The Portfolio MYTH-----Do you need to change your thinking?

[systematic]

It's fun to see we're all different. I'm equal weight with position sizing (due to being agnostic about each selection relative to any other). I'm also just starting to consider increasing number of positions in the portfolio as I'm not sure I want to continue being as concentrated as I have been thus far. Doesn't change the philosophy of the plan, just a minor rule change.

 

[rb250660]

I run 5 portfolios each with 5 positions giving a total of 25 slots. Volatility can be quite high in each portfolio but collectively its been very smooth.

I have only been running this set up for 12 weeks now but things are going well. One of the models I have been trading for years, the other 4 are new.

 

[minwa]

collectively its been very smooth.

Long term sanity this is the way to go.

 

[galumay]

It's fun to see we're all different.

Indeed! Although I have some in common with you, I too dont have different position sizes - other than a couple of speccy plays that I did allocate much less capital to. I am less concentrated than I now like, I have learnt a lot over the last few years and knowing what I know now I would hold a bigger position in less companies, with more conviction.

I am probably quite different to many as liquidity and volatility are not things I worry about and I dont subscribe to the view that risk can be mathematically and quantifiably assessed.

My personal portfolio has 13 companies currently held and the SMSF has 23.

 

[Triathlete]

For everyone, please pay attention to the increasing ability of machine learning / artificial intelligence techniques. They have already surpassed humans in almost all fields. They are already being used by the largest trading houses. Systems based on subjective judgement of individuals, particularly individuals without extensive expertise, will have a hard time competing.

Are you able to give us some feedback as to what type of returns are possible over yearly periods based on your own results / experience / etc based on machine learning / artificial intelligence techniques.

Also the amount of trade signals that is usually given to trade over the same yearly period.

Thanks in advance....

 

[howardbandy]

Are you able to give us some feedback as to what type of returns are possible over yearly periods based on your own results / experience / etc based on machine learning / artificial intelligence techniques.

Also the amount of trade signals that is usually given to trade over the same yearly period.

Thanks in advance....

Greetings --

Machine learning based trading systems, such as those developed using python and scikit-learn, differ from traditional trading systems, such as those developed using AmiBroker or TradeStation, in several important ways.
1. Traditional systems almost always use a model based on decision trees. Machine learning allows use of any of several models, including decision trees, support vectors, various regressions, neural networks, boosting, ensembles, and others. Individual decision tree models are easily overfit, leading to poor out-of-sample performance. When several alternative model techniques are applied to a trading system, individual decision trees are seldom the best performers.
2. Traditional systems typically use impulse signals (buy and sell that each occur on a single bar and indicate the beginning and end of a trade), while machine learning systems always use state signals (which have a state -- beLong, beShort, or beFlat for the next bar -- for each bar). Traditional systems can be written so that they use state signals, a technique I recommend even if you do not plan to use machine learning. Holding period for a trade defined by buy and sell is the number of bars between the buy and sell. Holding period for a trade defined by states is the number of consecutive states that are the same. Using state signals requires computation of the next signal for each bar, and encourages daily mark-to-market accounting.
3. Traditional systems are based on "compute an indicator, then see what happened after," while machine learning systems are based on "identify a desirable trade, then see what happened earlier."

I recommend developing separate models for each issue traded. For a given issue, first develop a model that trades long/flat. If that works (passes validation), then try to develop a second model that trades that same issue short/flat.

Pay attention to risk. Risk analysis applies whether machine learning or traditional development.

Begin with your personal risk tolerance, then measure the risk-normalized profit potential of whatever system you are thinking about trading. Keep position sizing out of the trading system model -- it belongs in the trading management model.

I recommend using an objective function that first calculates maximum safe position size based on recent trades (call it safe-f), then estimates the compound annual rate of return (call it CAR25) of that system on that risk-normalized basis. When risk-normalized, the single best use of funds is the system with the highest CAR25. Using some of the available funds to trade any other system is sub-optimal use of those funds.

If you want to trade more than a single issue, develop several systems, each trading a single issue, and trade the one that is currently best. Rotate systems according to recent performance.

Be wary of creating portfolios of multiple issues, as these are extremely difficult to validate.

The number of signals depends on how the developer (the person) defines the system.

The sweet spot for a given system is always:
A. Trade accurately,
B. Trade frequently,
C. Hold a short period of time.
D. Avoid serious losses.

Accuracy should be at least 65%. Holding period one or two days. Fifty or so trades per year. What defines a serious loss depends on your risk tolerance, but the more losing trades and the worse the losing trades, the lower the safe position size and the lower the profit potential. Prefer to pass potentially winning trades in order to avoid losing trades.

There is a learning curve.

Best regards, Howard

 

[Value Hunter]

The link below is from a Q&A session with Charlie Munger.

http://www.vgipartners.com/wp-content/uploads/2017/03/February-2017.pdf

If you read pages 13 and 14 Charlie recommends that investors who know what they are doing do not need to own more then 3 stocks.

 

[Triathlete]

If the outcomes just aren't material to you, as previously mentioned, then who cares - the alternatives basically merge. But if they are then it takes a special idea of what constitutes best to suggest a 2 stock portfolio is a responsible way forward for the prior assets. I suspect that T/As account really doesn't matter that much for him. In which case, do whatever makes you happy...just don't take it as useful general advice for those who's situation requires more careful consideration.

Sometimes having a 1 or 2 stock portfolio is justified if you are using it to trade effectively. The idea is that the cash flow generated from concentrating on this type of portfolio can then be used to build a medium to longer term portfolio if required.....

Currently I trade 1 stock in the ASX 50 using CFDs and have been able to increase the starting capital base 100% in 4 months by trading in and out.

Trading portfolio currently stands at:
15 trades all short trades
11 closed
4 open
win/loss...100%.... previous years 80%-86%
R:R.... 3.6:1

My previous years were also successful trading this way to generate cash flow.

However I will make the point that I only use 10% of my assets to leverage for this type of trading I find I have more focus this way without having to worry about a larger trading portfolio.

 

[JuniorR]

Triathlete you trade once a week?
Do you trade a specific day & time in that week?

 

[tech/a]

Short term?

DAX Futures
Periodic when an opportunity arises.

Smaller caps
Periodic also
Most weeks it depends on filters
And on scans along with number of open trades.
Exposure

 

[systematic]

It’s such a case of, ‘it depends’ that I don’t think you can give anyone a concrete rule. It depends what a trader is doing, and indeed what their goals are.

A systematic or ‘quant’ style takes a basket approach to stocks, relying on averages. By its nature, Munger’s comment doesn’t apply. A quant has no business being a stock picker. I’ve made comments in other threads in the past such as, ‘I’m just not that good.’ A systematic investor is like someone who casts a net in a proven area and overall, it works out to a decent catch. A stock picker is like a fishermen relying on their own special skill, fishing where others aren’t and catching something amazing (can you tell from my poor analogy I’m not a fisherman?).
If it works out you’ll be king of the world. But I’d suggest it doesn’t go that way for many.

That’s one of the core philosophies of a systematic approach, in my opinion. It’s saying you’re not a special fisherman and you need a net.

Just some extra food for thought. And that’s from someone who has struggled with this aspect of systematic trading.

 

[Triathlete]

Triathlete you trade once a week?
Do you trade a specific day & time in that week?

I am a Discretionary short term trader using CFD's

No specific day or time.....

I use charts to tell me when is the best time to trade....

I look for specific patterns such as Head and Shoulders ,Elliot wave theory and Timing cycles.....once I can see this clearly it gives me a few things such as:

I know based on the expectation of the "said theory" how far a stock price is likely to move too.....if I know this I can take calculated trades until the stock reaches those prices or until the market tells me I am wrong.

I do not care which stock I trade as long as it is in the top ASX 50 as with using leverage I want to make sure I can get out in a hurry.....

I only require one stock with enough movement (volatility) in a week to make money using this strategy.

I only require to know that the price is going up or down and how far the move is likely because I am only using this "calculated high risk trading strategy for cash flow only".......This strategy is not for beginners...

The money that is made over the year ( hopefully if successful) so far so good over previous years..... will then be split by putting into a medium/longer term portfolio, real estate, bonds and some to increase the trading portfolio going into the next years trading which in turn increases the amount of trading capital using leverage......

 

[tech/a]

The filter switches the method I use on and off.

Allowing long new trades to be taken or allowing open trades to run without new trades being taken, or standing aside. It’s part of the systematic approach.

Fishing in a big sea. All thrown back some quicker than others some with my bait.
Some are missed,the ones I keep aren’t known until on my hook. My net is cast in only the volatile seas best suited to my fishing.

DAX is purely discretionary. It can be often only held for a few minutes when smashing long or short.Once or twice a week I get on one. It takes patience and in the first 2 hrs.

 

[Triathlete]

The filter switches the method I use on and off.

Allowing long new trades to be taken or allowing open trades to run without new trades being taken, or standing aside. It’s part of the systematic approach.

DAX is purely discretionary.

How many trades does the system allow at any time???

 

[tech/a]

It allows a % at risk—- 20
But that could see a few trades with 2 and three positions.
Rarely more than 5

 

[Triathlete]

It’s such a case of, ‘it depends’ that I don’t think you can give anyone a concrete rule. It depends what a trader is doing, and indeed what their goals are.

+1

 

[Triathlete]

"for small funds and individuals I think concentration is the answer"

+1

 

[Triathlete]

Deep state my essential issue is against the middle ground which is the worst of both worlds. If an investor lacks skill and wants to own index funds that is a prudent decision. If an investor has skill and wants to own 5 individual stocks that is all well and good also. What I disagree with (unless its a mechanical approach) is a fund manager or individual that owns a diversified portfolio with no more than say10% in any single holding and owns say 20-30 stocks. This person should either throw in the towel and buy index funds or trim their portfolio to a maximum of say 4-8 positions. Now I understand some fund managers have too much money too manage and due to size constraints must diversify, but for small funds and individuals I think concentration is the answer.

+1

 

[tech/a]

It allows a % at risk—- 20
But that could see a few trades with 2 and three positions.
Rarely more than 5

Just to clarify

Rarely more than 5 stocks held

Agree V/H + 1

 

[So_Cynical]

Some interesting discussion in this thread, i have recently come to realise that the size of my portfolio is an important factor in its success, i enter new stocks on a rough equal position size ($) basis and have done since day one, the goal was to build a dividend stream via a mix of dividend and growth stocks, buy low sell higher but retain some shares in each div stock and move on.

The Portfolio currently has almost 40 stocks with 4 stocks out performing thus lifting the portfolio return out of the ordinary, take the 4 best performers out and im not doing so well, i have to admit that the big 4 are somewhat random successes, i didn't know at the time of entry i just bought them because i had enough money for a new stock and saw some cheapness and potential.

I have noticed over the years that the top performers in my Portfolio keep changing and surprising, i dont know what the next big thing is so i just keep doing what i have been doing for the last 11 years because it works, one in ten will be a superstar out performer, money management is important.

I add 4 or 5 stocks per year to the portfolio and sell out of 1 or 2 and thats how i ended up with near 40, i should probably trim down to 30 or so, have started to re-balance over the last 2 years so the average position size has grown by maybe 15 or 20%, i have achieved a couple of size milestones this year, its been a hell of a journey.
~