Perry Kaufman began his career as a “rocket scientist”, working on the navigation systems which were later using in the Apollo missions. In 1971 he became involved in the futures market and in that time has worked and consulted to a number of successful CTA, investment and prop trading groups, creating systematic trading and hedging programs.
Perry writes extensively on markets and strategies. His seminal book, Trading Systems and Methods + Website, now in its fifth Edition, has been called “remarkably insightful,” “the most authoritative and comprehensive work” in the industry; it puts the process of research and development into a “cohesive framework.” He has published fourteen books, some translated into Chinese, Russian, Italian, Spanish, and Japanese. He continues to lecture to economic forums, investor groups, and graduate students.
In this weeks chat we discuss market noise, the impact it has on trading styles and how it can be used to determine which strategies suit a particular market. We also talk about price shocks and how to mitigate their effects, how to use volatility in your favour, volatility parity for position sizing, the information ratio for strategy performance and some strategy ideas you can test yourself.
Topics discussed
- How market noise impacts trend following and mean reversion
- The effects of money flow during a crisis
- The types of strategies that work best in new markets and why
- How the efficiency ratio can be used to determine the best type of strategy for a market
- The best markets for trend following and mean reversion
- What strategy style to choose if you’re just starting out
- Using Volatility Parity for position sizing
- Impacts and dangers of price shocks on backtesting and how to handle them
- Mitigating the risk of price shocks
- Using the Information Ratio to measure strategy performance and detect possible over-fitting
- The effects of volatility on strategies and how to use volatility in your favour
- Fractal Geometry
- High Frequency Trading
- And some strategy ideas you can test
Resources mentioned in this episode
- To learn more about Perry and his work, checkout his websites Perry Kaufman and Kaufman Signals.
- Recommended books:
![]() |
![]() |
![]() |
![]() |
NOTE: You can get “Inside the Black Box” as a FREE audiobook here
Quotes
Top tips from this episode
- Noise in the market indicates the type of strategy that will work – trend following is suited to markets with low noise and mean reversion for high noise. So with the US markets being the most noisy you may want to consider Mean Reversion strategies and if you’re trading in a new market, which tends to trend more, then trend-following may be a better choice.
- Using the Information Ratio to measure strategy performance and detect possible over-fitting – the Information Ratio is calculated as Annualized Return/Annual Standard Deviation. Perry said a robust system should be able to get an Information Ratio between 1 and 1.5 but if it’s over 3 then you may have over-fit to historic data, so that’s a pretty simple measure to alert you that you may have curve-fit the system.
- “loose pants fit everyone”– If you have a system with very few parameters, it kind of has a sloppy performance but it works and gives you profits on most markets and then you reduce the risk by diversification. That is the correct outcome. If you try to hone in on the profits and eliminate the risks you’re just kidding yourself. You eliminate the risk in one place it pops up somewhere else.
Transcript
Got A Question, Topic or Guest you want to see on the Podcast?
Do you have a specific question, topic or guest you’d like to see on a future podcast episode?
Click here and submit your suggestion for a chance to have it featured on an upcoming podcast episode.
Subscribe to Better System Trader and never miss another episode!
![]() |
![]() |
![]() |
![]() |
Please support the podcast by giving an honest Rating/Review for the show on iTunes!
Episode Released:
7 June 2015
Hi Andrew
I have a question: You asked Perry about price shocks and how that related to backtesting. At about time 25:00, he states that backtests will pick up the price shocks and factor them in as positive, that is a profit for the system, whereas in real life most of those price shocks will be missed (he claims about 30% might be gotten by the system). I dont quite follow this. How/Why do the price shocks get included in the system backtest for a profit, but in real life get missed? I understand this adds risk to the system as the trader would trade the system based on the backtest results which showed profit from the price shocks when they wouldnt really (rather be losses), so if most of the system edge came from those and in real life it missed them then the system would have no edge. However, I’m a bit lost on the how & why price shocks get included in the backtest and not real life?
Keep up the excellent work on these podcasts!
Shawn
Hi Shawn, I raised this question with Perry, here is his response:
It’s a good question and I wasn’t clear enough in the podcast. I’ll attribute that to stream of consciousness, without time for me to consider all the issues.
Shawn seems to understand the problem, so I won’t trivialize this. Price shocks become part of the data history just like any other prices. When we select which parameters we use based on testing, we favor the best results. Those best results would have profited from large unexpected moves (shocks) or at least not hurt as much by them compared to other tests. I can’t tell you the percentage of price shocks that were favorable because that depends on the strategy. We never isolate past price shocks in our testing and say “this was a random event.” So, if you identify all the price shocks using some volatility measure, then check how many of them produced a profit in your test results, you’ll know exactly whether your parameter selection is unrealistic, or at least the returns from that selection are unrealistic.
So I expect the choice of parameters based on back testing will be the ones that reflect the most favorable price shocks. Some systems are not as sensitive to this, so I’m generalizing.
Then we get to real trading. Our positions often reflect the general direction of the market, especially if we’re trend followers. If we are trend followers, then a price shock is nearly always in the opposite direction to our position, because there is a void in the liquidity. Based on that premise, the difference between the number of favorable price shocks in your test, and the actual price shocks that were profitable in reality, will be your error in expectations of profits.
Are the profits from price shocks in the past larger than the total profits in the test? No, but they are larger than they seem because they would be more likely a loss in the future rather than a profit, so some of those past results would be reversed.
I’m not talking about a 2008 where prices turned and went down for weeks. I’m only discussing a single day shock, the result of some unexpected unemployment report or geopolitical even. It could be a loss of 2 to 5% in one day. The biggest issue with accepting test results with shocks will not be the returns, but the risk. When you assume a profit when you might have had a loss, you believe that the risk is much lower than it really is, and you are not prepared for the actual trading risk.
I suggest that you verify this yourself. Mark the days that are “price shocks” as larger than 2.5 ATRs compared to an average ATR. Then add up the profits and losses from those. If there are more than 50% in your favor, you are assuming that you can predict a price shock, or that a trend system is able to predict a price shock. I don’t believe that’s true.
Best regards,
Perry
Thanks for the reply, both Andrew and Perry.
It is alot clearer now thanks.
It is definitely something that you’d need to test with each system. I like the testing method Perry describes in his last paragraph there. I wonder, though, if this kind of thing would negatively affect longer term trading systems like trend trading systems. One day price shocks, even fairly large ones like 2.5SD moves, probably wouldnt be enough to trigger an exit on such a long term system. On a shorter term system, especially intraday systems, I think it would probably hurt and add risk. Not so sure about a longer term one though. That’s just my guess though, and my guesses have a bad habit of being wrong. Anyway, something to test.
Thanks
Shawn
Hi Andrew,
Allow me to thank you for the wonderful work that you and Better System Trader are doing. I am completely overwhelmed by the information and material you have on your website and the same has empowered me to do the required things in order to keep my head above water.
This was a specific comment related to this particular PodCast. Sir Perry Kaufman mentions that he is a fan of Sir John Ehler (PodCast time: 00.52 hr) and recommends his books and indicators. However, you have not mentioned the same under your “Resources & Links mentioned in this episode” section. I would kindly request you to take the same into consideration.
Thanks,
Neha
Hi Neha,
Thanks for the kind comments, I’m glad you’re finding value in the podcasts!
Good point about the Ehler book, I must have missed it. I’ve added it to the page now.
All the best,
Andrew.