Richard Donchian 5MA - 20MA question

How do you know when a trend has started? Ended? This forum is for discussions about trend indicators and signals.
Post Reply
rschiller
Contributing Member
Contributing Member
Posts: 8
Joined: Wed Nov 03, 2004 9:02 pm

Richard Donchian 5MA - 20MA question

Post by rschiller » Thu Feb 03, 2005 3:04 am

Greetings to all, my first post and hopefully in the right forum. I am trying to obtain further information and understanding of Richard Donchian's methods. Quoted from Perry Kaufmann:

"The current penetration must not only cross the 20-day moving average but also exceed any previous 1-day penetration of a closing price by at least one volatility measure. The 5-day moving average serves as a liquidation criterion (along with others) and is also modified by prior penetration and volatility."

What is meant by "one volatility measure"?
The signal is crossing the 20 day; is there a system that generates signals when the 5MA crosses the 20MA?
I would much appreciate any further information and understanding of Donchian's 5/20 method(s).

Best to all!

Rick

William
Roundtable Knight
Roundtable Knight
Posts: 238
Joined: Sun May 04, 2003 4:41 pm
Location: Manhattan, New York

Post by William » Thu Feb 03, 2005 11:09 am

You may want to look at c.f.' report Original Turtle Trading Rules. He discusses the concept of a volatilty measure.

rschiller
Contributing Member
Contributing Member
Posts: 8
Joined: Wed Nov 03, 2004 9:02 pm

I think I understand Volatility measure in Turtle System but

Post by rschiller » Thu Feb 03, 2005 11:45 pm

The volatility measure, or N, is the 20 day exponential MA of the ATR and is used for position sizing and stop placement. It seems Donchian is referring to perhaps the same thing or perhaps something else, but more as a filter as to when to enter a trade. This is the question I have.

thanks to all

RS

PruBear
Senior Member
Senior Member
Posts: 29
Joined: Sat Aug 23, 2003 11:03 am
Location: New Jersey

Post by PruBear » Sat Feb 12, 2005 12:01 am

I'm also interested learning more about Donchian's theorys.

I remember reading somewhere a quote from Daniel Dunn where he said Dunn Capital had run a bunch of tests on Donchians weekly rules and they had yielded interesting results. I seem to remember seeing some other references to them over the years (in market wizards I think?).

I haven't been able to pin down just what these weekly rules are but then maybe I'm looking in the wrong places.

I have read the tribute to Donchian on Seykota's Trading Tribe website but it really isn't specific enough to code and seems to be composed largely of guidelines.

Does anyone know what Donchian's weekly rules are or where a discussion of them can be found? I'd like to run some experiments using them when VT 2.0 becomes available. I'm a bit of an amateur but I can probably test them in XL in the meantime.
Any thoughts?

rschiller
Contributing Member
Contributing Member
Posts: 8
Joined: Wed Nov 03, 2004 9:02 pm

re: Donchian signal

Post by rschiller » Sat Feb 12, 2005 12:39 pm

I posted the original thread. Do you interprete Donchian's entry signal as the price only moving above or below the 20day MA and by some criteria that I originally asked about: "one volatility measure" or do you interpret the entry as the 5MA crossing the 20MA in the direction of the entry and being filtered by the "one volatility measure"?

And I'm still trying to determine what that measure is.

Rick

================

PruBear wrote:I'm also interested learning more about Donchian's theorys.

I remember reading somewhere a quote from Daniel Dunn where he said Dunn Capital had run a bunch of tests on Donchians weekly rules and they had yielded interesting results. I seem to remember seeing some other references to them over the years (in market wizards I think?).

I haven't been able to pin down just what these weekly rules are but then maybe I'm looking in the wrong places.

I have read the tribute to Donchian on Seykota's Trading Tribe website but it really isn't specific enough to code and seems to be composed largely of guidelines.

Does anyone know what Donchian's weekly rules are or where a discussion of them can be found? I'd like to run some experiments using them when VT 2.0 becomes available. I'm a bit of an amateur but I can probably test them in XL in the meantime.
Any thoughts?

Turbowagon
Roundtable Fellow
Roundtable Fellow
Posts: 82
Joined: Wed Oct 08, 2003 1:19 pm
Location: Washington, DC

Donchian

Post by Turbowagon » Wed Feb 16, 2005 1:25 pm

Prubear,
I found this information on Donchian:

http://www.chartfilter.com/articles/pricechannel.htm

Also, you may wish to search Seykota's site for the word "Donchian":
http://www.seykota.com/tribe/search/index.htm

Rick,
To help answer your original question, I believe Donchian specified certain units that the close had to penetrate the moving average by in order to validate a signal. It seems similar to the entry and exit threshold that can be used in many of the Veritrader systems. This is only my interpretation as I was not lucky enough to meet Richard Donchain myself. Apparenly Ed Seykota was so lucky.

See his site: http://www.seykota.com/tribe/Resources/ ... /index.htm

I have not found any information on how he calculated this volatility measure, but ATR would seem to be a decent choice.

Hope this helps,
eD

rschiller
Contributing Member
Contributing Member
Posts: 8
Joined: Wed Nov 03, 2004 9:02 pm

Re: Donchian

Post by rschiller » Thu Feb 17, 2005 12:15 am

Turbowagon, I posted the original thread here. I suspect it is some type of ATR and I have previously looked the links you included, thank you. The first is fairy extensive and inlcludes the following concerning 5/20:

Basic Rule A: Act on all closes that cross the 20-day moving average by an amount exceeding by one full unit the maximum penetration in the same direction of any previous closing when the closing was on the same side of the moving average.

Is the "one full unit" the amount the previous day's close penetrated and closed away from the 20MA or the range of that day, the ATR, if the close was not at the high or low? or was it . . . ??

You're probably right on using ATR but which: 10 MA or 20 EXP-ma or ??

thanks to all,

Rick


turbowagon wrote:Prubear,
I found this information on Donchian:

http://www.chartfilter.com/articles/pricechannel.htm

Also, you may wish to search Seykota's site for the word "Donchian":
http://www.seykota.com/tribe/search/index.htm

Rick,
To help answer your original question, I believe Donchian specified certain units that the close had to penetrate the moving average by in order to validate a signal. It seems similar to the entry and exit threshold that can be used in many of the Veritrader systems. This is only my interpretation as I was not lucky enough to meet Richard Donchain myself. Apparenly Ed Seykota was so lucky.

See his site: http://www.seykota.com/tribe/Resources/ ... /index.htm

I have not found any information on how he calculated this volatility measure, but ATR would seem to be a decent choice.

Hope this helps,
eD

mdtmn
Contributor
Contributor
Posts: 1
Joined: Sat Apr 30, 2005 11:18 pm

Re: Donchian

Post by mdtmn » Sat May 07, 2005 10:22 am

[Donchian's five- and 20-day moving averages.
Futures (Cedar Falls, Iowa) - November 15, 1995
Richard Donchian
Word count: 2423.
citation details
________________________________________
On Wall Street there are two conflicting adages:
1. "You'll never go broke taking a profit."
2. "Cut your losses short and let your profits ride."
Experience has shown that in commodities trading, the first of these "old saws" is dangerous and misleading, while the second may well be regarded as the one lesson the inexperienced commodity trader should learn if he wishes to have a better-than-even chance to come out ahead.
Every well-designed, trend-following, loss-limiting method for trading in futures (or stocks) rests on the basic principle that a trend in either direction, once established, has a strong tendency to persist, at least for a time. Among the many trend-following approaches now in use are the Dow Theory, point-and-figure chart techniques, swing methods (other than the Dow Theory), trendline methods, weekly-rule methods and moving average methods. We'll focus on moving average methods and, in particular, the comparatively simple five- and 20-day moving average method.
Building a moving average
According to Webster, an average is "the quotient of any sum divided by the number of its terms." A 15-day average of copper dosing prices, for instance, is the sum of 15 consecutive closes divided by 15.
A moving average is a progressive average in which the divisor (number of items) remains the same, but at regular intervals a new item is added at one end of the series.
To keep a five-day moving average of the closing price of December silver, you add up the last five closing prices and divide by five. That gives today's five-day average of closes. When tomorrow's close is available, compare it with the initial close of the nearest five-day series (the sixth counting back). If the new close is higher, add the difference to (or if lower, subtract the difference from) the previous total, then divide by five again. This gives tomorrow's five-day average of closes, and in sequence with today's five-day average of closes is the start of a five-day moving average of closes. Continue the process day by day.
Moving averages can be calculated for closing prices, for highs and lows, for volume, for open interest, or for any other factor that can be measured and which changes periodically. Various length moving averages may be employed to determine trends. A 10-week average of weekly closing prices, for example, can help measure the direction of major trends. A 10-day moving average of daily highs and daily lows can set up a tolerance band between the high and low averages, within which no action is taken. By buying only when the high average is exceeded on the upside and selling only when the low average is crossed on the downside, rather than by acting on every little crossing of the moving average, fewer whipsaw losses are incurred.
Some students believe that results shown by moving average can be improved by 1) employing an exponential moving average, that is, in which 1/x of the new figure is added to the previous x unit total and 1/x of the previous total is subtracted, or 2) employing a weighted moving average in which the more recent unit or units are multiplied by some factor that gives them a greater weight in the total than the earlier units. Others believe that it is helpful in plotting the moving averages on line charts to "push" the moving-average figure forward into the future by one or more days, weeks or other units so that when the next unit of market action takes place, guide points for penetration already will be available.
The chief value of moving averages as helpful tools in price analysis rests on the following simple premise: No commodity can ever stage an uptrend without first showing evidence of the preponderance of buying - over selling - by the price rising above a moving average. Likewise, no commodity can ever perform a downtrend without first showing evidence of more selling than buying - by the price falling below a moving average.
Rationale
The particular method submitted here is a simple one using unweighted five- and 20-day moving averages of closing prices. Signals to buy or sell are not given when the shorter average crosses the longer average but are dictated by precise rules outlined later.
Even though the five- and 20-day moving average method produced profitable results in markets that had good sustained moves, it shouldn't be regarded as any guaranteed or get-rich-quick panacea. In fact, the method has several apparent disadvantages.
Purchases, for example, are never made at or near bottoms, and sales are never made at or near tops. Action is only taken on the way up, after a bottom has been made, or on the way down, after a top.
Likewise, purchases often are made at strong spots where demand by others is great, and sales often are made when other sellers are competing actively for the relatively scarce bids. Consequently, orders are not always filled at "good" prices.
During periods that are characterized by narrow ranges, all trend-following methods tend to register small to moderate "whipsaw" losses. During protracted narrow fluctuating periods, a series of these small to moderate losses can add up to a fairly sizable cumulative loss. It is at such times that traders tend to stop following the methods often just before a pronounced money-making trend gets underway.
Despite these disadvantages, the method has some important advantages that tend to outweigh the drawbacks:
Diversification lowers potential but, more importantly, lessens loss risks. In futures trading, loss risks cannot be eliminated but can be controlled. Experience has demonstrated that a commodity trading method that pays strict attention to limiting losses can normally produce gratifying net annual gains.
Losses are always limited because positions must be closed out at definite points when the price fulfills certain requirements in relation to the moving average. The importance of definite loss-limiting rules cannot be overemphasized.
Profits are not limited. Whenever there is a lengthy, sustained move - and in most years there are many such important moves - the method automatically captures as profit a large slice out of the middle of the move. As long as the move continues without a valid crossing of the moving average in the opposite direction, positions are maintained. In this way, profits are unlimited.
Errors of judgment are greatly reduced. Because the method is almost entirely automatic, it virtually eliminates dependence on human market judgment. It is possible, and perhaps advisable, to use the method as an adjunct to fundamental factors and seasonal trends.
Indecision is greatly reduced. If you trade on a definite trend-following, loss-limiting method, you can do it without taking a great deal of time from your regular business. Because action is taken only when certain evidence is registered, you can spend a minute or two per commodity in the evening checking up on whether action-taking evidence is apparent. A definite method, which at all times includes precise criteria for closing out a trader's losing trades promptly, avoids emotionally unnerving indecision.
The Method
The rules for the five- and 20-day moving average method break down into two categories: general and supplemental.
General rules
1. The extent of penetration of the moving average is broken into units, depending on price level. For commodities selling over 400 (wheat, soybeans, silver), for example, a penetration of 40[cents] is required (Donchian had six price classes in the days before interest rates and stock index futures).
2. No closing penetration of the moving averages counts as a penetration at all unless it amounts to at least one full unit (39[cents] in Rule 1 was not enough for penetration - it had to be 40[cents] to count).
Basic Rule A: Act on all closes that cross the 20-day moving average by an amount exceeding by one full unit the maximum penetration in the same direction on any one day on a preceding occasion (no matter how long ago) when the close was on the same side of the moving average. For example, if the last time the closing price of cotton was above the moving average it stayed above for one or more days, and the maximum amount above on any one of the days was 64 points, then when the closing price of cotton moves above the moving average, after having been lower in the interim, a buy signal is given only if it closes above the average by more than 64 points (the unit in cotton is 0.10). This principle - the requirement that a penetration of the moving average exceeds one or more previous penetrations - is a feature of the five- and 20-day method that distinguishes it from other moving average methods.
Basic Rule B: Act on all closes that cross the 20-day moving average and close one full unit beyond (above or below, in the direction of the crossing) the previous 25 daily closes.
Basic Rule C: Within the first 20 days after the first day of a crossing that leads to an action signal, reverse on any close that crosses the 20-day moving average and closes one full unit beyond (above or below) the previous 15 daily closes.
Basic Rule D: Sensitive five-day moving average rules for closing out positions and for reinstating positions in the direction of the basic 20-day moving average trend are:
1. Close out positions when the commodity closes below the five-day moving average for long positions or above the five-day moving average for short positions by at least one full unit more than the greater of a) the previous penetration on the same side of the five-day moving average, or b) the maximum point of any previous penetration within the preceding 25 trading sessions. If the distance between the closing price and the 20-day moving average in the opposite direction to the Rule D close-out signal has been greater within the prior 15 days than the distance from the 20-day moving average in either direction within 60 previous sessions, do not act on Rule D close-out signals unless the penetration of the five-day average also exceeds by one unit the maximum distance both above and below the five-day average during the preceding 25 sessions.
2. After positions have been closed out by Rule D, reinstate positions in the direction of the basic trend a) when conditions in Rule D, point 1 above are fulfilled, b) if a new Rule A basic trend signal is given, or c) if new Rule B or Rule C signals in the direction of the basic trend are given by closing in new low or new high ground.
3. Penetrations of two units or less do not count as points to be exceeded by Rule D unless at least two consecutive closes were on the side of the penetration when the point to be exceeded was set up.
Supplementary General Rules
1. Action on all signals is deferred for one day except on Thursday and Friday, For example, if a basic buy signal is given for wheat at the close on Tuesday, action is taken at the opening on Thursday morning. The same one-day delay applies to Rule D close-out and reinstate signals.
2. For signals given at the close on Friday, action is taken at the opening on Monday.
3. For signals given at the close on Thursday (or the next to last trading day of the week), action is taken at the Friday (or weekend) close.
4. When there is a holiday in the middle of the week or a long weekend, signals given at the close of sessions prior to the holiday are treated as follows: a) for sell signals, use weekend rules; and b) for buy signals, defer action for one day, as is done on regular consecutive trading sessions.
A word of caution
The five- and 20-day moving average method, and most other trend-following methods, for that matter, are not good to follow unless you are prepared to include in your program a sufficient number of futures to provide broad diversification. Risks are increased to an inordinate degree if you try to follow the method in one or just a few selected contracts,
The commodities that are in a pronounced trend and are not giving, new signals are frequently the ones in which the best results are attained. Therefore, in starting a new program it might be advisable not to wait for new signals but to take positions in the direction of prevailing trends in those not giving new activation advice. Because the markets are moving so wildly, however, it might be best to a) go in the direction of the trend only after one or more days of counter-trend movement, plus a day move in the direction of the basic trend, and b) to use an arbitrary stop on positions taken without waiting for new signals.
Remember, five and 20 days are not necessarily the best lengths for moving averages. And, most probably, the action rules themselves, as outlined above, could be refined and improved. Also, it may be that exponential moving averages, weighted moving averages, moving averages based on highs or lows or daily means, or some combination of all these, would produce superior results.
In this field of technical study it is probably safe to state that the beginning of wisdom comes when you stop chasing rainbows and admit that no method is perfect. When you find yourself willing to settle for any comparatively simple method that in tests over a long period of time makes money on balance, then stick to the method devotedly, at least until you are sure you have discovered a better method.
Richard Donchian worked at Shearson Lehman Bros. while developing his technical analysis and trend-following methods that today many traders use as the base of their systems. He also launched the first managed futures fund in 1948. Donchian died in 1993 at the age of 87.

________________________________________

Citation Details
Title: Donchian's five- and 20-day moving averages.
Author: Richard Donchian
Publication: Futures (Cedar Falls, Iowa) (Magazine/Journal)
Date: November 15, 1995
Publisher: Oster Communications, Inc.
Volume: v24 Issue: n13 Page: p32(3)

Michael B

rschiller
Contributing Member
Contributing Member
Posts: 8
Joined: Wed Nov 03, 2004 9:02 pm

Michael B . . thank you!

Post by rschiller » Sat May 07, 2005 12:49 pm

Thank you for posting that article online!

Rick

rschiller
Contributing Member
Contributing Member
Posts: 8
Joined: Wed Nov 03, 2004 9:02 pm

Further question and think you on the Veritrader result

Post by rschiller » Sat May 07, 2005 3:16 pm

Garyboor, thanks for the Veritrader result, I do not have access to Veritrader. My original question concerned "one volatility measure" and when you tested the 5ma/20ma on Veritrader with the results you posted, was it a simple system using 5&20 or did it incorporate the ubiquitious "one volatility measure" which I stil do not have a definite answer for.

I asked Larry Williams about this, who knew Donchian personally. Larry alluded to the "one volatility measure" being a filter of 1% or so penetration of the 20MA. The 1995 Futures article also includes one and two unit rules. Did your Veritrader test utilize this or, again, just simple 5MA and 20MA crosses to enter and exit?

You are right on results for what you show, not spectacular. Roughly from 1980 until 1996-1997 the system returns about 17.5%, from there on to the present it is about flat.

Can you share any of the results and methodology you label as "astonishinly better results"?

thanks

Rick

Post Reply