Trade Management Functions The Relationship Between Entries, Exits and Stops

[email protected] LBRGroup.com

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

Testing for the Edge

Randomly entering a position can produce a profit 65% of the time. By applying simple filters, random entry can become breakeven or even slightly profitable. This data gives a benchmark for a systematic trader to beat in order ensure the system is successful due to more then capitalizing off noise. Noise is analogous with random price data.

Lets go through a 5 step exercise using a randomizer to generate trade entries as a departure point for modeling price behavior.
Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

ATR (Average True Range) Stops

ATR = Average True Range True Range includes gap area + Days range ATR is a moving average of the past Ndays True Range.

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

Relationship between targets & stops with random entries

Stop and reverse off randomly generated trade points yields a 50% win/loss ratio as well as a profitable system 50% of the time.

Random entry with Large Target and Large Stop (3 ATRs): The win/loss ratio still hovers around 50%. (3 ATRs is just outside the noise inherent in price action).
Random entry with Small Target and Large Stop (1ATR/3ATR). The win/loss ratio jumps up to 70%+. While the chances of getting 1 ATR before a 3 ATR stop is hit are greater then 70%, it still is not a profitable system. If we use a target of .5 ATRs and leave the large stop in place, the win % jumps to 82%. The increase in % win should be encouraging for the short term discretionary trader.

Random entry with Small Target and Time Stop (1ATR/8bars). Though the % win drops to 65%, the size of the average loss decreases. So, while a larger fixed stop gives a trade more time to be profitable, a discretionary trader may choose to exit if the trade does not work within a reasonable amount of time.
Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

A profitable system that is durable and robust based off random entries!
At this point we know we can randomly enter a trade and get better then an 80% win rate on most markets. By adding a trend filter, random entries can be made into a profitable system. Only long entries are taken in an up trend and only short entries are taken in a downtrend. (Again, all entries are at randomly generated points). The Profit Target is 4 ATRs and the Stop Level is set to 3 ATRs (just outside of the noise). This ensures that if we can get better then a 50% win rate, the system will be profitable. This system was run 100 times on each market with randomly generated trade entries each time. The time period tested was from 1995 to 2005 on 22 domestic futures markets that included currencies, bonds, index futures, metals, agricultural products, and softs.

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

Statistical profile of random system with one Trend Filter:

The system was profitable between 92 100% of the time on all markets with the exception of live cows and hogs, (which tend not to trend for extended periods of time), as well as silver and natural gas (probably due to its extreme spiky nature in recent years). All index futures were profitable 100% of the time over 100 randomly generated runs over 10 years of daily data! This is not a recommendation to trade this way, but RANDOMLY generated entries can lead to profitable strategies with use of one filter if stops are placed just beyond the noise and a larger target is played for.
Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

Summary:
Random entry produces a high win/loss rate when the stop is large enough to be just outside the noise. By using a time stop, we can avoid taking repeated large losses at the expense of a slightly lower win rate. With addition of one trend filter, a larger target can be played for. These raw statistics provide a hurdle that a system must exceed in order to produce better then random results. A system can be considered better then random entry when:
The drawdowns are smaller, or The P&L is substantially larger then random entry, or The win rate is significantly higher then that of random entry.

The first step to improving a system is by the addition of some type of trend filter. The greatest edge us trading in the direction of the trend.

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

Relationship between Entries, Stops, and Profits There is not always a relationship between a trade is entered and where a trade is stopped out. For example, with time based stops, the stop out level has no relationship to the initial entry price. Its not necessarily the entry that makes a model or system successful. Its all about the exit strategy!
The larger the target, the lower the % win rate. The smaller the target, the fewer times the stop will be hit (all else being equal). The wider the stop, the higher the % win rate.
Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

Volatility Breakout or Channel breakout as an Entry.

Volatility Breakout systems use an increase in range to enter a trade. The assumption is that an expansion in range indicates an imbalance in the supply/demand equation that is a driver behind trends. Multiple permutations of entry techniques and exit strategies were examined on 22 different domestic futures markets over the last 5 years of data. (2001 2005). Filters such as entering only after narrow range days were also examined.

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

Volatility Breakout System Daily SP

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

10

Results of Analysis of Breakout Entries:

There IS a statistically significant positive expectation to Volatility or Channel Breakout systems. Testing shows results are best when playing for a small target or shorter-period time exit (i.e., 1-2 days). Examination of multiple Trailing Stop strategies shows that they degrade system performance. Examination of three exit strategies - fixed ATR, time stop (such as next days close), or structural point (such as 1 or 2 day high or low) shows that the structural point as an exit is optimal for short term volatility breakout. (This is not the case with Channel breakout). Unfortunately, much of the edge from these systems, particularly short-term volatility breakout, tends to be eaten up by friction (lack of liquidity or slippage at entry points).
Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

11

Channel Breakout: Channels have widened out as markets have matured

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

12

It is folly to try and anticipate which breakouts will work!

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

13

Momentum Driven Models

While Volatility and Channel breakout systems are purely price driven, momentum functions are based on a derivative of price. Momentum has many interesting properties, including a leading function for price. The tradeoff is an increase in noise over smoothed functions such as moving averages, which have a lag. What is the best combination of exit strategies, stops and targets to compensate for this increase in noise? A variety of look back periods for momentum functions were examined and tested against permutations of fixed targets ranging from .5 ATRs to 3 ATRs. Fixed stops, time stops, as well as delayed entries (so as not to be entering on a momentum spike were also tested. Similar conclusions could be drawn as what was modeled with random entry systems: Smaller targets are more profitable then Large targets, Time stops are better then fixed stops, and Time stops combined with Fixed stops are optimal. Momentum driven models perform better with a weighting given to additional use of time stops which was not the case with random generated entries.
Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

14

Comparison of 3 & 10 period momentum functions. Both forecast higher (or lower) prices about 66% of the time.

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

15

Last Word on testing Momentum Models

Momentum models are profitable across a wide range of parameters (robust) and they have a higher then above average win rate. Despite the high win/loss ratio, a system tends to be profitable only 66% of the time for a given market for a given year. Momentum systems work best on a systematic basis if applied across a portfolio of markets as opposed to one individual market. The forecasting value for momentum models is of short duration only and trailing a stop in attempt to capture a higher degree of trendiness tends to degrade their performance. Momentum functions may have their greatest value for shorter term discretionary trading or as a trigger component for a more complex strategy.
Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

16

Momentum Systems + Trend Filter

A trend filter increased the profitability of our random entry system. What happens when a trend filter is added to a Momentum System to create a more complex strategy? In a nutshell: the win % rate stays the same, the number of trades is reduced, and the net profit may increase slightly. On any given year, a basket of 20 markets will still show only 66% of the markets as being profitable, (though one that was unprofitable on year may be profitable the next). System profitability may also decrease since fewer trades are taken. Momentum strategies provide a tradable edge, but have a short period of forecasting value regardless of whether a trend filter is used or not.

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

17

Modeling Extended Runs

The term Impulse is used to note a sudden change in supply/demand (via an increase in range or momentum). However, another way of quantifying degree of trend is through the use of moving averages. An above average number of closes on one side of a moving average can be an indication of a stronger period of trendiness. Capturing a stronger degree of trendiness is one of the places that a Big Win come from. We will use the term extended run to describe a more persistent imbalance in the supply/demand relationship. This imbalance leads to the type of trend that is also characterized by an increase in market efficiency, decrease in noise, and thus lends itself to trailing stops as a trade management tool.

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

18

Profile of an Extended Run

The most persistent trends tend to be marked by multiple consecutive closes on one side of a short moving average. For purposes of this presentation we will examine consecutive closes on one side of a 5 period simple moving average. 7 closes on one side of this SMA are considered to be significant. The same properties found in our momentum studies were found to hold true with time functions on one side of a moving average. Here is the profile for entering a trade on the first close on the opposite side of the SMA and exiting it on the first close that is in the direction of the original trend:
%Win = +70%; average loss ~ 2x average win; profitable in 90 - 95% of a basket of 22 markets on any given year. Win/loss

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

19

Retracement Trades after Extended Run

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

20

Nat Gas losing long trade(note: current extended run underway to downside).

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

21

Additional Observations on Extended Runs

We modeled the number of consecutive closes on one side of the 5 SMA testing between 5 to 25+ consecutive closes. As an interesting aside, the % win on entering on the first retracement trade after an extended run dropped off significantly after 18 consecutive closes on one side of the 5-period simple moving average. The implications are that the market was more likely to have had a buy or sell climax after such an extreme run or the market was prone to a more significant reaction in the opposite direction after an extreme markup or mark down move.

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

22

23 days on one side of SMA perhaps TOO extreme!

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

23

Summary of Retracement Strategies after a move of above average price duration:

Extended runs are a different way of measuring the degree of trend. The first reaction in the opposite direction leads to a trade opportunity for a high probability trade (high win%). Once again, it tests out best when the trade target is small and the initial stop is large. There is no advantage to trailing a stop for trade management purposes and in fact this will degrade the system results.

ULTIMATELY, to capture a bigger win, we must examine the conditions that precede an Extended Run. Can this be done through torturing the data? Our testing has shown us that this is difficult to do with any degree of consistent success. Instead, we must now depart the wonderful world of statistics and move into a more abstract analysis of STRUCTURE!

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

24

Trailing Stop Functions

ONLY should be used at specificf structural points! Breakouts, Trend reversals as defined by the wave structure, and after A-B-C formations in a trending market

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

25

Market Waves

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

26

Comparison of different structural points small target versus large target

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

27

Sweet spot with detail on short time frame

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

28

Stops trail stop on HIGHER time frame

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

29

Trailing stops- 3 bar moving average of the lows overlaid on lower time frame

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

30

Daily parabolic versus daily 3-bar HIGH overlaid on the hourly chart

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

31

No trailing stop is going to get you out at the highs, but it will probably keep you in your trade then otherwise. At the breakout point, it is preferable to trail a stop then play for fixed target.

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

32

Volatility Stops

ATR = Average True Range 3 ATR +/- highest high or lowest low of last 10 bars 3 ATR +/- moving average of close, high or low.
Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

33

DAILY CRUDE VOLATILITY STOP

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

34

BEANS, daily chart, ATR trailing stop

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

35

OCO Bracket Functions

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

36

Execution Platforms

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

37

Original Keltner channels on 5 minute SP chart

Copyright (c) LBRGroup 1996-2006. All Rights Reserved.

38

The End!

Thanks for Attending.