Dump it Here

[Skate]

Let's talk about opportunities
Value investing is a timing-related issue. What looks good one week can be a "terrible buy" the next. Remembering, this statement works both ways.

Let's talk about timing
There is a great article to read (hyperlinked below) about the luck of timing when entering the markets. The article is about a novice investor who entered the Aussie market after the COVID plunge. Whether you are a trader or investor "timing" can be a matter of luck.

There is never a good time to start trading
Luck & timing plays a significant role in the performance of any portfolio

Sue Park, a novice investor entered the markets when others were fearful
Sue decided it was a great time to buy shares for the first time ever when panicked investors were hitting the "sell" button (March 16, 2020). The Australian markets had plunged 30 percent from its record high just three weeks earlier. I should also point out that the short-lived plunge lasted just 11 weeks. The rapid "vee-shaped" plunge & recovery even fooled the most senior traders.

Opportunities
The COVID-19 flash crash gave me the perfect opportunity to take advantage of what was being offered. Opportunities like these don't come around all that often

Story condensed
Ms. Park decided to buy shares in banks, the hard-hit travel sector, Telcos, and buy-now, pay-later (BNPL) companies. But she refused to buy oil and alcohol stocks as she considers them to be "very obviously evil". It's fair to say the fledgling investor has made a killing on the market.

There are plenty of opportunities
At any given time, there are plenty of opportunities in the share market as there are some excellent quality companies that are growing and out performing their competitors right now & the trick is finding them.

FOMO & TINA
The article starts out with the story of Sue Park & then goes on to explain the progress of the markets over the last 12 months. Along the way, the article articulates the meaning of two acronyms (FOMO - Fear Of Missing Out) & (TINA - There Is No Alternative) both drivers of the markets.

Please have a read
Record-breaking market frenzy will come to an 'abrupt halt', experts warn (msn.com)

Skate.

 

[Wilham]

On another line of thought.

Has anyone converted a system from daily to weekly and what was the impact on CAR/MDD? I they would go down and up respectively but keen to hear anyone's experiences with that.

 

[Skate]

On another line of thought.

Has anyone converted a system from daily to weekly and what was the impact on CAR/MDD? I they would go down and up respectively but keen to hear anyone's experiences with that.

Let's talk about a BBO Strategy
The BBO Strategy can be traded from "minutes to months" as the idea is sound across all time frames. When trading a similar idea - the strategy construction is completely redesigned with differing periodicity. Trading the same idea as a Daily or Weekly Strategy is not a simple matter of changing one or two parameters. The idea may be the same but the coding similarities normally end there.

Periodicity impacts CAR/MDD
Trading is a constant trade-off & there are advantages of trading different time frames but I prefer & gravitate towards weekly systems, but that's just me. I'm more concerned with making the most amount of money from the least amount of effort.

Your question about trading weekly versus daily allows me the opportunity to express my views
Putting the question about the impact that periodicity has on CAR/MDD I would rather make a boxing comparison. Trading weekly is like boxing in slow motion, meaning you have time to "respond" when trading is not going your way. Trading on a daily time frame you are always "reacting" to price movements.

Responding gives you time to think
I make it one of my trading rules never react always respond. (it's one of my life rules as well)

Summary (Weekly versus a Daily)
There are merits to trading both time frames but for "me" I prefer trading a weekly system as it reduces the workload & stress. There is always a robust discussion when Daily versus Weekly trading is raised but I'm a believer "if it works for you, it's right for you". By posting consistent messaging - reinforces my point of view.

Skate.

 

[qldfrog]

Hey man
Are you using position score as part of your code if you are or planning to then that isn't the best method.
i used a different method which was posted by @LoneWolf which works really well. Have it as part of your buy method. Just be very careful, make sure you take it out of your production code. made the mistake many times and had a few WTF is going on moments.

step = Optimize( "step", 1, 1, 1000, 1 );
Buy = cond1 AND Random() >= 0.20

I found the best way to differentiate between strategies is to plot CAR v MDD as below. CAR is X and MDD is Y. Green strategy 1 blue strategy 2.View attachment 121441

Is there a way to highlight , bold and underline that post.i remember reading about this way of testing systems consistency from @Lone Wolf months and months ago,and could not find it back and here it pops up again.Credit to the wolf and also @othmana86 .
This is a gem of a code snippet ??

 

[Lone Wolf]

Is there a way to highlight , bold and underline that post.i remember reading about this way of testing systems consistency from @Lone Wolf months and months ago,and could not find it back and here it pops up again.Credit to the wolf and also @othmana86 .
This is a gem of a code snippet ??

On this topic, it's worth repeating the advice @MovingAverage threw in:

That is my prefered way of doing monte in AB. Simply export the optimization results into Excel and you can do some great visual/charting analysis of your systems performance. This is much better than doing a single run. Although I'd suggest you push the 1000 value to much higher to get a better perspective. Only thing I'd add is that you're better to use mtRandom() over the regular Random() function. Be very careful about selecting the random comparison value (you use 0.2), choose the wrong value and the number of trades you take will have an adverse impact on your system.

 

[Skate]

Investing versus trading
I normally refrain from posting in other threads but @peter2 raised the idea of a Large Cap Portfolio, a portfolio that I have already constructed. Peter's suggestion was for a 5-position portfolio & it's exactly the size I had set for my portfolio. Why 5 positions? It allowed a heavy concentration of funds in a few positions - making it a worthwhile exercise.

Worthy of a repost in the "Dump it here" thread

peter2 said:
The large caps (LC) move the index and it's going to be very important to buy these at the earliest opportunity. I will start the large cap portfolio (5 positions) soon. It'll be a low activity portfolio and I'll post actions and progress in this thread. The main aim for the large cap portfolio is to earn more than a term deposit (TD) + tax. This is not a very high standard but I don't want to leave the money at the bank earning next to nothing when I have the skill to earn a little more.

@peter2 from my perspective I'll detail how I handle my Large Cap "Investment Strategy"

Spare funds hanging around
There is a distinction between trading & investing. I wish to make a comment on how I handle my "Large Cap Portfolio". This portfolio currently holds 5 positions & is an investment strategy seeking (a) capital gains & (b) dividends.

Investing is different from trading
Unfortunately, investing in "mispriced" positions rely on doing a few things differently (from trading) as far as I'm concerned. The main difference between the two is (a) the value invested in each position increases significantly & (b) the position relies on the "time in the markets" - not the "timing of the entry" to develop its long-term potential.

Horses for courses
With investing, timing the entry isn't critical but "time in the markets is". In this game, we have one objective "which is" to skim the price differential between the entry & exit price. Coding an idea allows you to target opportunities that you believe you can profit from.

Mean Reversion Strategy
Years ago, I traded a mean reversion strategy, the idea was sound but not that profitable. This concept allowed me to expand on this idea to develop a strategy specifically targeting positions that historically become quickly oversold than exhibiting strong signs of a recovery.

The Bee Strategy sprung to life
This Bee Strategy (the idea) was dormant for quite some time & I was resolved to the fact that this strategy wouldn't give many signals - if any. The last quarter of 2018 was the first sign the Bee Strategy sprung into "life" giving a few signals but the signals were short-lived.

The COVID flash crash
Near the end of March 2020 (the 19th March to be precise) the "Bee Strategy" started to produce signals. At that time, I was more concerned about actioning the GTFO filter, exiting over a hundred positions to reduce the hemorrhaging.

Investment signals
At that time, the Bee Strategy signals were ANZ, BHP, MQG & BHP (all positions were taken). Recently the "Bee Strategy" flashed another buy signal being MFG which was taken immediately, as the first four signals turn out to be very rewarding.

Four weeks later another MFG signal
I've reported that I've taken an extra position in MFG on that second signal. These 5 positions now make up an investment portfolio - a buy & hold strategy to be precise. My thinking & actions of a Large Cap Portfolio is similar but different, giving an alternative method.

Skate.

 

[Skate]

Be very wary of some of these sites as apart from professionals like Radge they will send you broke.

Look up @Skate, @peter2, @tech/a in no particular order, actually just about anyone on here. You will learn a lot more but it may not be as pretty or appealing and you will have to do the hard yards but it will save losing money and confidence.

Where does the motivation & inspiration come from?
The "Dump it here" thread has evolved over time but the theme still remains the same - helping others understand the importance of self-education with all things trading.

Investing
I normally don't post about investing as it's the poor cousin to trading. Simply, trading requires more skill with an upscale in risk. The other issue is "developing a profitable trading system" that takes a lot of work & time (requiring upskilling).

Technical Analysis
Technical Analysis to me is a vague terminology - a method of predicting the market's future movements.

Predicting is an "inappropriate terminology"
I'll use a softer word (which basically means the same) - "probability".

Probability is a "get out of jail" terminology
Using the word "probability" implies ambiguity because with all probabilities - it will either happen or not. The word is perfect to cover your ar$e when things go south. Technical analysis at times gets a "bad rap" because we are working with an "in-exact science". The markets are a combination of emotions. Technical analysis suffers at the hands of some posters because it is a "discipline" normally practiced without discipline.

Skate.

 

[Skate]

Entering the market
With any endeavour there are multiple techniques to achieve the same outcome.

Fundamental analysis is one technique & technical analysis is another
Combining technical analysis with systematic (mechanical) trading gives mathematical proof of the "efficacy of applied rules". I appreciate that "efficacy" is being overused at the moment with the "Covid vaccine" but the "efficacy" of testing results gives the confidence to start trading it. I used at times a combination of methods & ideas gleaned from other respected posters. I've shared the general concepts & at times they have been useful to others. Sometimes it exciting to watch an idea go from paper trading & ultimately go live.

There is a multitude of traders who prosper from trading systematically
Mechanical trading (systemically) is all about a process by which we look for trading signals & on that evidence of "profitability" the statistical significance allows us the possibility to trade those technical signals with confidence. At times simple technical & trading rules have often led to profits - extensive evaluation of the performance is the key ingredient.

The Action Strategy
This Action Strategy is currently parked but @ducati916 idea for the strategy has been proven beyond doubt to be sound.

The Sphere Strategy
This "Sphere Strategy" came from an idea @peter2 posted back in 2016. Peter traded the idea (a concept developed by others) with success. Twisting these original ideas means I don't have to re-invent the wheel. I'm just approaching the idea from a different (unique) angle. I have already outlined the buy condition of the "Sphere Strategy" a few posts back saving me from re-posting the concept again. I'm excited to experience if using a short four-week lookback period of a stable position (IMO) can be successfully traded when momentum has been confirmed.

What am I hoping to achieve?
Well, other than keeping the "Dump it here" thread active & relevant - I wanted to paper trade (experiment) with the crazy idea (in real-time) confirming that entries (into a trend) are a dime-a-dozen & timing the exit becomes an important factor of profitability of any strategy.

Skate.

 

[ducati916]

If I may
instead of trying a period 1992 to present, work on a recent year;
In 1992, was there many qant around?, fast trading? were Central bank covering your asses?
what I mean is you try to compare the performance of a ferrari in 2020 against the speed of a roman charriot in a mud track in 10BC..
totally irrelevant IMHO;
I would dare saying testing anything before 2006 and expecting a relevance in 2021 is a folly;
Anyway:
then try to run and compare with 2011 as per Monsieur Skate; the 2019 crash?
how does your system behave and handle the crash then the recovery ?
what about the yoyo period we have had since nov 2020 to now basically?
We are REALLY lucky to have many types of markets within the last 2 y..use these to your advantage
PS I noticed you have a 42% exposure during some of the longest bull market in history,
if not during that given period, when would your system get in ? A bad sign IMHO
also: std error: 1986636?????
what does the MC show?

This post seemingly was overlooked. It really should not have been. It is central to the issue of backtesting mechanical based systems.

The basic premise that seems to be predominating is; the more information I have as a trader and use in my backtest, the more (statistically) certain I can be of my proposed system.

The (statistical) confidence level increases (however) in a non-linear proportion to the number of observations (n) and which results in an improved confidence level, as the square root of (n).

The issue that Monsieur Frog has identified is (and why this is the case) that the 'market's' distributions are non-symmetric, particularly in the ones that we are primarily interested in: market dislocations (lower) or (black swan) events. Looking at 1987, 2019, 2020 will provide you with far more relevant information than 1992 - 2021 inclusive.

jog on
duc

 

[Cam019]

This post seemingly was overlooked. It really should not have been. It is central to the issue of backtesting mechanical based systems.

The basic premise that seems to be predominating is; the more information I have as a trader and use in my backtest, the more (statistically) certain I can be of my proposed system.

The (statistical) confidence level increases (however) in a non-linear proportion to the number of observations (n) and which results in an improved confidence level, as the square root of (n).

The issue that Monsieur Frog has identified is (and why this is the case) that the 'market's' distributions are non-symmetric, particularly in the ones that we are primarily interested in: market dislocations (lower) or (black swan) events. Looking at 1987, 2019, 2020 will provide you with far more relevant information than 1992 - 2021 inclusive.

I fundamentally disagree. The more data you have to run a back test, the more statistically significant the results will be.

The less parameters a system has and the longer the historical data set, the bigger the trade sample size will be and therefore the more confidence you can have that your system will perform within the backtested results, over time.

I never understood the idea that people would try and backtest a system over a small historical data set of say, 1 year, to try and either reduce drawdown or get a positive return over a black swan or economic crisis event. If your system trades all the markets the same and the longs and shorts the same, the bigger the trade sample size, the more statistically significant the results and the more certainty you have regarding the range of expected outcomes the system could face over time.

 

[MovingAverage]

This post seemingly was overlooked. It really should not have been. It is central to the issue of backtesting mechanical based systems.

The basic premise that seems to be predominating is; the more information I have as a trader and use in my backtest, the more (statistically) certain I can be of my proposed system.

The (statistical) confidence level increases (however) in a non-linear proportion to the number of observations (n) and which results in an improved confidence level, as the square root of (n).

The issue that Monsieur Frog has identified is (and why this is the case) that the 'market's' distributions are non-symmetric, particularly in the ones that we are primarily interested in: market dislocations (lower) or (black swan) events. Looking at 1987, 2019, 2020 will provide you with far more relevant information than 1992 - 2021 inclusive.

jog on
duc

I mean no disrespect when I say this, but I don't agree with the generalized proposition that old data (pre 1992 as suggested by frog) or that '87, '19 and '20 will be more relevant that 1992 - 2021. What is important first and foremost is that you have a statistically relevant number of trades in your backtest--in other words your sample size in statistics parlance. To suggest that you need to test over a certain time period without regard to the system characteristics (e.g., hold time, trade frequency etc) is naïve at best. For example, for a short hold time system such as a swing system you may get a statistically relevant number of trades within a six month period, but in contrast a long hold system may require a backtest period of 5 years to get a statistically relevant number of trades. What specific period you backtest over is irrelevant to a degree--what is critical is that post backtesting (optimization) the systems performance on out of sample data is carefully reviewed. So by all means backtest your system on pre '92 data but make sure you forward test (not optimizing) on data post '92 to see if it consistently performs to your requirements. This should of course include forward testing on out of sample data that represents a range of market conditions, pullback, sideways and heading north. Also, (and I'm probably misunderstanding your point) but the z-score / t-score is the common measure of statistical confidence and there is nothing non-linear about them.

 

[qldfrog]

I mean no disrespect when I say this, but I don't agree with the generalized proposition that old data (pre 1992 as suggested by frog) or that '87, '19 and '20 will be more relevant that 1992 - 2021. What is important first and foremost is that you have a statistically relevant number of trades in your backtest--in other words your sample size in statistics parlance. To suggest that you need to test over a certain time period without regard to the system characteristics (e.g., hold time, trade frequency etc) is naïve at best. For example, for a short hold time system such as a swing system you may get a statistically relevant number of trades within a six month period, but in contrast a long hold system may require a backtest period of 5 years to get a statistically relevant number of trades. What specific period you backtest over is irrelevant to a degree--what is critical is that post backtesting (optimization) the systems performance on out of sample data is carefully reviewed. So by all means backtest your system on pre '92 data but make sure you forward test (not optimizing) on data post '92 to see if it consistently performs to your requirements. This should of course include forward testing on out of sample data that represents a range of market conditions, pullback, sideways and heading north. Also, (and I'm probably misunderstanding your point) but the z-score / t-score is the common measure of statistical confidence and there is nothing non-linear about them.

With my english as a foreign language, i might have been misunderstood, but i preach the opposite: in clear words:
Testing on old data, and by old i mean anything pre GFC is irrelevant.
I disagree with your idea of the more data the better: this would be true in a static environment, but the market is not static, forever changing and i am talking here about the mechanics, the mental response of traders and qants.
Would you be happy to analyse the average speed of a butterfly and take into account its chrysalis and caterpillar stage to have more data?
I want to follow mr Skate animal analogies?
If you use a monthly system, you have no better choice i agree than spanning years and years..good luck for it to be relevant.
Hope it clarifies my point of view..which is just that, and never pretented to be a truth
The only constant in the market is the change

 

[Skate]

Opinions & differing points of view
The recent exchanges (between differing points of view) in this thread demonstrate why the "Dump it here" thread is different to most other threads as it allows every member the right to express their views on a subject or topic respectfully.

Opinions are welcomed in the 'Dump it here' thread
It's a perfect segue to remind others that we all enjoy reading differing points of view because that's how we learn. "Refrain" is sometimes advisable because we're all "wordsmiths to a point" & challenging poster serves no purpose, it's much like masturbating in public, it may feel good to you, but it looks disgusting to everyone else.

Express your views
Whether your view is right or wrong isn't important, what's more important, this thread gives you the ability to express your views without being ridiculed or challenged. Every member enjoys a different level of experience & expertise. Posting an alternative view is the "heart & soul" of this thread as displayed in the last series of posts. I'm just saying, without self-moderation, the tone can quickly escalate.

Skate.

 

[ducati916]

1. I fundamentally disagree. The more data you have to run a back test, the more statistically significant the results will be.

2. The less parameters a system has and the longer the historical data set, the bigger the trade sample size will be and therefore the more confidence you can have that your system will perform within the backtested results, over time.

3. I never understood the idea that people would try and backtest a system over a small historical data set of say, 1 year, to try and either reduce drawdown or get a positive return over a black swan or economic crisis event.

4. If your system trades all the markets the same and the longs and shorts the same, the bigger the trade sample size, the more statistically significant the results and the more certainty you have regarding the range of expected outcomes the system could face over time.

1. This statement is true if we are talking about the height/weight/lifespans of humans or similar. It is categorically false when you are talking about financial markets.

2. Your 'statistical confidence' will undoubtably be higher. Unfortunately, it means nothing in financial markets.

3. Because that is where, depending of course how you trade (leverage etc) is where you blow-up.

4. Simply incorrect.

jog on
duc

 

[ducati916]

1. I mean no disrespect when I say this, but I don't agree with the generalized proposition that old data (pre 1992 as suggested by frog) or that '87, '19 and '20 will be more relevant that 1992 - 2021.

2. What is important first and foremost is that you have a statistically relevant number of trades in your backtest--in other words your sample size in statistics parlance.

3. To suggest that you need to test over a certain time period without regard to the system characteristics (e.g., hold time, trade frequency etc) is naïve at best.

4. For example, for a short hold time system such as a swing system you may get a statistically relevant number of trades within a six month period, but in contrast a long hold system may require a backtest period of 5 years to get a statistically relevant number of trades.

5. What specific period you backtest over is irrelevant to a degree--what is critical is that post backtesting (optimization) the systems performance on out of sample data is carefully reviewed. So by all means backtest your system on pre '92 data but make sure you forward test (not optimizing) on data post '92 to see if it consistently performs to your requirements.

6. This should of course include forward testing on out of sample data that represents a range of market conditions, pullback, sideways and heading north. Also, (and I'm probably misunderstanding your point) but the z-score / t-score is the common measure of statistical confidence and there is nothing non-linear about them.

1. And no offence or otherwise taken.

2. I disagree. What is important is that your system can hold up to the infrequent dislocations that blow you up. Now some of that risk will be in how you actually trade (Futures, Options, Stocks, CFDs, etc.) and the leverage you have in the system. That is not a function of the number of observations in total that you have.

3. Not a certain time period: specific market conditions.

4. Irrelevant.

5. There was a (brief) discussion on Monte Carlo analysis. Unfortunately it didn't really go anywhere. Essentially, MC is an attempt to look at everything that could happen, but didn't. The answer lies in a much more robust understanding and analysis of MC.

6. See above.

jog on
duc

 

[MovingAverage]

With my english as a foreign language, i might have been misunderstood, but i preach the opposite: in clear words:
Testing on old data, and by old i mean anything pre GFC is irrelevant.
I disagree with your idea of the more data the better: this would be true in a static environment, but the market is not static, forever changing and i am talking here about the mechanics, the mental response of traders and qants.
Would you be happy to analyse the average speed of a butterfly and take into account its chrysalis and caterpillar stage to have more data?
I want to follow mr Skate animal analogies?
If you use a monthly system, you have no better choice i agree than spanning years and years..good luck for it to be relevant.
Hope it clarifies my point of view..which is just that, and never pretented to be a truth
The only constant in the market is the change

I'm not suggesting more data is better--I'm suggesting that a statistically relevant amount of data be used. If you used more data than is statistically relevant then no adverse influence will result other than allowing you to have a higher level of confidence in your results--there is no down side to using more data than is statistically relevant. However, use a sample size that is not statistically relevant then you must would apply a low level of confidence in your results. Your reference to monthly systems is very apt here--I said on this forum before that I abandoned my monthly system because I just simply could not have any faith in the backtesting results--could not get a statistically relevant result even over 20 years so very low confidence. The key is understanding the statistical confidence you have in your sample size and then applying the appropriate level of error to your results. I appreciate the real world analogies, but system trading is nothing more than applied statistics. Stock market price data is nothing more than discrete time series data and statistical analysis is a well established form analysis for such data. Anyway, each to their own and whatever works for everyone is good--just sharing my thoughts.

 

[ducati916]

Opinions & differing points of view
The recent exchanges (between differing points of view) in this thread demonstrate why the "Dump it here" thread is different to most other threads as it allows every member the right to express their views on a subject or topic respectfully.

Opinions are welcomed in the 'Dump it here' thread
It's a perfect segue to remind others that we all enjoy reading differing points of view because that's how we learn. "Refrain" is sometimes advisable because we're all "wordsmiths to a point" & challenging poster serves no purpose, it's much like masturbating in public, it may feel good to you, but it looks disgusting to everyone else.

Express your views
Whether your view is right or wrong isn't important, what's more important, this thread gives you the ability to express your views without being ridiculed or challenged. Every member enjoys a different level of experience & expertise. Posting an alternative view is the "heart & soul" of this thread as displayed in the last series of posts. I'm just saying, without self-moderation, the tone can quickly escalate.

Skate.

And this (particular) topic seems to have stirred some very different and opposing viewpoints, which is a good thing.

jog on
duc

 

[MovingAverage]

2. I disagree. What is important is that your system can hold up to the infrequent dislocations that blow you up. Now some of that risk will be in how you actually trade (Futures, Options, Stocks, CFDs, etc.) and the leverage you have in the system. That is not a function of the number of observations in total that you h

Hang on a minute--I was referring about a statistically relevant sample size. What you are talking about is something very different, unrelated and not relevant to my point. In fact your response is confirming one of my very points in which I said it is important to do out of sample testing across a range of market conditions--which as you point out MUST include infrequent dislocations that blow you up.

 

[ducati916]

Hang on a minute--I was referring about a statistically relevant sample size. What you are talking about is something very different, unrelated and not relevant to my point. In fact your response is confirming one of my very points in which I said it is important to do out of sample testing across a range of market conditions--which as you point out MUST include infrequent dislocations that blow you up.

Your general point is that more data is better than less data. I am saying that that is simply not the case. We agree that specific samples of data are required.

jog on
duc

 

[MovingAverage]

Your general point is that more data is better than less data. I am saying that that is simply not the case. We agree that specific samples of data are required.

jog on
duc

No my general point is not that more data is relevant--like I said, a statistically relevant amount of data is what is need. They are two very different things and should not be confused. It may seem like one and the same, but trust me in the world of statistics they are chalk and cheese.