In this episode we’re discussing the results of a quantitative study on stop losses completed by Cesar Alvarez of Alvarez Quant Trading. Cesar was also a guest of the show way back in Episode 3.
Cesar was director of research for Connors Research for almost 9 years, developing quantitative trading models for individuals, prop traders and hedge funds.
In this episode he’s going to share the results of a quantitative study on stop losses, also testing out some common pieces of trading advice to see if they’re actually true.
Stops can have such a huge impact on trading results so I’m sure traders of all levels will find this research invaluable.
We will be discussing backtesting results and some charts. We’ll do our best to explain them for those listening along but if you’d also like to see the results while we discuss them, you can download a copy or even watch as a video in the sections below.
I hope you enjoy Cesars discussion of ‘Stops – the Good, the Bad and the Ugly’.
In this episode we discuss
- Different types of stop losses, their application and performance results
- Percentage vs Volatility based stop losses
- Intraday vs End of Day stop losses
- Trailing stops vs Targets
- Some common trading statements that are often assumed to be true and the results of testing them – do they hold up?
- The levels of Stop knowledge, which level are you at?
Show your support
While you’re listening, it would be great if you could show your support by leaving a brief review of the podcast in iTunes. This really helps to spread the word and reach even more listeners!
Resources mentioned in this episode
- Cesar can be followed on his website alvarezquanttrading.com
- Here is an article Cesar published called ‘Maximum Loss Stops: Do you really need them?‘
Watch the Video
Download the supporting files
Get a free copy of the slides and the excel sheet of the test results:
Quotes
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:
27 December 2015
047 – Backtesting & Execution – Nitesh Khandelwal
035 – Trading the VIX & using Fundamental data in trading models – Andrew Gibbs
031 – Trend Following & Market Breadth – Greg Morris
122 – Competing with Machine Learning and PHDs
088 – Protect and Grow Capital during Market Corrections with Ivanhoff
Jay Daly
December 29, 2015 at 11:57 amAndrew,
Loved the podcast with Cesar Alvarez on testing stops. Quick question, with max loss end of day ATR stops, can you confirm that the stop was adjusted down every day of the holding period based on 2.5 times the ATR as of end of day? Just want to confirm that the stop would keep going down every day if one was in a drawdown.
Thanks,
Jay Daly
Andrew Swanscott
December 29, 2015 at 12:04 pmHi Jay,
Glad you enjoyed it.
No, the max loss ATR stops are not adjusted every day, they are calculated on the date of entry and fixed until the position is closed.
Apologies if that was unclear but I hope the explanation helps.
All the best for 2016,
Andrew.
Julius
December 30, 2015 at 12:19 amAndrew,
Nice work on the podcast! I’ve been following for a few months and you’ve done a great job with it.
I was unclear on the difference in Cesar’s stop loss discussion between a max loss ATR and a trailing ATR. The spreadsheet does not seem to include an example of how these are calculated. I understand ATR, but I thought it was always calculated as a trailing x period. Maybe an example calculation of the two in the show notes would be helpful.
Thanks and best wishes for 2016,
Julius
Andrew Swanscott
December 30, 2015 at 9:46 amHey Julius,
Glad you enjoyed it!
The difference between the two stops is the Max Loss ATR is calculated on entry only and is fixed until the trade is closed, it doesn’t adjust over time, however the trailing ATR is recalculated every day and only ever moves up (for a long position) so it trails prices as it goes up and if price drops the position is closed.
Hope that helps.
Regards,
Andrew.
Ken Jarrad
December 30, 2015 at 7:45 amOne of the requirements per stock is “Close is greater than $10”.
I wonder whether this filter was applied to the ‘actual price’ or the ‘adjusted price’. The price tested at a particular date would ideally be independent of subsequent corporate actions (e.g. stock splits).
Does PremiumData provide both prices? (Yahoo! Finance has both prices for daily close.)
Also, do the historical prices include cash dividends? On the Premium Data website it says they do not include cash dividends as standard but there is an option to include them. (Yahoo! Finance adjusted prices do include cash dividends.)
[The findings of the research would not be materially altered by these factors but it is necessary to clarify them to reproduce identical results and extend the research.]
Cesar Alvarez
January 3, 2016 at 3:08 amKen, I applied the actual price that existed that day before splits and dividend adjustments. Premium Data has both prices. The historical data is also adjusted for dividends.
Ken Jarrad
January 3, 2016 at 8:01 amCaesar, thank you for the clarification.
What you have is not Premium Data’s standard product. I contacted them and they replied:
Access to the alpha testing program for the new updating platform we are developing (“PDU”) is only open to existing subscribers upon request. No date has been set for its commercial release.
Our existing platform (and the historical data available through it) does not offer any of the key features you are looking for.
The PDU platform offers all of them.
At the moment, access to the data updated by PDU is only available via a plug-in into the AmiBroker program.
The historical data cannot be purchased on its own.
If you are an AmiBroker user, and think you might be interested in becoming an alpha tester for PDU, you need to have subscribed to our existing update service for US Stocks and have purchased the existing historical data for both “currently-listed” and “delisted”.
Matt
January 3, 2016 at 9:21 amThat’s what PremiumData told me 1 year ago.
“We only intend to provide the capability for users to test on historical index constituents within a supported analysis package.”
Looks long they are going to be in a LONG alpha phase!
Andrew Swanscott
January 3, 2016 at 9:38 amI’ve been using the Alpha PDU package for more than a year now and it was running for a while before I joined. It’s a good package though!
As far as I’m aware it’s still open for anyone to join, all you need to do is subscribe to the relevant data then send them an email and they will switch you over.
You do need Amibroker though.
Cheers,
Andrew.
Glenn
December 30, 2015 at 1:40 pmHi Andrew,
Once again a highly professional podcast. To me the stop loss question is now a big part of my trading plan puzzle. I’ve recently adopted the Mark Douglas technique (Adam Grimes does as well), with my own variation, of scaling out 1/3 at the first R profit target and another 1/3 at the 2nd R profit target and letting the remainder run until a 2.5ATR trailing stop is hit. Do you know of any testing that has quantified this method (do you think Cesar has explored this).
It seems that I may be shooting myself in the foot with a trailing 2.5ATR. I’m a swing trader looking to stay in a trade for 2 days to two months so Cesar’s results may have no bearing on swing trading per se.
All the best Glenn
Andrew Swanscott
January 1, 2016 at 2:05 pmHi Glenn, glad you enjoyed the podcast.
I’ve not actually tested scaling in/out of positions myself and I’m not aware of any studies that have, although as you’ve mentioned the results can vary depending on the actual system so it’s probably best to test it out on your own strategies to determine the impact it is having on results.
All the best for 2016.
Andrew.
Ken Jarrad
January 3, 2016 at 8:07 amThe research compares exit triggers using prices that are either “intra day” or “end of day”. The finding is that this is a major factor affecting the profitability of the strategies.
I deliberately used the term “trigger” because the trade execution might not be at the price that triggers the exit. I think Caesar said that the EOD trigger is executed at the next market open.
What is the trade execution price for the intra-day trigger? Is it also the price at the next market open? Or is it the trade executed at the trigger price? i.e. the intra-day stop price.
If the two triggers result in different trade execution prices then the findings of the research are undermined. This is because a price that stops the trade is probably quite extreme thus subsequent prices are expected to be higher (i.e. revert to the mean). The price at the next market open is expected to be higher, thus explaining the improved performance.
[See http://li209-143.members.linode.com/018-scott-andrews for a discussion of opening gaps.]
A fair comparison would use the end of day price as the trade price (for end of day triggers). In practice this can be approximated by placing a stop order that becomes effective in the last minute of the market session.
Cesar Alvarez
January 3, 2016 at 10:05 amKen, the EOD stop is evaluated at the close using the closing price and then if it is beyond our target, then the exit is executed at the next open. For the intraday stop, the execution price is the stop price (unless there was a large gap beyond the price, then it is the open price).
I don’t understand why using these two types of stops ‘undermine’ the research. These are the two most common type of stops that I see people use and write about.
I actually fine EOD entries/exits based on some specific price to be unrealistic. It works fine most of the time but they will be times that you get in when you should not have and did not get in when you should of.
Cesar
Cameron Moore
January 8, 2016 at 9:29 pmHi Andrew,
Well done on another great podcast.
I have a question about the method used to determine the R based metrics particularly for the baseline case where no stops, other than a time stop, was used. As I understand it, R is the initial risk for for a trade. Given that there were no stops used, what was the amount assumed for the initial risk? Was something like the average loss used or was the assumption that 100% of capital would be risked for each trade?
Many Thanks,
Cameron.
Cesar Alvarez
January 9, 2016 at 6:37 amCameron, I used avg % loss as the amount risked for the base case with no stops. This is what Dr. Van Tharp suggests when you don’t have a stop.
Cesar