How to research and verify trading ideas

[TSGannGalt]

Quote from traderlux:

My thoughts exactly. Although somewhat strange in tone, I thought his posts were helpful, and I think he was one of the few that did some background homework and presented results. (Where is john galt? And bowo is still here?)

I wasn't banned completely. I got myself banned by posting a little joke in the reporting myself as a spammer. I was already planning on getting my main alias (this one) back so I told Joe that he should just keep my "whoisjohngalt" temp alias banned for good.

I like to joke around a bit here and there... so the previous alias very much deserved it.

Plus... I really have nothing more to add to this thread.

 

[talontrading]

Can I ask what people find so complicated about this? I guess I admit this thread got badly derailed by my "interaction" with BoWo and I was also pretty thrown off by people saying this little system was complicated.

Is it that it requires "outside" information? Meaning information not contained within the price bars?

Just trying to understand why so many people (and I know from the PM's that it seems to be A LOT of people) find this complicated.

And I'm discouraged that if this is complicated, it's really difficult to go any farther with this discussion because it only gets thicker from here.

Quote from bobneedham:

I find it a little complicated too, but it could really help out other traders. I just hope there will be new posts

 

[4XQs]

People who find it complicated want stuff like "Buy when MA10 crosses over MA20".

 

[talontrading]

more on the second strategy i wrote about here (earnings drift):

http://www.cxoadvisory.com/blog/external/blog11-24-09/

 

[talontrading]

<sigh>... ok then... for those folks here's another kind of idea:

Buy the close of this bar when:
1. it is above the 50th percentile of the past 200 days' closes.
2. it is below the 10th percentile of the last 15 days' closes.
3. It is the third consecutive close down.

Exit conditions:
1. Sell when today's close is higher than the 50th percentile of the past 5 day's closes.

Rules for shorts are the same, just flipped.
Do not enter a position if you already have a position.

Happy Thanksgiving! This is a purely technical idea that works. Discuss? Questions?

Quote from 4XQs:

People who find it complicated want stuff like "Buy when MA10 crosses over MA20".

 

[talontrading]

I received a PM question about percentiles. I didn't intend to make this difficult but it seems there may be a perception that I'm using obtuse concepts here deliberately. I think percentiles are much more simple than most of the indicators people use... what we're saying here is you take, for instance for #2 below, all of the last 15 trading days' closes, rank them high to low and then today's should fall in the bottom 10% of that ranking.

Sorry if this made this system seem intimidating. We do virtually everything here with nonparametrics, but that's a whole other story.

Quote from talontrading:

1. it is above the 50th percentile of the past 200 days' closes.
2. it is below the 10th percentile of the last 15 days' closes.
3. It is the third consecutive close down.

Exit conditions:
1. Sell when today's close is higher than the 50th percentile of the past 5 day's closes.

 

[cvds16]

I still don't get the concept, the way I was thaught using percentiles it had a totally different meaning. Could be semantics but who knows ...
Maybe a practical example would explain this better; take for example AAPL and give us the calculations, please ...

 

[talontrading]

The basic idea behind a percentile is to rank the observations and then take the Nth percent of the observations, interpolating if necessary.

This is a pretty simple concept.

Using Microsoft Excel and data available from Yahoo I calculate for AAPL that the 10th percentile of the past 15 days is 200.16 and the 50th percentile of the past 200 days is 142.72. The last close of 200.59 is above the 200 day 50th but is not below the 15 day 10th so there is no trade. Furthermore, there have not been 3 consecutive down closes in AAPL so there could not be a trade.

After I posted that system a few days ago, I was almost immediately PM'd by two posters who basically said "do you know what you're giving away here??" I am surprised (becoming less so) no response from the forum.

This idea is a simple, robust system that is a gem. We trade a very similar concept every day.

Quote from cvds16:

I still don't get the concept, the way I was thaught using percentiles it had a totally different meaning. Could be semantics but who knows ...
Maybe a practical example would explain this better; take for example AAPL and give us the calculations, please ...

 

[JScott]

Sticking with the original purpose of the thread, my contribution of “how to research” for this particular setup is rather simple. While this particular strategy lends itself well to analysis using Excel (if you have the raw data), I prefer a bit more visual approach. Because the criteria are basic, I was able to quickly code the criteria on the trading platform (TradeStation for me) and plot where the conditions are met.

Looking back over 2 years, this strategy doesn’t appear to have many trades (assuming my work is error-free). Looking at the DJIA cash, I see three sell signals for 2008 in total and five trades in 2009 (three sells; two buys).

I was able to quickly determine a few things from Talon’s method . . .
1) The stop loss designation is a fairly critical component as I noticed that most entries had minimal heat in the first few days, but the target doesn’t tend to be big either, so maybe a few trades would normally have been stopped out
2) Holding time ranged from 1 to 7 days – looking at the where the signals were generated, this strategy definitely has opportunity for trailing part of the position to catch a bigger move

Note: BTW, I took "third consecutive close down" to mean three "red candlesticks"as opposed to three lower closes (which could have an upday in the middle after a big gap down).

Anyway, I didn’t provide specific data for testing results I know, but easy enough to do so with a stop loss definition. And if Talon doesn’t provide his stop loss, next step would be to look at how the profit targets perform and determining an appropriate risk/reward outcome.

I prefer to “manually” backtest where possible to help internalize price action around the entry signals for a better understanding. Anyway, my first pass at looking at this suggests that it’s plenty good to keep looking at the potential – maybe even on smaller timeframes.

Thanks again, Talon.

Keep trading.

J. Scott

 

[talontrading]

You can't overestimate the importance of manual backtesting. Even in a quantitative intraday test that may test more than 100,000 occurrences of a pattern (no exaggeration) it is still a good idea to look at as many by hand as possible. This is a very good point, and this system is a good candidate for that kind of work.

We don't trade this with a stop loss, but we make other adjustments to cover some of the risk. I realize that seems very scary, but for instance we traded this through the financial crisis and made money. There is always the possibility of an outsized loss of course.

Good start. I should also point out that this is another idea I know to be good from my results trading it. Haven't really looked at it on lower timeframes, so that's an interesting area for exploration. There are reasons I think it may not work, but I could be wrong.

Maybe we'll find something useful for everyone here.

Quote from JScott:

Sticking with the original purpose of the thread, my contribution of “how to research” for this particular setup is rather simple. While this particular strategy lends itself well to analysis using Excel (if you have the raw data), I prefer a bit more visual approach. Because the criteria are basic, I was able to quickly code the criteria on the trading platform (TradeStation for me) and plot where the conditions are met.

Looking back over 2 years, this strategy doesn’t appear to have many trades (assuming my work is error-free). Looking at the DJIA cash, I see three sell signals for 2008 in total and five trades in 2009 (three sells; two buys).

I was able to quickly determine a few things from Talon’s method . . .
1) The stop loss designation is a fairly critical component as I noticed that most entries had minimal heat in the first few days, but the target doesn’t tend to be big either, so maybe a few trades would normally have been stopped out
2) Holding time ranged from 1 to 7 days – looking at the where the signals were generated, this strategy definitely has opportunity for trailing part of the position to catch a bigger move

Note: BTW, I took "third consecutive close down" to mean three "red candlesticks"as opposed to three lower closes (which could have an upday in the middle after a big gap down).

Anyway, I didn’t provide specific data for testing results I know, but easy enough to do so with a stop loss definition. And if Talon doesn’t provide his stop loss, next step would be to look at how the profit targets perform and determining an appropriate risk/reward outcome.

I prefer to “manually” backtest where possible to help internalize price action around the entry signals for a better understanding. Anyway, my first pass at looking at this suggests that it’s plenty good to keep looking at the potential – maybe even on smaller timeframes.

Thanks again, Talon.

Keep trading.

J. Scott