Tharp's Thoughts Weekly Newsletter (View On-Line)
Introduction to Position Sizing™ Strategies E-Learning Course
$99 Price Offer Expires Next Week
This course is perfect for busy professionals who need a practical way to understand risk and how to keep losses to a minimum. It will open your eyes to the importance of position sizing strategies and how they can reduce your losses and increase your profits. This limited-time introductory price of only $99.00 ends July 30th!
Learn More or Buy Now
An Interview with 2011 Wagner Award Winner Thomas Krawinkel: Part 2
Thomas Krawinkel recently won the 2011 Wagner Award for his submission to NAAIM’s annual research paper competition. He has been studying Van Tharp’s material for several years and cited Van Tharp in the References section of his paper. We conducted this two-part interview with him after he returned from presenting his paper at the NAAIM annual conference. To view Part 1, click here.
What are your thoughts about backtesting?
First, I want to differentiate between backtesting and optimizing. I think many people tend to confuse the two. Many people seem to think that backtesting is optimizing their rules on a set of data. This is dangerous because you can optimize each variable and get fantastic results but all you have really done is to curve fit the rules to the data sample. You will not be able to generate the same kind of results with live trading.
To me, backtesting is merely testing an idea on data. I have a set of rules that I can run on historical data. What I like to do is to pick some trades at random out of a backtest and then look at the charts for those trades.
I collect all of the trade results and analyze the expectancy, draw time lines of key indicators, apply position sizing strategies. I try to learn why each trade worked or didn’t work. If I see enough of a pattern, I consider what I might change. I would never let a computer optimize or change the system for me.
My way of optimizing or changing the system comes from looking at a very small sample of trades compared to what people tend to do with testing platforms. Affordable computing power now allows people to use a platform to optimize a set of rules for all trades over a very long period of time. I think that’s a recipe for disaster.
Besides curve fitting, testing problems occur in large part due to data issues. I have learned to pay a lot of attention to the data I use for backtesting.
One general problem with long term testing is survivorship bias. Only the stocks that are still around show up as part of the backtesting results, which may make the results appear to be better than what the system might actually generate.
Then there are other data problems, simple things like how the data provider handles dividends. For example, Stockcharts.com subtracts dividend payments from the stock prices prior to the payment. Google, however, does not make this adjustment.
If you look at Stockcharts.com, it appears that the price on a specific date in the past was lower than where it really traded at the exchange or as reported by Google. The dividend adjustment affects the 52-week high prices, the moving averages, and multiple other indicators. It all depends on which data provider you use.
Many people do not know about these adjustments, but they should if they base their trading system decisions on price action.
Would there be a similar adjustment for one-time events, say a special five dollar dividend?
Certain data providers would reduce all historical prices by five dollars. In some cases, these adjustments can take historical prices below zero. Others data providers will adjust the historic prices by the percentage of the dividend against the stock price at the time of payment, which in my opinion is the best way to handle that situation. This way, the price never gets below zero.
And we are talking about dividends here. Imagine the adjustments the data providers make for stock splits or reverse stock splits.
Data issues like these become extremely important when you backtest over ten or more years because dividends, splits, etc. happen regularly and affect the stock price data. People seem to think that testing more stocks over more years gives you better test results. However, more years of data can actually skew your test results rather than make them more valid.
Then why backtest at all?
Backtesting gives me a general feeling whether an idea will work. That's what I use it for. Humans tend to look for support for our opinions and ignore evidence against them. I see that in Ken Long's chat room and at the poker table. I have heard people say “I don’t play Aces” even though they know that’s the best hand; however, they might believe they always lose when they play Aces. They may have lost just once in ten hands playing Aces, but they remember that one loss more than the other nine wins. Backtesting helps me see that playing an Ace is a good idea—regardless of what I have heard from others—and others definitely affect your beliefs.
You can read about or hear about a trading idea that applies to a particular set of stocks at a particular time but that opens the question as to whether it works on other stocks over other time frames. If you know other traders who talk about the same idea, that can seem to make the idea more "real." But it's not real, it's just a perception that can be distorted the more you see it, hear it, or think it. The question is whether an idea is valid. That's what testing is for—to see if an idea has an edge or if it is nothing extraordinary.
When I design system rules, I get an idea of how the system works for a particular market. But when I run those rules on a different market type or in different time periods, I expect this system to generate similar results. I want to test an idea for validity and an intrinsic edge—not simply for a good curve fit.
I prefer chart patterns as a basis for trading systems because I believe they represent psychological patterns. I contend that these kinds of systems should work on data today as well as on data from 10 years ago as I don't think human psychology has changed much in the last 10 years.
Price pattern based systems work because of psychology rather than other factors—at least that’s my personal belief.
I believe part of the purpose of back testing is to help me gain confidence in a system. I have seen other traders who can trade a new idea with real money right off. I don't like to do that. I like to test an idea first to see if it has worked before, rather than just in the current market.
When I was testing swing systems, I was surprised to see how widely their performance varied from year to year. One year the SQN score for a system might be 3.0, another year it was 5.0. Another year the system might have had a negative expectancy. I think it’s important to see the size of the swings. Looking at a single number over the backtest period as a whole, even the SQN score, won't tell you if you could actually trade the system and stomach the variation over different periods. My conclusion from this is that people who want to trade swing systems need to be prepared for long drawdowns.
Most people look at win rate, average R-multiple, accumulated earnings, and results like that. I think you've got to be able to take a large set of data and split that into different chunks and examine each chunk separately. How do aspects of the system change from chunk to chunk?
Did you look at the market types for these chunks?
I did; however, I didn’t find much correlation between system performance and market types. Although for example, a certain swing system might get a lot fewer signals in a bear market than a bull market. The way I was testing may have affected the correlation results but the differences were very low—in the range of statistical noise for me.
What’s your advice for others doing backtesting?
If you want to achieve the kind of live trading results you saw in your backtesting results, you will have to trade a system the exact same way it was run in the tests. Here are the questions you need to ask:
- Can it be done?
- Can I do it? Is it possible with my execution capabilities (i.e., automated vs. manual), my psychological setup and my buying power?
If you can't actually trade the system the way the test was run, revise the test to reflect the way you can actually trade it.
I would also add that if I'm sick or on holiday and not trading, when I return, I simulate the trades that I missed while I was out. I will enter those simulated trades in my trading log in order to help me generate a full sample for an accurate comparison to the backtested results.
Missing trades (anything more than even just 5% of the trades) can really skew your system results. I was very surprised at how big of an effect missing just a few trades can have on system results whether because of a cluster problem as I wrote about in the paper, you're on vacation, or whatever else causes you to miss those trades.
When I research trading systems now, I look for little clustering and short time horizons—I don’t try other system “adjustments.” For example, I've heard people say “Let me increase the expectunity of the system by trading a larger pool of instruments.” They think that they can go from trading 10 ETFs to trading all 500 companies in the S&P. Unfortunately for them, the larger pool increases the problem of clustering: they won’t have the buying power to take all of the system’s trades. All they do then is add randomness that may take their real life performance far from what they would expect to see by following a tested system.
I want to get back to the point about looking at your test and understanding what you can actually trade—you really have to focus on that. You may want to make $50,000 this year and the key performance indicators of a system derived from all trades in a test may suggest that you can do that. But you have to determine whether your buying power will cause you to skip a significant number of trades. I designed a Quick Check spreadsheet that is available here. Through it you may find that your “$50,000” system will only make you $10,000 given your personal circumstances. It would be better to know you have a $10,000 system before you quit your job or count on that trading income in any way.
Very true. Thank you very much Thomas.
You are welcome. And I’d like to thank the Van Tharp Institute. I wouldn’t be where I am right now without Van’s materials and workshops, so I’m happy to give back where I can.
You can download his paper from the NAAIM web site and you can download Thomas’ Buying Power tool by clicking here.
This Excel file will be available for downloading for a limited time only. You will need Microsoft Excel 2007 (or 2010) to use the tool. If you are asked for a password in the download process, just “x” the box to close that window and the file should still open.
The Buying Power tool spreadsheet helps traders understand the relationship between margin, 1R, and the position sizing strategy. It also provides a calculator to help identify limiting factors. Thomas provided this tool to the NAAIM evaluation committee in support of his paper. There are no instructions with the Buying Power spreadsheet tool, and you will need a good working knowledge of position sizing concepts in order to interpret its results.
You can contact Thomas by email - “TKrawinkel at GMX.de”
2011 Fall Workshop Schedule
Learn eight core areas of successful trading.
This is a great way to learn from in a small group.
This popular workshop usually sells out.
Register this week to ensure your seat.
Peak Performance 101
Advanced Peak Performance 202
$3,995 |Mechanical Swing and Day Trading Systems
||Discretionary Swing and Day Trading Systems
Click here to see our full workshop schedule with details
Click here to see locations, logistics, etc.
Navigating Sideways Markets
" All I can say is if Paul Tudor Jones, Louis Bacon, and John Paulson can’t make money in this market, I bet you can’t either. Better to watch in awe from the sidelines and until the dust settles and let others do the bleeding."
-- Hedge fund trader John Thomas
The quote above was buried deep in last week's article, yet it is so à propos to this market environment that I thought we should revisit it!
In the past several weeks, we've looked at the markets from several perspectives, but we can largely sum up the first half of 2011 in this way: sideways volatile action.
Every move up in most markets (e.g., equities, commodities, etc.) has been met with an equally big pullback and vice versa.
I've been on the phone over the past several weeks with a handful of frustrated traders who are getting chopped up by these market "pops and drops."
I'll share with you some of the counsel I've given these folks, but first, some encouragement: Don't beat yourself up! These market conditions are giving fits to the best traders in the world. Indeed, particular luminaries mentioned by John Thomas in today's quote are down double-digits for the first half of the year.
What You Can Do
The markets will eventually break out of this log jam. In the mean time, consider this:
- Long-term Trend Followers: Even though you (Paulson, et al) are getting chopped up right now, part of fishing for big trend moves involves consistently taking your stops when the markets move sideways. Stay patient and consider reducing your initial position size. Then scale into moves that gain some traction.
- Swing Traders: Are you seeing trades earn money only to watch profits disappear? I know traders who are taking quicker profits by waiting for the first little move against profitable positions and then quickly taking whatever profits they have. That will be the swing modus operandi until the markets become less twitchy.
- Day Traders: Intraday moves have become news dominated, and intraday reversals are becoming the norm. Tighter profit targets and quick stop-outs are the order of the day.
A Profound Experience
Client Feedback from Our Germany Workshops
Just a short note to express my gratitude for the recent courses in Berlin. I can’t overstate how profound the experience was for me or what an impact it is already having. When I returned home, I felt much more at ease with myself and my trading. Add the benefit from incorporating some of the course material into my daily routine and I can see a tangible improvement in my trading performance already. After only three days of trading, the courses have already paid for themselves several times over in terms of incremental (not just absolute) profits. Somehow that seems to be almost beside the point when compared to the bigger and more exciting prize of self-mastery and personal growth.
I’m very much looking forward to working with you further, and I’m excited about the journey ahead.
Everything that we do here at the Van Tharp Institute is focused around helping you improve as a trader and investor. Therefore, we love to get your feedback, both positive and negative!
Click here to take our quick, 6-question survey.
Also send comments or ask Van a question by using the form below.
Click Here for Feedback Form »
Back to Top
Email us firstname.lastname@example.org
The Van Tharp Institute does not support spamming in any way, shape or form. This is a subscription based newsletter.
To change your e-mail Address, click here
To stop your subscription look at the very bottom, left corner of this email and click on that link.
How are we doing? Give us your feedback! Click here to take our quick survey.
800-385-4486 * 919-466-0043 * Fax 919-466-0408
SQN® and the System Quality Number® are registered trademarks of the Van Tharp Institute
Back to Top