Investing

Trading idea- Based on Graham (TSX)

  • Last Updated:
  • Nov 20th, 2017 6:10 pm
Tags:
None
[OP]
Deal Addict
User avatar
Dec 14, 2010
4304 posts
3260 upvotes
porticoman wrote:
Oct 10th, 2017 11:41 am
Rod on the OP I see three holdings that have aged 200 days, the other seven are between 18 to 90 days.

In your first report there is a term "turnover 86%", what does that mean?

On the three position that are over 200 days, is there a time when these would be sold?

What if any criteria are you using to sell those long hold positions or any position?

How would you describe your trading behaviour - could it be 'swing trading', or buy & exit on a certain increase, or 'it doesn't matter', it could be buy & hold long more than 365 days?
What @boyohboy said. The model might buy a stock and sell it on the following week. It could also buy a stock and sell it after a few years. The selling criteria is based on ranking deterioration or market timing to switch everything to cash (or bonds ETF). Statiscally, each stock is held for about 6 months. And the model usually buys a stock or 2 per month. That's average - some months have lots of buy and sells, while others go with no action for a while.

This model is a form of short term investing - it relies on fundamentals to find price disconnected from true value, and it usually takes time to catch up. It's a mechanical model, so sometimes a stock will be sold for little or no gain, but simulation (and the current short out-of-sample) has shown decent total return.


Rod
Everything about my Investing and automated Trading strategies to boost your income: https://boostyourincome.ca
Member
Jul 27, 2017
489 posts
138 upvotes
GTA
Thanks Rod, although I do not fully understand the 'sell point' on a stock that has a gain?

In post #1396 above you mention ...
rodbarc wrote:
Oct 10th, 2017 3:09 pm
The selling criteria is based on ranking deterioration or market timing to switch everything to cash (or bonds ETF).

This model is a form of short term investing - it relies on fundamentals to find price disconnected from true value, and it usually takes time to catch up.

It's a mechanical model, so sometimes a stock will be sold for little or no gain, but simulation (and the current short out-of-sample) has shown decent total return.
What or who decides when the sell point is for any stock that is either stagnant or has increased by a certain percentage?

In your list of 10, (exclude the ones that have dropped) you have stocks that have done really well in the short term or longer than 90 days (200+ days) that are still in position , then sell stocks that have only increased a few points?

Are you yourself deciding the sell point when to take the profit & exit the position, or is it based purely on technical trigger or a combination of both, When do you decide to hit the sell button?

In your model I believe that you have a sell rule <80, would there be a rule that applies to 'sell' when a stock has given it's best increase from it's bought price?...if that makes any sense?
[OP]
Deal Addict
User avatar
Dec 14, 2010
4304 posts
3260 upvotes
porticoman wrote:
Oct 10th, 2017 4:02 pm
Thanks Rod, although I do not fully understand the 'sell point' on a stock that has a gain?

In post #1396 above you mention ...



What or who decides when the sell point is for any stock that is either stagnant or has increased by a certain percentage?

In your list of 10, (exclude the ones that have dropped) you have stocks that have done really well in the short term or longer than 90 days (200+ days) that are still in position , then sell stocks that have only increased a few points?

Are you yourself deciding the sell point when to take the profit & exit the position, or is it based purely on technical trigger or a combination of both, When do you decide to hit the sell button?

In your model I believe that you have a sell rule <80, would there be a rule that applies to 'sell' when a stock has given it's best increase from it's bought price?...if that makes any sense?
The selling decision is solely done by the ranking. This model doesn't have sell rules based on gains or losses or specific trend / oscillators crossovers, like other models do. The fact that we might sell a stock for a small gain is a coincidence - the ranking determines how that stock is doing from a combination of growth, momentum, value and quality, and the top 20% is the most profitable one. So if a stock is not in that quintile anymore, it's sold to either be replaced with another stock or wait in cash until another buy opportunity arises. Since the whole process is based on a mechanical model, some stocks are sold for a small gain or even for a loss. However, overtime, this is a profitable strategy, since most stocks are sold for decent gains than losses. See the trading stats on the first post, which is also updated weekly.

After a list of stock is created to be purchased, following Graham's principles, another screening needs to take place to find the top 10 holdings, since the model holds up to 10 stocks. That's when the ranking comes in. It assigns scores to criteria based on growth, momentum, value and quality. You can see the current score distribution on the first post, under rankings, which is updated weekly. All stocks that made the list are purchased if they have a rank > 80, which means, they are on the top 20% of the quality-value-momentum-growth combination. Once they drop from that quintile, the stock is sold - regardless if the stock was purchased a month ago or a year ago, and regardless if the stock is profitable or not. This is because statistically, the stocks on the top 20% ranking are the most profitable - here are the stats (backtested from 1999 until today) of how ALL stocks in the TSX ranks according to this ranking - rebalance is done at every 4 weeks, which is the minimum hold time of this model:

Image


Rod
Everything about my Investing and automated Trading strategies to boost your income: https://boostyourincome.ca
Deal Addict
User avatar
Jun 19, 2009
4512 posts
729 upvotes
Scarborough
Is PLC = Park Lawn Corporation? Doesn't look like it fits the volume requirement
Deal Addict
User avatar
Jun 19, 2009
4512 posts
729 upvotes
Scarborough
Odd, it was showing 22k average volume on Google Finance.
Thanks for the reconfirmation
[OP]
Deal Addict
User avatar
Dec 14, 2010
4304 posts
3260 upvotes
Sorry, confusion from my side: The liquidity formula is daily average total, which is price * volume. So it calculates price * volume from the last 60 days, and the average from last 60 days has to be > 200,000. BTW, the minimum has been 200,000 (not 100,000) since we made this change on the live port. Not all small caps will meet Graham's rules and rank higher anyway. Given the volume for PLC, I don't think we'll have a liquidity issue to buy and sell.


Rod
Everything about my Investing and automated Trading strategies to boost your income: https://boostyourincome.ca
[OP]
Deal Addict
User avatar
Dec 14, 2010
4304 posts
3260 upvotes
tribe1689 wrote:
Oct 17th, 2017 4:54 am
Hey Rod,

not trying to be an ass or anything, but would the returns in your paid model not make you a multimillionaire already simply from being able to crush the market with what seems like such ease?

Is there a page that explains how the buying/selling process works? how do we get the indicator, how fast does it need to be done to be effectively following the model, etc...

Any chance you can give RFDers a discount? I seem to remember a few months ago it was going to be $15/month range, now its $30.
Thanks for the feedback. I'm responding on this thread as this is not relevant to dividend investing and it's more relevant to this thread, as it's related to algo trading (and some of your questions are applicable for this model too).

I'd be a multimilionaire if I started trading those models in 1999 or if I had put 6 or 7 digits when they went live in 2015. I started with a lot less, and most of models went live in 2015. However, the live performance, updated weekly, reflects how it's actually doing from a return perspective.

It seems easy to crush the market because the hard work has been done under the covers. It's transparent to the a user that a platform is scanning the market for rules that took a while to create and to rank them accordingly. Quant investing is a career and it takes time and effort to do it properly. When properly automated a lot of that work becomes transparent, as a user simply sees a buy and sell instruction. But keep in mind there will be periods of underperforming the market, by design, because the models always use the same rules while the market has different drivers for price change. There is no magic here, just an automation to always scan and buy or sell stocks according to a set of rules. These rules, of couse, will not always work all the time.

Every model has a rules section, but I will add another tab, common for all models, focusing on how to buy, sell, how does one get notified, etc. The models rebalance weekly or monthly, so they are not meant to day-trading, these are meant to give a decent return over a period of time. They are also liquid enough so that it shouldn't matter if you buy at open, middle of the day or close. The live model actually calculates the buying and closing price as an average of the day to reflect the closest price that you bought / sold at. Trading commissions ($4.95) and high slippage (fire safety) are included, which makes the live preformance actually looking worse than reality (which is a good thing).

Income based models are cheaper than growth based models, and both are cheaper than any quant algo trading subscription. Discounts are available when subscribing to more than one model. Lowering the price further wouldn't be fair for existing subscribers. The value of a subscription is the automation in scanning the market every week or every month according to those rules, and ranking it appropriately. The idea and high level rules are published, so one can build a DYI model for free. Or tag along on a model that has been tested and it's ready.

Thanks,


Rod
Everything about my Investing and automated Trading strategies to boost your income: https://boostyourincome.ca
Deal Addict
Dec 6, 2006
3791 posts
698 upvotes
Toronto
rodbarc wrote:
Oct 17th, 2017 10:42 am
...
Income based models are cheaper than growth based models, and both are cheaper than any quant algo trading subscription. Discounts are available when subscribing to more than one model. Lowering the price further wouldn't be fair for existing subscribers. The value of a subscription is the automation in scanning the market every week or every month according to those rules, and ranking it appropriately. The idea and high level rules are published, so one can build a DYI model for free. Or tag along on a model that has been tested and it's ready.
...
Considering price & cost. I guess if someone likes manyof the models and has the know-how to DIY them, it can be cheaper to replicate them on his own.
But for only perhaps 1-2 models and/or without knowing how to DIY replicate the model, it's just easier to subscribe. One will need a paid Portorlia123 account to do these full algo and backtrack test too I believe(?) so that isn't free either.
Jr. Member
Apr 8, 2017
118 posts
29 upvotes
rodbarc wrote:
Oct 17th, 2017 10:42 am
Thanks for the feedback. I'm responding on this thread as this is not relevant to dividend investing and it's more relevant to this thread, as it's related to algo trading (and some of your questions are applicable for this model too).

I'd be a multimilionaire if I started trading those models in 1999 or if I had put 6 or 7 digits when they went live in 2015. I started with a lot less, and most of models went live in 2015. However, the live performance, updated weekly, reflects how it's actually doing from a return perspective.

It seems easy to crush the market because the hard work has been done under the covers. It's transparent to the a user that a platform is scanning the market for rules that took a while to create and to rank them accordingly. Quant investing is a career and it takes time and effort to do it properly. When properly automated a lot of that work becomes transparent, as a user simply sees a buy and sell instruction. But keep in mind there will be periods of underperforming the market, by design, because the models always use the same rules while the market has different drivers for price change. There is no magic here, just an automation to always scan and buy or sell stocks according to a set of rules. These rules, of couse, will not always work all the time.

Every model has a rules section, but I will add another tab, common for all models, focusing on how to buy, sell, how does one get notified, etc. The models rebalance weekly or monthly, so they are not meant to day-trading, these are meant to give a decent return over a period of time. They are also liquid enough so that it shouldn't matter if you buy at open, middle of the day or close. The live model actually calculates the buying and closing price as an average of the day to reflect the closest price that you bought / sold at. Trading commissions ($4.95) and high slippage (fire safety) are included, which makes the live preformance actually looking worse than reality (which is a good thing).

Income based models are cheaper than growth based models, and both are cheaper than any quant algo trading subscription. Discounts are available when subscribing to more than one model. Lowering the price further wouldn't be fair for existing subscribers. The value of a subscription is the automation in scanning the market every week or every month according to those rules, and ranking it appropriately. The idea and high level rules are published, so one can build a DYI model for free. Or tag along on a model that has been tested and it's ready.

Thanks,


Rod
Rod

How much money do you recommend to start one of the models?

Also if you don't mind me asking, are you working in quant investing or did you get the experience from reading and programming?
If i want to do it for fun do uou recommend any prog language and free data from yahoo or google?

Thanks
[OP]
Deal Addict
User avatar
Dec 14, 2010
4304 posts
3260 upvotes
ShuttleBoy wrote:
Oct 17th, 2017 3:17 pm
Rod

How much money do you recommend to start one of the models?

Also if you don't mind me asking, are you working in quant investing or did you get the experience from reading and programming?
If i want to do it for fun do uou recommend any prog language and free data from yahoo or google?

Thanks
You need to factor the costs (trading commissions and subscription) to determine that. If you start with $7k (putting $1k on each position), on a growth model, then your cost is about 5%. But if you start with $10k, your cost now drop to 3%. These models are meant to give you a decent return overtime, so that would be the cost to have access to that. The more your money grows and the more funds you allocate to it, the lower your chat becomes. So it all comes down to your cost comfort level and awareness of the risk if we have a bad year.

I don't work with Quant Investing, if I did, then I probably would need my own firm to advertise models, since conflict of interest would prevent me to setup something like this. It started as a hobby and it grew as the only way for me to trade, since it's pretty hard to keep emotions apart and ensure consistency. It also gives me more confidence, as I use more than price and volume in a chart. These are the books I used and recommend. For programming language, learn Python, since most of algo trading platforms require that knowledge.


Rod
Everything about my Investing and automated Trading strategies to boost your income: https://boostyourincome.ca
Jr. Member
Apr 8, 2017
118 posts
29 upvotes
rodbarc wrote:
Oct 17th, 2017 5:15 pm
You need to factor the costs (trading commissions and subscription) to determine that. If you start with $7k (putting $1k on each position), on a growth model, then your cost is about 5%. But if you start with $10k, your cost now drop to 3%. These models are meant to give you a decent return overtime, so that would be the cost to have access to that. The more your money grows and the more funds you allocate to it, the lower your chat becomes. So it all comes down to your cost comfort level and awareness of the risk if we have a bad year.

I don't work with Quant Investing, if I did, then I probably would need my own firm to advertise models, since conflict of interest would prevent me to setup something like this. It started as a hobby and it grew as the only way for me to trade, since it's pretty hard to keep emotions apart and ensure consistency. It also gives me more confidence, as I use more than price and volume in a chart. These are the books I used and recommend. For programming language, learn Python, since most of algo trading platforms require that knowledge.


Rod
Great information...thanks!!
Sr. Member
Jul 30, 2008
776 posts
16 upvotes
Rod,

Few final questions.

1) Is the $28/month the only price we have to pay (other than commission costs, obviously) or do we have to open other paid account(s) like someone mentioned above?
2) Which one model would you recommend for total highest return. Would it be Value, Sentiment and Momentum (Canada)?
3) You mentioned you'll add info in the future on how we get notified on when/what to buy. In the mean-time, is it as simple as getting emails and then just making a sale/purchase? Or too complicated to give a quick answer?
4) It sounds like Value, Sentiment and Momentum (Canada) you're doing about 4 total trades per week or so, based on the stats? So not too labour-intensive, I guess.

I guess my savings are large enough that I can throw 10% in, and if the returns are what's stated, that 10% will quickly grow to be a lot higher. And quite frankly, I'm not sure what conventional ETFs, etc to invest in right now since everything seems to be peaking and there's not much that isn't at close to all-time highs.

Top