Do Trendlines work?

[John Merchant]

I refuse NOT to entertain!

OK, so you won't entertain. Perhaps then we could have a dialogue on how YOU set up an intraday BT of trend line formation. Then we could understand why YOURS worked, and MINE didn't. I just love talking to people about backtesting. Little matters like time of day, definition of a top or bottom, experienced slippage, stops, profit targets, lalala.

Mind, I'm not asserting that trend lines have no value, only that they are worthless by themselves.

Of course, if you look at all the trades I post, they are rollouts from low volume midday consolidations MINUTES before the rollout occurs. Try backtesting that!

 

[MAESTRO]

Quote from John Merchant:

I would be properly humbled if you posted an execution report. I do, although mine are not as magnificent you claim yours to be. I stand by my contention that intraday trend line formation does not backtest with positive expectation.

I have two years, 100 securities back test results. Yes, you are correct. Trend lines in its bare form do not produce positive results. IMO trend lines could be used in conjunction with some other techniques (RSI, bollinger etc). However, having said that I do not have any results to back that up. Its just a hypothesis.

 

[John Merchant]

Abogdan. Thank you. It is believable coming from you, whereas I am patently an idiot. I wasted quite a lot of time tweaking the simplest possible models of trend line formation. Hence my thinly disguised bitterness. I do SO want money for nothing! What I DO believe is that trendlines have value when they are corroborated by buying/selling pressure, certain relationships of cash to futures price, certain price manipulations, and certain total volume formations. Absolutely none of which is backtestable because it verges on clairvoyance. A la Jack.

 

[MAESTRO]

Quote from John Merchant:

Abogdan. Thank you. It is believable coming from you, whereas I am patently an idiot. I wasted quite a lot of time tweaking the simplest possible models of trend line formation. Hence my thinly disguised bitterness. I do SO want money for nothing! What I DO believe is that trendlines have value when they are corroborated by buying/selling pressure, certain relationships of cash to futures price, certain price manipulations, and certain total volume formations. Absolutely none of which is backtestable because it verges on clairvoyance. A la Jack.

Difficulties of the trend lines back testing is obviously related to their "fuzzy" definition. I haven't met two traders yet that would have exactly the same trend line definition or the same algorithm to produce them on the chart. It often happens that 10 traders would draw 10 different sets of trend lines on the same chart. I did this experiment in our office. That is why I at least suggested to draw trend lines using pivot points that could be defined more precisely.
Cheers.

 

[John Merchant]

El Maestro. Thanks. Makosgu wisely declines to engage. So I will continue without him (or her). The problem is that as you define a trend line system for backtesting you develop an unwieldy number of variables. (That is the joy oif Jack: he doesn't worry about it.) To wit:

the trading instrument

the time of day you start looking for a trend

the time you stop looking for a trend

the earliest time of day that you will believe a found trend to be valid

the latest time day that you will believe a found trend to be valid

the definition of the supposed tops which form a trend (this can be more than one rule depending on how much verification or delay you want)

ditto for bottoms

the stop for longs

the stop for shorts

the profit target for longs

the profit target for shorts.

This is at least eleven rules, which WAY violates my criteria for valid backtesting.

 

[makosgu]

I like quizzes?!!! Are some of these trick Q's?

Quote from John Merchant:


...
1. the trading instrument

irrelevant. Well, somewhat irrelevant (illiquidity is only a minor nuissance)

2. the time of day you start looking for a trend

you can always start with now. Orientation is then immediately corrected continually.

3. the time you stop looking for a trend

Look??? Trends are recognizable immediately and with respect to your entrance and fractal.

4. the earliest time of day that you will believe a found trend to be valid

Trends are valid immediately

5. the latest time day that you will believe a found trend to be valid

Redundant Q, Redundant Answer

6.the definition of the supposed tops which form a trend (this can be more than one rule depending on how much verification or delay you want)

irrelevant (top???)

7.ditto for bottoms

ditto on relevance

8.the stop for longs

reverse into short

9.the stop for shorts

reverse into long

10.the profit target for longs

most irrelevant, reverse at market?

11.the profit target for shorts.

same answer as 10...

Of course, this all may seam bold, but it is taken from the context of exactly how I trade.

Are you not entertained???

 

[John Merchant]

Makosgu. Thank you for your response. I suppose, in retrospect, that they WERE trick questions. Evidently my approach to composing trading algorithms is FAR too pedantic. The code engine I use is SO literal.

 

[Lefty62151]

Oh My God:
Still beating this to death. So, how many times can you beat a horse to death. Apparently quite a few.

Look, trendlines are simply "lines in the sand". Certainly there is an element of "art" in that you have to decide where to place them. Currently I don't use them intraday, but if I did, I would TRY to start at a swing high or low, tying together consecutive highs or lows to project a price "barrier" into the future. Certainly I agree with Dr. Abogdan that no two people seem to draw them in the same way. I am just proposing an example..

The limitation of backtesting is also the skill level of the programmer. I have not tested trendlines. Don't need to. I am looking to evaluate what happens when price "tests" the line. Again an example. If price approaches a trendline (using 5 mn bars) and I see the bar is a wide range bar (compared to previous bars), and it takes out the line, I have gained an edge. Specifically, I KNOW that if price takes out the line, probably it will continue (how far is unknown). I evaluate the context. How far will price continue. Is there a natural price level above or below that will stop price? If not, how likely is it that price will continue to the next pivot or price level? Is that level far enough away that I can make a reasonable profit vs. the risk of failure? I use the experience of years of observation of a particular market. I use the information gained from weekly characterization of market price action. I know where the levels are and how far away from this point they are. MY BACKTEST is happening in realtime. MY COMPUTER is my nervous system, and over time it has proven that it is faster, and has access to more sophisticated software than Tradestation, Wealthlab, TradersStudio, etc.
I place my bet and switch to my internal position management program, that mostly consists of keeping my hand off the mouse. If I have calculated my stop loss correctly, I know from experience that I have about a 72% chance of being right.

I'm not trying to be sarcastic. I just want to make a point. There is plenty to be gained from backtesting, IF one has the skills to do it right. There is plenty to be gained from Trendlines, or any other tool one chooses, IF one has the skills to use them productively. The idea (and it is common here) that one has to provide some currency of truth in the form of a backtest doesn't seem as compelling to me as it used to.

Good luck tomorrow.
Lefty

 

[Epiphany]

Trendlines work if you are prepared to use a wide stop

 

[hans37]

I think this discussion of trendlines working or not working is ridiculous.

Of course trendlines work and they are damn accurate too: for past price action.


PLEASE , if anybody knows how to trade the past and extract profits I'm all ears!


Oh how I would love to buy put future options on the s/p 10/15/87!! closing out 10/19/87