(binned per thread starter's request) Researching and testing | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Mismatch between live forward testing and back testing 4 replies

Researching Saadscalper ea 3 replies

Researching Litewave expert 12 replies

Researching Admiral Markets Ltd 19 replies

Attachments: (binned per thread starter's request) Researching and testing
Exit Attachments

(binned per thread starter's request) Researching and testing

  • #1
  • First Post: Edited Jun 1, 2024 11:52am May 28, 2024 7:32pm | Edited Jun 1, 2024 11:52am
No long intros, straight to the point.

There is a lot of academic articles about patterns (e.g .https://www.sciencedirect.com/science/article/abs/pii/S1568494619304818) and other methods of analyzing financial time-series. But ~99.99% of them are too difficult (for me) to understand and reproduce results.
Let's try to make things more down to earth.

Feedback, ideas, etc. are welcome. The further development of the thread depends on it.

Hammer (https://www.forexfactory.com/thread/post/14886848#post14886848) = Inverted Hammer - No better than a random candle regardless of its location
Air force
  • #2
  • Edited 8:05pm May 28, 2024 7:45pm | Edited 8:05pm
Starting with a Hammer.
GU. Hourly tf. Dataset: 31 167 candles.

lw - lower wick
uw - upper wick
bd - body
&& - logical operator 'and'

Logical rules:
Inserted Code
lw > (bd*2) && bd < 50 && uw < 2 → (lower wick is 2 or more times bigger then a body,
body occupies less then 50% of a range, upper wick is less then 2% of a range)
For simplicity location of a patterns (downswing, upswing) doesn't matter.

Found 496 items. Occurrence ~ 1.6%
Hammer is a bullish pattern, so the point of interest is how high price can go after conditions are met. Let's see 200 candles into the future.
Below is a sample of 20 examples.
Attached Image (click to enlarge)
Click to Enlarge Name: ham.png Size: 30 KB



Quantiles.
Left - maximum profit
Right - max DD. (lowest price since Hammer's close)
e.g 0.0022 = 22 pips
Attached Image (click to enlarge)
Click to Enlarge Name: Mathematica_hp8e9KtOPP.png Size: 15 KB




On which candle was the highest price in a sequence
Attached Image (click to enlarge)
Click to Enlarge Name: Mathematica_EOp2qN0XJp.png Size: 15 KB



and the lowest
Attached Image (click to enlarge)
Click to Enlarge Name: Mathematica_4W7wxTEQsW.png Size: 18 KB

Ratio max_p/min_dd from each case. Values below -1 indicating profit is always greater then DD. However, as we can see from this and other stats/plots - it's symmetric. But in reality there is always a profit, that can be managed/increased with multiple entries.
Attached Image (click to enlarge)
Click to Enlarge Name: Mathematica_LBks19h6co.png Size: 4 KB



Details for each case

Attached File(s)
File Type: pdf table.pdf   26 KB | 198 downloads
Air force
  • #3
  • May 30, 2024 3:02am May 30, 2024 3:02am
You need to rule out the null-hypothesis "the hammer has no predictive power". You have stats about 496 hammers on G/U H1. Now randomly pick 496 non-hammer candles and do the same stats. If the results are too similar the hammer is nothing special.
No greed. No fear. Just maths.
  • #4
  • May 30, 2024 6:18am May 30, 2024 6:18am
Quoting PipMeUp
You need to rule out the null-hypothesis "the hammer has no predictive power". You have stats about 496 hammers on G/U H1. Now randomly pick 496 non-hammer candles and do the same stats. If the results are too similar the hammer is nothing special.
Thanns, PipMeUp for very useful suggestion.
After dozens of random experiments the conclusion is (p-value varies, but..)
The null hypothesis that the datasets have the same distribution is not rejected at the 5 percent level based on the Kolmogorov-Smirnov test.

I'm going long at:
1. Random bearish candle

-Profit and drawdown quantiles

Attached Image (click to enlarge)
Click to Enlarge Name: Mathematica_7s8O7CPB4i.png Size: 8 KB


-Hypothesis tests

Attached Image (click to enlarge)
Click to Enlarge Name: Mathematica_iQN60FBhr2.png Size: 5 KB

Attached Image (click to enlarge)
Click to Enlarge Name: Mathematica_zpX9wf2F1X.png Size: 8 KB


2. Random bullish candle

-Profit and drawdown quantiles

Attached Image (click to enlarge)
Click to Enlarge Name: Mathematica_oTV0nNWukm.png Size: 8 KB


-Hypothesis tests. In this particular sample, h0 is tossed out.

Attached Image (click to enlarge)
Click to Enlarge Name: Mathematica_BOKVCRvX4a.png Size: 6 KB

Attached Image (click to enlarge)
Click to Enlarge Name: Mathematica_bwtI7IwcO2.png Size: 7 KB



So, the Hammer pattern is nothing special.

I don't think that this research is useless, at least there is some stats about market moves n-candle further and profit/DD. Will continue tests against other 2-candle patterns, etc and markets

Air force
  • #5
  • May 30, 2024 6:39am May 30, 2024 6:39am
Quoting W0lfram
But in reality there is always a profit, that can be managed/increased with multiple entries.
Can you please elaborate on that? Based on your graphs there doesn't seem to be SL/TP levels that on average give a profit.

I love your systematic detailed analysis! No gut feeling but exact science. What is not to love? ;-)
Looking forward to your next installments!
  • #6
  • May 30, 2024 6:40am May 30, 2024 6:40am
Quoting W0lfram
the Hammer pattern is nothing special.

A hammer is an individual candle, not a pattern. A "pattern" is a formation comprising a structure of two or more candles/bars.
  • #7
  • May 30, 2024 7:33am May 30, 2024 7:33am
Quoting yoriz
{quote} Can you please elaborate on that? Based on your graphs there doesn't seem to be SL/TP levels that on average give a profit. I love your systematic detailed analysis! No gut feeling but exact science. What is not to love? ;-) Looking forward to your next installments!
Hi, thanks!

Exactly, there is no stop loss. The DrawDown table shows the potential losses one might experience. And 'Green' table helps to see probabilities where to exit or lock profits if the trade instantly goes in favor.

This is suitable if I trade with a leverage of 1:2 or less. Additionally, there are always new signals to average the position. Or hedge if opposite signal comes.
Air force
  • #8
  • Edited 8:26am May 30, 2024 7:55am | Edited 8:26am
Quoting KarlKraus
{quote} A hammer is an individual candle, not a pattern. A "pattern" is a formation comprising a structure of two or more candles/bars.
I believe, in every book and research article about that topic,
it's called a pattern because it occurs at a specific location (not necessary) and is typically followed by supposed price action. And falls into subcategory 'single-candle' patterns.
Number of involved candlesticks doesn't matter.
Air force
  • #9
  • May 30, 2024 8:39am May 30, 2024 8:39am
Quoting W0lfram
Ratio max_p/min_dd from each case. Values below -1 indicating profit is always greater then DD. However, as we can see from this and other stats/plots - it's symmetric. But in reality there is always a profit, that can be managed/increased with multiple entries. {image} Details for each case {image}
What? Can you explain why you use maximum Profit / minimum drawdown and in which manner? Over 200 hammers? Thank you!

I love to read it. Straight to the point. However, can you also please explain the 0% to 100%? what does that number says?
In 10% of the cases we have max. Profit of 0.0026 (26 Pips)? I'm not that fit in stats.
  • #10
  • May 30, 2024 9:37am May 30, 2024 9:37am
Quoting JackintheBox
{quote} What? Can you explain why you use maximum Profit / minimum drawdown and in which manner? Over 200 hammers? Thank you! I love to read it. Straight to the point. However, can you also please explain the 0% to 100%? what does that number says? In 10% of the cases we have max. Profit of 0.0026 (26 Pips)? I'm not that fit in stats.
I've used confusing terms here, sorry. Minimum - because is the lowest price in a sequence, so its actually maximum dd
Quote
Quantiles
Left - maximum profit
Right - max DD. (lowest price since Hammer's close)
e.g 0.0022 = 22 pips
Attached Image (click to enlarge)
Click to Enlarge Name: Mathematica_Y1zZI1qWzH.png Size: 19 KB
Quote
Over 200 hammers?
After hammer appears, we see at the high/lows of the next 200 candles. 496 hammers was found.
Thanks to PipMeUp, now we know that hammer is no different from a any random candle.


Quote
In 10% of the cases we have max.
Yes. 10% of values in dataset are below 26 pips. So if we gonna trade it, ~90% of the time we get 26 pips profit or more. But be ready for big DD, that's why only leverage 1:1 for this kind of 'strategies'
Air force
  • #11
  • Edited 10:53am May 30, 2024 10:32am | Edited 10:53am
Quoting W0lfram
I've used confusing terms here, sorry. Minimum - because is the lowest price in a sequence, so its actually maximum dd
I believe commonly used names for this are: Maximum Adverse Excursion (MAE) and Maximum Favorable Excursion (MFE).

EDIT: Uh, no my bad. Since we are looking for both long and short opportunities there is no "favorable" direction. So maximum/minimum excursion is indeed more accurate.
  • #12
  • Edited 12:07pm May 30, 2024 11:42am | Edited 12:07pm
As a proper scientist I tried to reproduce the results of a peer... (Well, actually I was just curious ;-))

I get the same results on GU. Another way to look at the data is to plot the minimum/maximum excursion of both a random sampling and the hammer pattern on top of each other. There is no significant difference:
Attached Image (click to enlarge)
Click to Enlarge Name: screenshot.png Size: 65 KB


I wondered whether perhaps the "hammer-effect" was just short-lived. So I did the same exercise on the M5 timeframe (i.e. detect a M5 hammer-candle and measure excursion over 200 M5 bars). Again no significant difference:
Attached Image (click to enlarge)
Click to Enlarge Name: screenshot.png Size: 56 KB
  • #13
  • May 30, 2024 2:41pm May 30, 2024 2:41pm
Quoting W0lfram
Starting with a Hammer. GU. Hourly tf. Dataset: 31 167 candles. lw - lower wick uw - upper wick bd - body && - logical operator 'and' Logical rules: lw > (bd*2) && bd < 50 && uw < 2 → (lower wick is 2 or more times bigger then a body, body occupies less then 50% of a range, upper wick is less then 2% of a range) For simplicity location of a patterns (downswing, upswing) doesn't matter. Found 496 items. Occurrence ~ 1.6% Hammer is a bullish pattern, so the point of interest is how high price can go after conditions are met. Let's see 200 candles...
You are right, hammer is not much of a reason to enter. In your conditions, did you make sure to add as condition that a bullish hammer has new low in an above average sequence of bear candles? Because, without that, it is not a bullish hammer. Also, please compare hammer and inverted hammer. In my studies years back, inverted hammer was better. Unfortunately lost that data, but your nice thread is bringing back old memories. Cheers and can't wait for more !!!!
If I don't mention SL and/or TP = longer-term setup possibly in the works!
  • #14
  • May 30, 2024 3:42pm May 30, 2024 3:42pm
From what I had tested, the best patterns are hammers / shooting star (PB) and outside bars (BEOB, BUOB)
  • #15
  • May 30, 2024 4:43pm May 30, 2024 4:43pm
Quoting KarlKraus
{quote} A hammer is an individual candle, not a pattern. A "pattern" is a formation comprising a structure of two or more candles/bars.
I see what you are saying Karl, and I get your point.
Consider the other attributes of a hammer bar: it is also true, that the hammer could be classified as a composite of an underlying pattern, and simultaneously part of a pattern of a higher time frame candle.
Such is the fractal nature of price movement.

The context of where the bar is matters. Is it close a weekly support or resistance zone. Did market change direction at that particular pivot? Day of week, etc....

Market behaviors such as the morning session having twice the normal ATR, or bars generated by news events, moon phases, seasonal tendencies, economic policy, ... all the funny mentals (fundamentals).
I saw a test suggesting that the bearish engulfing bar, is actually a bullish indicator.

Then there is all the statistical work done by "The Rumpled One" with regard to high low break outs.

Other opinions: https://www.forexfactory.com/thread/256823-there-are-no-secret-patterns-and-there-will
There ain't no use in cryin' It doesn't change anything." QOTSA
  • #16
  • May 30, 2024 4:47pm May 30, 2024 4:47pm
Attached Image (click to enlarge)
Click to Enlarge Name: -5078290943798519990_120.jpg Size: 7 KB
  • #17
  • May 30, 2024 4:47pm May 30, 2024 4:47pm
Attached Image (click to enlarge)
Click to Enlarge Name: -5078290943798519992_120.jpg Size: 12 KB
  • #18
  • May 30, 2024 4:47pm May 30, 2024 4:47pm
Attached Image (click to enlarge)
Click to Enlarge Name: 107926964_173662.jpg Size: 8 KB
  • #19
  • May 30, 2024 4:48pm May 30, 2024 4:48pm
Attached Image (click to enlarge)
Click to Enlarge Name: Inverted Hammer.jpg Size: 8 KB
  • #20
  • May 30, 2024 4:48pm May 30, 2024 4:48pm
Attached Image (click to enlarge)
Click to Enlarge Name: 806622710_179418 copy.jpg Size: 130 KB
0 traders viewing now