD ... 9951 quotes ... 2/01/1970 until June 4th 2009.
> >>>>>> Default objective (fitness) function = CAR/MDD.
> >>>>>>
> >>>>>> fast = Optimize( "MA Fast", 1, 1, 30, 1 );
> >>>>>> slow = Optimize("MA Slow", 1, 1, 30, 1 );
> >>>>>>
> >>>>>> Buy = Cross(MA(C,fast),MA(C,slow));
> >>>>>> Sell = Cross(MA(C,slow),MA(C,fast));
> >>>>>>
> >>>>>> - when I optimized on the total range I found that the top
> >>>>>> values, were inverted (as per Howard's examples in this forum and
> >>>>>> his books) but when I left out the 2008/09 extreme market
> >>>>>> conditions I found this did not hold.
> >>>>>>
> >>>>>> Why does sucn a relatively small change in the test range make
> >>>>>> such a radical difference in the outcomes?
> >>>>>>
> >>>>>> Here are some of the reported metrics from AB .. notice that they
> >>>>>> are similar in some cases and markedly dissimilar in others.
> >>>>>>
> >>>>>> I am not sure if that leads to a question but it certainly gets
> >>>>>> my attention considering that I am in the business of engineering
> >>>>>> reward/risk.
> >>>>>>
> >>>>>> Note that I am using ProfitFactor because it is typical in the
> >>>>>> industry but it has some question marks over whether is it the
> >>>>>> best CoreMetric to use (I am investgating PowerFactor and
> >>>>>> assymetricalPayoffRatio which might be more apt ... I hope to
> >>>>>> post more on these metrics later).
> >>>>>>
> >>>>>> Opt1:
> >>>>>>
> >>>>>> using all data
> >>>>>>
> >>>>>> top model = CAR/MDD == 0.25 AND periods == fast 10, slow 7;
> >>>>>>
> >>>>>> NETT PROFIT 1749%
> >>>>>> Exposure 44.9;
> >>>>>> CAR 7.68;
> >>>>>> RAR 17.07
> >>>>>> MAXDD 31.57
> >>>>>> RECOVERYFACTOR 2.64
> >>>>>> PF 1.62 (WIN 68% * PR 0.75)
> >>>>>> #TRADES 588
> >>>>>>
> >>>>>>
> >>>>>> Opt2:
> >>>>>>
> >>>>>> using data range from 2/01/1970 to 31/12/2007 (that's Dec for
> >>>>>> the benefit of timezoners).
> >>>>>>
> >>>>>> top model = CAR/MDD == 0.42 AND periods == fast 1, slow 6;
> >>>>>>
> >>>>>> NETT PROFIT 1921%
> >>>>>> Exposure 54.51;
> >>>>>> CAR 8.23;
> >>>>>> RAR 15.09
> >>>>>> MAXDD 21.67
> >>>>>> RECOVERYFACTOR 5.07
> >>>>>> PF 1.35 (WIN 40% * PR 2.03)
> >>>>>> #TRADES 1049
> >>>>>>
> >>>>>>
> >>>>>> I hope I reported the metrics correctly but anyone can replicate
> >>>>>> my tests and report otherwise.
> >>>>>>
> >>>>>> I think it also demonstrates that if PoF (PowerFactor) is a
> >>>>>> better CoreMetric than ProfitFactor it will need to be
> >>>>>> standardized on a returns/time basis (choose your time period =
> >>>>>> the basetimeframe you trade ... PoF is related to GeoMean per
> >>>>>> bar?)
> >>>>>>
> >>>>>> --- In amibroker@xxxxxxxxxxxxxxx, "brian_z111" <brian_z111@>
> >>>>>> wrote:
> >>>>>>>
> >>>>>>> Following recent discussions on benchmarking and using rule
> >>>>>>> based systems to engineer returns to meet 'clients' profiles
> >>>>>>> i.e.Samantha's MA(C,10) example, I did some follow up R&D with
> >>>>>>> the intent of expanding the examination a little further via a
> >>>>>>> zboard post.
> >>>>>>>
> >>>>>>> I may, or may not, get around to that so in the meantime I
> >>>>>>> decided I would share a couple of things while they are still
> >>>>>>> topical.
> >>>>>>>
> >>>>>>> I made up some quick and dirty randomly generated eq curves so
> >>>>>>> that I could optimise MA(C,10) on them (out of curiosity).
> >>>>>>>
> >>>>>>> Also, out of curiosity, I decided to see how the example signal/
> >>>>>>> filter code that I made up, as the study piece for Yofas topic
> >>>>>>> on benchmarking, would actually perform.
> >>>>>>>
> >>>>>>> Buy = Ref(ROC(MA(C,1),1),-1) < 0 AND ROC(MA(C,1),1) > 0 AND
> >>>>>>> ROC(MA(C,10),1) > 0;
> >>>>>>> Sell = Cross(MA(C,10),C);//no thought went into this exit and I
> >>>>>>> haven't tried any optimization of the entry or the exit
> >>>>>>>
> >>>>>>> By chance I noticed that it outperformed on one or two of the
> >>>>>>> constituents of the ^DJI (Yahoo data ... 2005 to 2009) and to
> >>>>>>> the naked eye the constituents all seem to be correlated to a
> >>>>>>> fair extent over that time range.
> >>>>>>>
> >>>>>>> Also, to the naked eye, it outperforms on randomly generated
> >>>>>>> stock prices around 50% of the time and the outperformnce
> >>>>>>> doesn't appear to be correlated to the underlying(I haven't
> >>>>>>> attempted to find an explanation for this).
> >>>>>>>
> >>>>>>> Here is the code I used to make up some randomly generated
> >>>>>>> 'stocks'.
> >>>>>>>
> >>>>>>> As we would expect it produces, say, 100 price series with a
> >>>>>>> concatenated mean of around zero (W/L = 1 and PayoffRatio == 1)
> >>>>>>> etc.
> >>>>>>> When plotted at the same time ... individual price series are
> >>>>>>> dispersed around the mean in a 'probability cone' ... in this
> >>>>>>> case it is a relatively tight cone because the method doesn't
> >>>>>>> introduce a lot of volatility to the series.
> >>>>>>>
> >>>>>>> /*P_RandomEquity*/
> >>>>>>>
> >>>>>>> //Use as a Scan to create PseudoRandom Equity curves
> >>>>>>> //Current symbol, All quotations in AA, select basetimeframe in
> >>>>>>> AA Settings
> >>>>>>> //It will also create the curves if used as an indicator (add
> >>>>>>> the appropriate flag to ATC)
> >>>>>>> // but this is NOT recommended as it will recalculate them on
> >>>>>>> every refresh.
> >>>>>>> //Indicator mode is good for viewing recalculated curves (click
> >>>>>>> in whitespace)
> >>>>>>> //CommentOut the Scan code before using the indicator code.
> >>>>>>> //Don't use a very large N or it will freeze up indicator
> >>>>>>> scrolling etc
> >>>>>>>
> >>>>>>> n = 100;//manually input desired number - used in Scan AND
> >>>>>>> Indicator mode
> >>>>>>>
> >>>>>>> ///
> >>>>>>> SCAN
> >>>>>>> ///////////////////////////////////////////////////////////////////
> >>>>>>>
> >>>>>>>
> >>>>>>> Buy=Sell=0;
> >>>>>>>
> >>>>>>> for( i = 1; i < n; i++ )
> >>>>>>>
> >>>>>>> {
> >>>>>>>
> >>>>>>> VarSet( "D"+i, 100 * exp( Cum(log(1 + (Random() - 0.5)/
> >>>>>>> 100)) ) );
> >>>>>>> AddToComposite(VarGet( "D"+i ),"~Random" + i,"X",1|2|128);
> >>>>>>> //Plot( VarGet( "D"+i ), "D"+i, 1,1 );
> >>>>>>> //PlotForeign("~Random" + i,"Random" + 1,1,1);
> >>>>>>> }
> >>>>>>>
> >>>>>>> /*
> >>>>>>> ////PLOT/////////////////////////////////////////////////////
> >>>>>>>
> >>>>>>> //use the same number setting as for the Scan
> >>>>>>>
> >>>>>>>
> >>>>>>> for( i = 1; i < n; i++ )
> >>>>>>>
> >>>>>>> {
> >>>>>>>
> >>>>>>> PlotForeign("~Random" + i,"Random" + i,1,1);
> >>>>>>>
> >>>>>>> }
> >>>>>>>
> >>>>>>>
> >>>>>>> ////
> >>>>>>> OPTIMIZE
> >>>>>>> ///////////////////////////////////////////////////////////
> >>>>>>>
> >>>>>>> //use the filter to run on Group253 OR add ~Random + i
> >>>>>>> PseudoTickers to a Watchlist and define by AA filter
> >>>>>>>
> >>>>>>>
> >>>>>>> //fast = Optimize( "MA Fast", 1, 1, 10, 1 );
> >>>>>>> //slow = Optimize("MA Slow", 4, 4, 20, 1 );
> >>>>>>>
> >>>>>>> //PositionSize = -100/P;
> >>>>>>> //Buy = Cross(MA(C,fast),MA(C,slow));
> >>>>>>> //Sell = Cross(MA(C,slow),MA(C,fast));
> >>>>>>>
> >>>>>>> //Short = Sell;
> >>>>>>> //Cover = Buy;
> >>>>>>>
> >>>>>>> I also stumbled on this, which seems to have some relevance:
> >>>>>>>
> >>>>>>> http://www.scribd.com/doc/6737301/Trading-eBookCan-Technical-Analysis-Still-Beat-Random-Systems
> >>>>>>>
> >>>>>>>
> >>>>>>> It contains a link to a site that has a free download of some
> >>>>>>> RNG produced datasets.
> >>>>>>>
> >>>>>>> There hasn't been much discussion on using synthetic data in the
> >>>>>>> forum ... Patrick recommended it for testing? OR
> >>>>>>> benchmarking? ... Fred is against using it ("If we knew enough
> >>>>>>> about the characteristics of the data, in the first place, to be
> >>>>>>> able to create synthetic data then we would know enough to
> >>>>>>> design trading systems to exploit the data's profile anyway", OR
> >>>>>>> something like that).
> >>>>>>>
> >>>>>>> I was titillated enough by my first excursion into benchmarking
> >>>>>>> with synthetic data to bring me back for some more.
> >>>>>>>
> >>>>>>
> >>>>>
> >>>>
> >>>
> >>>
> >>>
> >>>
> >>> ------------------------------------
> >>>
> >>> **** IMPORTANT PLEASE READ ****
> >>> This group is for the discussion between users only.
> >>> This is *NOT* technical support channel.
> >>>
> >>> TO GET TECHNICAL SUPPORT send an e-mail directly to
> >>> SUPPORT {at} amibroker.com
> >>>
> >>> TO SUBMIT SUGGESTIONS please use FEEDBACK CENTER at
> >>> http://www.amibroker.com/feedback/
> >>> (submissions sent via other channels won't be considered)
> >>>
> >>> For NEW RELEASE ANNOUNCEMENTS and other news always check DEVLOG:
> >>> http://www.amibroker.com/devlog/
> >>>
> >>> Yahoo! Groups Links
> >>>
> >>>
> >>>
> >>
> >
> >
> >
> >
> >
> > ------------------------------------
> >
> > **** IMPORTANT PLEASE READ ****
> > This group is for the discussion between users only.
> > This is *NOT* technical support channel.
> >
> > TO GET TECHNICAL SUPPORT send an e-mail directly to
> > SUPPORT {at} amibroker.com
> >
> > TO SUBMIT SUGGESTIONS please use FEEDBACK CENTER at
> > http://www.amibroker.com/feedback/
> > (submissions sent via other channels won't be considered)
> >
> > For NEW RELEASE ANNOUNCEMENTS and other news always check DEVLOG:
> > http://www.amibroker.com/devlog/
> >
> > Yahoo! Groups Links
> >
> >
> >
>
------------------------------------
**** IMPORTANT PLEASE READ ****
This group is for the discussion between users only.
This is *NOT* technical support channel.
TO GET TECHNICAL SUPPORT send an e-mail directly to
SUPPORT {at} amibroker.com
TO SUBMIT SUGGESTIONS please use FEEDBACK CENTER at
http://www.amibroker.com/feedback/
(submissions sent via other channels won't be considered)
For NEW RELEASE ANNOUNCEMENTS and other news always check DEVLOG:
http://www.amibroker.com/devlog/
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/amibroker/
<*> Your email settings:
Individual Email | Traditional
<*> To change settings online go to:
http://groups.yahoo.com/group/amibroker/join
(Yahoo! ID required)
<*> To change settings via email:
mailto:amibroker-digest@xxxxxxxxxxxxxxx
mailto:amibroker-fullfeatured@xxxxxxxxxxxxxxx
<*> To unsubscribe from this group, send an email to:
amibroker-unsubscribe@xxxxxxxxxxxxxxx
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
|