How to research and verify trading ideas

[tubfarter]

He, who stops the flow of rivers, gives liquidity to the rest of the market.

He, who stops the flow of rivers, does not need to research or verify anything. And he doesn't.

He, who stops the flow of rivers, simply drives the price against the majority.

It doesn't matter what you do, buy or sell, you are going to be wrong, unless you are positioned in the same direction of he, who stops the flow of rivers.

 

[Spydertrader]

Quote from talontrading:

FWIW, I disagree with a lot of these criteria, but it depends on what yo're trying to accomplish with the screen.... daytrading? swing trading? investing? longs only? shorts only?

The screen accomplishes several items.

In general, the screen provides a list of 'quality' companies to trade. In addition, the culling criteria provide a list of companies, that by definition, are going to 'move.' After all, without a change in Price, no money can be made. For example, the 'float' criteria, attempts to hit the 'sweet spot' with respect to available shares (to trade). Too small a float, and the company remains ripe for manipulation. Too high a float, and you see retarded Price movement in relation to volume. The same holds true for Institutional Ownership of Shares or Insider Owned Shares.

The screen works for any time horizon, and has a neutral bias. In other words, whether one trades intra-day, intermediate term or trades (invests) with a eye toward a more distant future, seeking out quality does have its advantages. Moreover, these same 'quality' stocks not only provide significant opportunity for entering long, but also, provide ample short signals as time moves on.

By running the screen every day, the trader creates a 'self-cleaning' list where those companies which fail to 'measure up' fall off - finding themselves replaced by far better candidates as time moves forward.

Although I trade futures exclusively now, I used these very same criteria when I traded equities.

Good Trading to you all.

- Spydertrader

 

[talontrading]

I can tell you from some fairly extensive quantitative fundamental work that is not a neutral bias screen.

I agree with your ideas of finding companies that will move, but I disagree with the idea of finding "quality" companies. For me, if the company is a piece of crap but everyone's buying it, I'll jump on that trend while it lasts.

I would screen something more like this:
$5 < Price < $300
Avg daily volume > 1M
Average true range > 1% of the underlying or $1, whichever is larger.

I would especially be interested in companies that show a high ratio of short term (say 2-5 day) volatility to long term (say 100 - 250 days).

I think that accomplishes what you're trying to do more directly without bringing in fundamental factors that have a directional bias (insider ownership, etc.)

Quote from Spydertrader:

The screen accomplishes several items.

In general, the screen provides a list of 'quality' companies to trade. In addition, the culling criteria provide a list of companies, that by definition, are going to 'move.' After all, without a change in Price, no money can be made. For example, the 'float' criteria, attempts to hit the 'sweet spot' with respect to available shares (to trade). Too small a float, and the company remains ripe for manipulation. Too high a float, and you see retarded Price movement in relation to volume. The same holds true for Institutional Ownership of Shares or Insider Owned Shares.

The screen works for any time horizon, and has a neutral bias. In other words, whether one trades intra-day, intermediate term or trades (invests) with a eye toward a more distant future, seeking out quality does have its advantages. Moreover, these same 'quality' stocks not only provide significant opportunity for entering long, but also, provide ample short signals as time moves on.

By running the screen every day, the trader creates a 'self-cleaning' list where those companies which fail to 'measure up' fall off - finding themselves replaced by far better candidates as time moves forward.

Although I trade futures exclusively now, I used these very same criteria when I traded equities.

Good Trading to you all.

- Spydertrader

 

[xburbx]

Ok that makes sense. I will have to research it myself a bit more as well. Assuming volatile stocks have been identified, how much of a factor does something like Beta become in a swing strategy in this kind of market that has been a raging bull? Sorry to jump topics.

Quote from talontrading:

I can tell you from some fairly extensive quantitative fundamental work that is not a neutral bias screen.

I agree with your ideas of finding companies that will move, but I disagree with the idea of finding "quality" companies. For me, if the company is a piece of crap but everyone's buying it, I'll jump on that trend while it lasts.

I would screen something more like this:
$5 < Price < $300
Avg daily volume > 1M
Average true range > 1% of the underlying or $1, whichever is larger.

I would especially be interested in companies that show a high ratio of short term (say 2-5 day) volatility to long term (say 100 - 250 days).

I think that accomplishes what you're trying to do more directly without bringing in fundamental factors that have a directional bias (insider ownership, etc.)

 

[Robert Yanks]

Quote from talontrading:

I can tell you from some fairly extensive quantitative fundamental work that is not a neutral bias screen.

I agree with your ideas of finding companies that will move, but I disagree with the idea of finding "quality" companies. For me, if the company is a piece of crap but everyone's buying it, I'll jump on that trend while it lasts.

I would screen something more like this:
$5 < Price < $300
Avg daily volume > 1M
Average true range > 1% of the underlying or $1, whichever is larger.

I would especially be interested in companies that show a high ratio of short term (say 2-5 day) volatility to long term (say 100 - 250 days).

I think that accomplishes what you're trying to do more directly without bringing in fundamental factors that have a directional bias (insider ownership, etc.)

Thank you for this post Talon. I do want to experiment a bit with volatility in my newbe backtesting.

I understand and can code the first two items. But …

“Average true range > 1% of the underlying or $1, whichever is larger.”

What number do I use for the ATR? I’m starting with 14 as that’s my default. Also, should the > than 1% be 10%?

“I would especially be interested in companies that show a high ratio of short term (say 2-5 day) volatility to long term (say 100 - 250 days).”

So if I used something like Mike805 suggested and go with (ATR5/ATR100) and then buy when it’s greater than 1.1 (or some figure above 1) is that the general idea here?

Thanks again.

 

[u21c3f6]

Quote from charts:

Isn't it ... (?):
Expectancy = (Probability of Win * Average Win) - (Probability of Loss * Average Loss)

Also ..., in my opinion, your trading shouldn't be "very closely related" to gambling.

Trading can be extremely closely related to gambling. I use the same methodology to "trade" live sportsbetting wagers as I do my options trading (I just use different data to make the "wagering" decisions).

IMO, one's trading expertise and success can be increased by learning how to "gamble" properly.

Joe.

 

[Mike805]

Quote from wutang:

I've begun looking at ways to filter out trades myself. Less trades and commisions and greater stability in returns sounds great. My concern with creating things like this volility spread indicator are how do I pick the proper time frames? I guess there's some fudge room on daily data but how do you come up with proper time frames on intraday data? Is this just a matter of trial and error, experience? Something like ATR(15)/ ATR(60) on a one minute chart seems feasible but I get nervious that I'm curve fitting when I try to get more specific than that on the period numbers. How do you know if the nature of volitility has changed and your filter is no longer accurate (before it's too late)?

A time filter has seemed beneficial to me on intraday trades. Only trading during certain hours of the day when volitility will be higher. How about only taking signals within a certain range of key prices (yesterday's close, overnight high/low, opening range high/low etc.)?

Even with a filter such as these, if you're using it with a mean reverting system how do you protect yourself from fading volatility that turns into a trend and fails to revert to the mean? Big spike in volatility, you get your signal, and then volatility dies but price is slowly drifting against you. Some sort of a time stop? Thanks again guys.

In reading this post and some of the others here I think its best to remind anyone following along that I don't intend to provide any hard (set in stone) answers to such questions as hard answers do not, in my opinion, exist.

Your job is to do the research yourself. You have to identify, based on the fundamental concept you are developing, what type of quantification is appropriate and where. The risk of curve fitting is always there, however, as I mentioned in another thread a while back, there is a difference between constructive and destructive optimization/curve-fitting, respectively. I got a lot of s--t from some posters regarding the above definitions, so bear with me.

It is my belief that one can utilize certain filters and optimize them constructively, i.e. finding best-fit variable settings. The opposite is also true; one can fool themselves quite easily if they're performing a type of data mining.

The difference lies in the aspect of the system you're trying to optimize. A case example is "pattern" finding. Algorithms that data-mine and search for patterns are IMO fundamentally flawed. As many who are experienced in learning algortihm development will tell you, a model is only as good as its initial rule set. It is very difficult to construct a non-curve fit model from a set of "found" (i.e. data mined) rules.... Optimization, when properly applied, can "adjust" certain rules to benefit the model. Filtering volatility is one of those adjustments IMO. That's not to say you should go along and "fit" ATR lengths without doing the research and analysis.

So that the rub of it, don't use optimization to find out which rules work per se, use optimization to "adjust" fundamentally sound rules.

How do you know if the rule you're working with is fundamentally sound? Well, talon's analysis of his volatility filter is a solid approach. Note that experience and good judgement help here as well.

Mike

 

[black diamond]

Quote from wutang:

Something like ATR(15)/ ATR(60) on a one minute chart seems feasible

If you are doing intraday vol calculations you need to be aware of the normal behavior of what you are trading and correct for it somehow. I think the ratio you posted would just go down at lunch time and up at the end of the day most of the time.

 

[Muskoka Joe]

Quote from black diamond:

If you are doing intraday vol calculations you need to be aware of the normal behavior of what you are trading and correct for it somehow. I think the ratio you posted would just go down at lunch time and up at the end of the day most of the time.

=============

Initially you should just try to verify your premise re volatility not assume its correct. So whatever you use it has to show and improvement over the basic model over a wide range of parameters. If the premise looks to have merit then one could improve the particulars. If it does not, try the opposite.

 

[SomeYoungGuy]

Talon,

I haven't been paying as much attention to this thread as I wanted over the last couple weeks, what with Christmas, New Year's and getting married. I worry you have gotten away from neophytes like me. Is is alright if we back up a bit?

From the posts of respected traders here, we have had a worthwhile trading strategy gifted to us. But how would we know that if we came up with it on our own? Meaning, when you wake up in the morning with an idea, what do you look for to help you decide to pursue or discard it?

Let me make a few of my own "strategies". I will keep them long-side only for simplicity.

A) go long when our stock goes up and the S&P goes down, and close out the position after two consecutive down days.
B) go long on our stock when yesterday's high temperature for Las Vegas is in the top half of it's decade in Farenheit (ie 65-69 deg or 85-89 deg), and close out when it is in the bottom half of it's decade (60-64 or 80-84)
C) go long on our stock when any professional New York sports team wins, close out the position after two days regardless.

Yes, I know these are silly and you know these are silly. But we don't get to just say "lame" and walk away. When we backtest, what do we look for? What makes the seed of an idea smell good enough to keep looking at?

Let's pretend that the above strategies backtest like this with a $100 bet:

A = 50% correct, winners win an average of $1, and losers lose an average of $1
B = 30% correct, winners win an average of $10, and losers lose an average of $5
C = 60% correct, winners win an average of $3, and losers lose an average of $7

What kinds of returns are we looking for? (Not exact numbers, but what do we want our returns to look like? Slightly profitable across the backtest universe? Hugely profitable at every stage? Personally, I would want strong steady gains, high probability of correct bets, and a low drawdown. Is that just my own bias showing through and pointing me in a non-optimal direction?)