NexusFi: Find Your Edge


Home Menu

 





Research: trading pullbacks in CL


Discussion in Trading Journals

Updated
      Top Posters
    1. looks_one Georgii with 100 posts (38 thanks)
    2. looks_two missionatsea with 4 posts (0 thanks)
    3. looks_3 robert880 with 3 posts (0 thanks)
    4. looks_4 Big Mike with 2 posts (0 thanks)
    1. trending_up 16,689 views
    2. thumb_up 40 thanks given
    3. group 23 followers
    1. forum 113 posts
    2. attach_file 97 attachments




 
Search this Thread

Research: trading pullbacks in CL

  #91 (permalink)
Georgii
New York, NY, USA
 
Posts: 105 since Jun 2013
Thanks Given: 2
Thanks Received: 46

3/31/14, Monday



Another breakout pullback.

1. The market was in a trading range, with a pretty strong bearish trend toward the close. The Globex session reveals more confusion. By news time we see a bearish trend form.
2. The bearish trend is pretty strong and accelerates as it breaks through the low of the day.
3. The pullback is fairly strong.
4. More aggressive shorts would have taken the short red candle on the left, it did provide good risk reward but it would have failed immediately. The marked entry bar is a textbook example and offers good enough risk reward to be taken. The follow through is immediate and relentless.

Well, that wraps up March!

Reply With Quote

Can you help answer these questions
from other members on NexusFi?
New Micros: Ultra 10-Year & Ultra T-Bond -- Live Now
Treasury Notes and Bonds
NT7 Indicator Script Troubleshooting - Camarilla Pivots
NinjaTrader
Better Renko Gaps
The Elite Circle
Exit Strategy
NinjaTrader
Futures True Range Report
The Elite Circle
 
Best Threads (Most Thanked)
in the last 7 days on NexusFi
Get funded firms 2023/2024 - Any recommendations or word …
61 thanks
Funded Trader platforms
39 thanks
NexusFi site changelog and issues/problem reporting
26 thanks
The Program
18 thanks
GFIs1 1 DAX trade per day journal
18 thanks
  #92 (permalink)
Georgii
New York, NY, USA
 
Posts: 105 since Jun 2013
Thanks Given: 2
Thanks Received: 46

March 2014: overview

The month of March provided 16 days with potentially tradable signals out of 21 trading days (no market holidays this month). That's a total of 75% of all trading days.

January - March 2014: overview

As we total the results, we have 43 potential signal days.

January 16/21
February 11/19
March 16/21

For the period that constitutes 43 signal days out of 61 tradable days total had some form of potential opportunity, or 70% of trading days. That's in between two signal days every three days and three signal days every four days.

On some of those days we'll have one opportunity, on others more than one, so this does not mean we have 43 potential trades, we have more in our initial sample. Sometimes we may get re-entries (if we choose to allow them), on others not.

A quick calculation revealed 65 analyzed opportunities if we allow reentries.

Again, this is just three months' data, and this is without qualifying our signals (getting rid of those which would not constitute a high probability opportunity, the C|D category). We may well end up discarding 50% of our signals after we create more accurate classifications.

We have a number of things to study, having obtained this sample...

1. Should we treat breakout pullbacks differently from other pullbacks?
2. How should we treat higher time frame bias? Should we even have one? Should it matter where we are in the larger chart, in relation to yesterday's highs/lows? Should yesterday matter?
3. Should we filter for trend strength/weakness?
4. Should we filter for pullback strength/weakness?
5. Are there times when we should only wait for a complex pullback (second entry)?
6. Should we allow re-entry?
7. How tight should we really be with our initial stop?
8. How do we manage the trade in the multiple scenarios we've imagined? When do we trail our stop, when do we give it more room? What parameters of time and volatility will we take into account?
9. Are there situations where we should reverse position when the trade fails?
10. How do we treat the 512 tick chart in relation to the 5 minute?

These and possibly more questions will need to be carefully examined in the coming posts.

Let's also be humble and remember, we're looking at one small slice of the pie. We need a larger sample size to draw truly statistically valid conclusions, but if we were to begin our study with 200+ samples we could easily get lost in the heaps of data. We want to start with a sample large enough to show variation, yet manageable enough to allow us to focus. Afterward, we can start applying our ad-hoc conclusions to a larger sample and see how it performs, then make adjustments accordingly.

Reply With Quote
Thanked by:
  #93 (permalink)
Thxo
West Java
 
Posts: 145 since Apr 2014
Thanks Given: 143
Thanks Received: 123



Georgii View Post
March 2014: overview

We may well end up discarding 50% of our signals after we create more accurate classifications.

We have a number of things to study, having obtained this sample...


Hi, congrats for the journal and the result, it's seems very promising ...
btw, i'm using pretty similar pullback concept my self, the difference is just at the execution process ...

From my very limited experience, i found that if i add more rules/ classifications (curve fitting), more prone i am to the sudden behavioral change in the markets, and my losing streak probabilities are higher at this time periode (as far as 2x higher from the normal (curve fitted) market conditions).
but off couse, like i said before, this is just my very limited experience, you should test it for your self ...

Happy Trading

Reply With Quote
  #94 (permalink)
Georgii
New York, NY, USA
 
Posts: 105 since Jun 2013
Thanks Given: 2
Thanks Received: 46


Thxo View Post
From my very limited experience, i found that if i add more rules/ classifications (curve fitting), more prone i am to the sudden behavioral change in the markets, and my losing streak probabilities are higher at this time periode (as far as 2x higher from the normal (curve fitted) market conditions).
but off couse, like i said before, this is just my very limited experience, you should test it for your self ...

Thanks! Yes, we'll have to see just how many rules we'll actually need for this. I want to definitely increase the sample, this is just an initial test to get ideas.

Reply With Quote
  #95 (permalink)
Georgii
New York, NY, USA
 
Posts: 105 since Jun 2013
Thanks Given: 2
Thanks Received: 46

Trade Analysis: Instant Successes


Here I've selected the trading ideas that would have resulted in nearly immediate wins, that is trades that would likely survive tight trade management rules, such as swinging a stop below/above the extreme of each bar.

These are all simple pullbacks, or complex pullbacks that would not have triggered a simple pullback initially.



As we can see, there are 10 signal days and 13 marked potential opportunities which resulted in nearly instant profits. That's just 23% of signal days (42 signal days total, I excluded my first signal from the sample after analyzing it and determining I made a mistake so now we have 42 signal days vs 43), and 20% of all observed potential signals (13 out of 65 observations).

The quickest time to target was 1 bar (5 minutes), the longest time to target was 5 bars (25 minutes, example 2a, trade 4a). The average time to target was 2.3 bars, or 11 minutes and 30 seconds.

So as one can see given this particular modest sample, in less than a quarter of cases can we expect our trades to meet the target with virtually no hesitation or reentry required. This is important to know. It means that on average, on 16% of trading days we'll be able to have a 'perfect' pullback trade, with no stumbles or fumbles.

Reply With Quote
  #96 (permalink)
Georgii
New York, NY, USA
 
Posts: 105 since Jun 2013
Thanks Given: 2
Thanks Received: 46

Trade Analysis: Complete Failures (no reentries permitted)


Here are what I define as 'complete failures': trades that don't make it past 50% of the target and reverse to take us out at our initial stop. This assumes we are only trading simple pullbacks, or complex pullbacks that didn't trigger simple pullbacks, and we do not allow ourselves to re-enter if we get a second opportunity (a complex pullback).



The results at first glance are bit unpleasant, we get 18 potential trading opportunities that fail completely, 5 more than instant successes. That means 28% of our trades on average won't make it past 50% of our take profit target, and will end up hitting our stop.


This shouldn't be a cause for despondency however. Without exit management rules and analyzing risk reward, we can't draw conclusions on this alone. Also, we may be able to do a better job filtering our trades, and reentries may alter the picture in our favor. Finally, we haven't taken a look at the 'in between' trades and see if we can apply trade management rules to them that would skew the results in our favor.

Our average time to failure in this sample is 2.83 bars, or 14 minutes 9 seconds, noticeably longer than our average instant success time.

The breakdown of failures is as follows:

During the first bar: 5
During the second bar: 4
During the third bar: 2
During the fourth bar: 3
During the fifth bar: 3
During the sixth bar: 1

And as such, the cumulative measurements are:

Within the first bar: 5 (28%)
Within the first two bars: 9 (50%)
Within the first three bars: 11 (61%)
Within the first four bars: 14 (78%)
Within the first five bars: 17 (94%)
Within the first six bars: 18 (100%)

Reply With Quote
  #97 (permalink)
Georgii
New York, NY, USA
 
Posts: 105 since Jun 2013
Thanks Given: 2
Thanks Received: 46

Trade Analysis: "Jagged" Successes (no re-entries permitted)


Next we look at those simple pullbacks that hit our profit target but not without some hesitation and retracement. None of these would have hit the initial profit target with a simple 'trailing the stop just under/just over the last candle" exit strategy.

To clarify, the criteria are simple pullbacks, or complex pullbacks IF a single pullback was not triggered based on our rule set, which requires the counter-trend candle to be broken. We do not permit re-entry so far, so we're not analyzing those trades we would have been in potentially after sustaining a loss on an initial signal (note, this assumes the same setup - there could be more than one setup in a trading day).



There are five such samples here. While they are technical 'successes', it would be a bit brazen to classify them as such because depending on our stop trailing rules we may have taken only a bit of profit from them or maybe even a loss.

However, if we do add them to the 'instant success' category, we end up with not 13 but 18 observed potential opportunities that do hit their profit target within a reasonable amount of time. What this communicates to us is that out of 18 trades that do hit the profit target, 13 out of 18 (72%) will do so cleanly - at least in these sampled market circumstances.

Here we also get an important statistic: the maximum time it takes for one of our 'jagged' successes to hit its target is 7 bars or 35 minutes. This number (or an even lesser one) could serve as a maximum time cap on our trades. Granted, as I am prone to repeating, we are analyzing only a small sample here, and the given subset here of 5 trades is way too little to make any statistically meaningful conclusions. But with our general understanding of how these trades play out, we already see overall what can reasonably be expected, at least to the point where it can guide our further research.

Reply With Quote
  #98 (permalink)
Georgii
New York, NY, USA
 
Posts: 105 since Jun 2013
Thanks Given: 2
Thanks Received: 46

Trade Analysis: "Not Quites" and "Frozen", no re-entries permitted


This is the sort of the 'catchall' category for observations that came closer than 50% to the initial profit target, in some cases missing it by just a tick, then either stopping out or freezing for such a time that any prudent trade management system would have likely gotten us out before its final resolution as a complete win or loss. In essence these are trades that resolve neutrally and are potential scratches: some would be managed to a slight win, others to a slight loss, just as with the 'jagged' wins.



We have a total of 7 observations here. The maximum time with this sample for a losing resolution was 12 bars (60 minutes), and 11 bars (55 minutes) for a winning resolution - although this time is not really relevant, it is not likely we'd ever build a rule set that would allow us to stay involved for so long.

Reply With Quote
  #99 (permalink)
Georgii
New York, NY, USA
 
Posts: 105 since Jun 2013
Thanks Given: 2
Thanks Received: 46

Initial Conclusions: CL pullbacks, no re-entries permitted, 1/14 - 3/14

Let's review our trading template...

Rules:

1. We wait for a trend to emerge, either as a continuation of a prior trend or a breakout of a visible trading range, then wait for a pullback. We do not trade trends that are inside of the intra day range. We observe levels on the five minute chart that shows regular trading hours in crude (9:00 - 2:30 PM EST), and consult with a 512 tick chart that shows Globex levels as well.
2. We enter on a stop, on the first bar that breaks the extreme of the last counter-trend bar in the with trend direction.
3. For this given rule set only, we take simple pullbacks or complex pullbacks that would not have initially triggered a simple pullback (to clarify simple versus complex, go back to the first page of this topic) and we do not allow re-entries.
4. For this given rule set only, we take our signals from a five minute chart.
5. The profit target is the extreme of the trend.
6. We enter 'all-in' and exit 'all-out'. We do not scale in or out of our positions with multiple entries and exits.
7. We only enter trades from 9:05 - 11:30, we make no new entries after 11:30 (we are allowed to manage an existing position).
8. We cannot enter a position five minutes before an economic report, or fifteen minutes before a crude oil report.
9. We must exit five minutes before a crude oil report.

Our initial test over a span of 61 tradeable days reveals the following:

Potentially tradeable observations appear in 43 of 61 days (70%). There are a total of 43 observations given this rule set (no re-entries allowed).

13 observations (30%) follow through to the initial profit target in a direct, clean manner with very minimal retracement.
5 observations (12%) follow through to the initial profit target with some difficulty.
18 observations (42%) fail to reach 50% of the profit target, and half of these observations (9, or 20% of all observations) will hit the initial stop within 10 minutes.
7 observations (16%) get closer than 50% of the profit target but fail to hit it in a timely manner or eventually fail, requiring an early exit that would likely result in a 'scratch'.

Looking at the information the way it is, it's very difficult to really tell if there is an edge here or not. Of course, life would be very easy if we saw that 60% of our observations hit our profit target without hesitation - it wouldn't matter if the other 40% went straight for our stop on the first bar, we would end up ahead assuming we never took a risk that was greater than our profit target, and that is net commissions and slippage.

If we just left the trade open and kept our stop and profit target orders in the order book and walked away, it's very unlikely we'd have an edge with this rule set. Assuming the last mixed category of "Not Quites" and "Freezes" balanced out (they almost do, 3 of them end up as winners and 4 as losers if kept on for 60 minutes), we'd have an equal number of trades that hit the profit target and an equal number that failed in this sample set.

Assuming we had 1:1 risk reward on each trade, we also have commissions and slippage to worry about which would put us at a minus. Even if the risk reward is slightly skewed in our favor (which may be the case in our sample set), we would have only a razor thin edge. We also have to take into account we may miss some of these trades or make execution errors, and we should assume out of conservatism that the errors won't work in our favor.

There are three more variables here:

* Trade selectivity.
* Re-entry opportunities: permit or no?
* Exit strategy.

Each of these variables can influence the edge of this strategy in one or the other direction.

Become too selective in the trades, and you may curve fit your system too strongly and miss out on good opportunities. A 90% accurate system that trades just once a month will make you less money than a 60% system that trades twice a week.

Re-entries may be a good idea IF they offer the same or better odds. Obviously if our odds get worse with re-entries or they cause no difference to our PnL in the long run, we would want to avoid them.

Exit strategy has multiple variables to it that influence profitability. There needs to be a golden mean between trailing the stop too quickly and keeping it too loose. We may have rules that allow us to exit manually without waiting for a stop to get hit. We may bring our take profit closer in certain circumstances and 'bracket'.

All three of these variables also need to take into consideration our psychological preferences. A certain approach may mathematically work out better but we won't be able to execute it properly, making it impossible to improve our edge and possibly even deteriorate it. For example, re-entries if not kept in check may cause a bout of revenge trading.

These are things we'll take into consideration as we continue testing this idea, and once again let me emphasize - we are only looking at a small wedge of the pie here. Let's not be too fast to make conclusions based on what we see here, we need to allow the law of large numbers to play out properly.

Reply With Quote
  #100 (permalink)
Georgii
New York, NY, USA
 
Posts: 105 since Jun 2013
Thanks Given: 2
Thanks Received: 46


Test Results: Swing stop to candle extreme, no re-entries

Now we've moved onto testing specific exit strategies, which is already the beginnings of a complete system: we have rules to enter and exit. At the moment we're operating without permitting re-entries.

First, let me say that when I began testing the trades I noticed that some of my marked entries (a total of 14 out of 48 analyzed observations) yielded less than a 1:1 risk reward as I had initially relied on my eye to select them. Still, I kept them in my analysis pool because I didn't want to distort the sample I selected. 50% of them were winners and 50% were losers.

The first set of exit rules I decided to test was the 'swing under/over the previous candle' rule. What this simply means is that if you're in a long trade, once you get triggered into the trade you swing your stop underneath the last candle's low, and if you're in a short trade you swing your stop above the last candle's high. This is done as soon as the candle closes in both cases.

There are no other rules here: it doesn't matter where the price happens to be within the trade, or how far the candle moves in our favor - e.g. if it covers 20% of our profit target or 99%, the stop is still underneath the last candle until the current one closes. Every five minutes we trail the stop to the next candle extreme, until the market hits our stop or we hit our take profit target. There is no time stop here either.

It's a very simple, stupid type of trade management strategy which requires virtually zero thought or discretion.

In my test, I included 1 CL point for slippage and 0.5 CL points for commissions to account for our trading frictions. Limit orders were only filled if the price traded through them (e.g. if we were long and the sell limit order was at 99.45, we would only be filled if the price got to 99.46). In each trade, I set the take profit target to one point below the extreme of the trend, so if the trend peaked at 100.33, the limit sell order was set to 100.32.

Our results from our selected samples from 1/14 - 3/14 using this rule set are as follows:

Hypothetical trades analyzed: 48
Met profit target: 15 (31%)
Positive: 17 (35%)
Negative: 31 (65%)

Net PnL (incl. comm + slippage): -23 CL points, -2.51x R

Average CL points per trade: -0.5
Median CL points per trade: -5

Average initial take profit target in relation to %R (before commissions and slippage): 1.35x
Median of above: 1.19x

Account peak: +76 CL pts
Account trough (max drawdown): - 44 CL pts

Below is a graphical representation of our cumulative profits in %R terms, and our trade by trade performance in %R terms (in case you're wondering, when expressed in point or dollar terms the graphs are nearly identical).


Reply With Quote




Last Updated on October 19, 2014


© 2024 NexusFi™, s.a., All Rights Reserved.
Av Ricardo J. Alfaro, Century Tower, Panama City, Panama, Ph: +507 833-9432 (Panama and Intl), +1 888-312-3001 (USA and Canada)
All information is for educational use only and is not investment advice. There is a substantial risk of loss in trading commodity futures, stocks, options and foreign exchange products. Past performance is not indicative of future results.
About Us - Contact Us - Site Rules, Acceptable Use, and Terms and Conditions - Privacy Policy - Downloads - Top
no new posts