.
If you are on the blog page in a web browser from a computer, please subscribe to this using the "Follow by Email" link to the left. If you're on a mobile device you should see something in the frame that allows you to subscribe. Having your email helps me to notify you when Google mucks up email distribution.
~~~~~~~~~~
Administratia:
Please read Forward Testing blog entry regarding my Twitter CSP alerts. The link is here:
https://fwdtest.blogspot.com/2018/09/csp-updates-for-friday-sep-28.html
~~~~~~~~~
The alerts files (monthly historical as well as the alerts generated as of the last trading day, after market open), are available here (
https://goo.gl/WbuJhS ). The archive subfolder contains historical alerts files that you can review.
Real-time Q&A with me, if I'm available, is through this link:
https://discord.gg/4QAUqyd This is Dr. Jeff Scott's HGSI Discord forum and it's worth your time to join (free). I am @PaulDuncan at Discord and I typically watch the #cashsecuredputs-n-coveredcalls channel. Come say "hello"!
~~~~~~~~~~
Fisher Transform
The more I dive into the Fisher Transform ("FXfrm") the more I'm impressed with it's performance as an indicator to reveal turning points. If you are not familiar with the Fisher Transform, the equity-applicability of the transform was demonstrated by John Ehlers and you can read about it here:
https://www.mesasoftware.com/papers/UsingTheFisherTransform.pdf
The paper has some math, relies on knowledge of transfer and filter functions (concepts in control theory and electrical engineering), and can be tedious if you are not educated in those areas (or at best, rusty). Luckily, I'm a practicing electrical engineer, so plowing through the details was a flashback to my college days ( decades ago ) and the mental gymnastics were not insurmountable.
The take away Dr. Ehlers wants us to believe is this: turning points in a price series, after you apply the Fisher Transform, are amplified and as a result, there is unambiguous signal capability to enter or exit a trade.
It is a different post that I need to write, but I can confirm that "yes, the Fisher Transform has an advantage
over moving average entries and exits", mostly because it has minimal to zero lag. Let's take that statement as a starting point.
~~~~~
The business problem that I'm attempting to solve is "How do I reduce the number of stock assignments when selling put options?"
When you sell a put option, if the price of the underlying equity is LOWER than the strike price of the put you sold at the close of markets on the Option Expiration (OE) day, the stock will be assigned to you at the strike price. So, if you sold the $50 put strike in XYZ, and the stock was $45 at the close of markets on OE day, then you will be assigned the stock at $50 and will have a $5/share paper loss (offset by any premium that you collected). To dig yourself out of the hole you need to:
- sell calls against the position (now that you own at least 100 shares), further reducing the position break even by collecting additional option premium and betting that the stock will recover and move higher than the call strike price, and remain there until the call OE;
- sell another put below your breakeven, and if assigned, it will lower the total position basis,
- BUY another lot (minimally 100 shares) of XYZ at the lower price, lowering the position break even
- Go back to step 1, rinse, repeat.
It should be obvious that if you are going to sell put options, then you want to be in stocks that are in an uptrend. Stocks that are in a downtrend will guarantee assignment and you'll be catching the proverbial "falling knife".
So, how do we identify stocks that are in an uptrend?
Enter the Fisher Transform.
The Fisher Transform has virtually zero lag and allows us to identify uptrends on any scale that we desire -- intraday, daily, weekly, monthly, quarterly, etc. It doesn't matter. What is important is that we can clearly see, using the Fisher Transform, whether the equity is in an uptrend or downtrend.
It's important to note that the Fisher Transform uses a past window of data, just like a moving average. Unlike a moving average, the window only tells the Fisher Transform what the current price is in regard to where it has been in recent history. So, if we have a 5-bar window, we're looking at evaluating today's price action with respect to the past 5 days, etc.
You get the idea.
Ehlers defaults his window at 10 bars for buying and selling and I'll state, without proof here, that this does appear to be the best window size for
daily data. I have TradeStation Portfolio Maestro, which is backtesting software, and have been doing weeks of testing of this configuration. So, if we are looking at applying the Fisher Transform to daily data, we will consider the last 10 days of data to make buying and selling decisions. Hence, the daily trend is defined based upon this 10-day window.
Remember though, the Fisher Transform applies to any bar setting: hourly, weekly, monthly -- it doesn't matter, as the indicator works well independent of the bar setting. What DOES NOT work well is the default 10-bar window setting applied to other time frames, a 10-week-bar window does not appear to be the best setting for considering the weekly trend.
Repeat the same statement for monthly. Or hourly. Same caution applies.
My testing suggests that *if* I am to consider multi-time frame analysis, and require that they both be confirming each other, then a 5-week bar, combined with the 10-day-bar, provide the best portfolio performance in terms of bars to review. There are all sorts of nuances and caveats in that prior statement so please ensure you read it again and understand it. Put another way: if I'm running TWO windows, the daily window will be 10 bars in length and the weekly window will be 5 bars in length.
Some combination of daily and weekly bar status (both in uptrend, both in downtrend, one in uptrend, one in downtrend) give the best performance for entering and exiting a long position.
As it turns out, the *BEST* performance, using the SP 500 over multiple 10-year periods, results when we:
- ignore the weekly status to enter a stock
- enter on a change in the daily status from downtrend to uptrend
- sell ANY TIME the weekly AND daily trends change from uptrend to a downtrend, OR
- sell ANY TIME the daily trend changes from an uptrend to a downtrend, independent of what the weekly is doing.
This is good, but it's not exactly intuitive. A few comments:
- I was thinking that the best time to enter a position was when both were confirming a new uptrend, so I tested for the daily changing to an uptrend and then waiting for the weekly to confirm. While this does work, it is not the "best" long-term performer. My only explanation is that the change in the shorter time frame from downtrend-to-uptrend catches the new trend in a quicker fashion, IF it materializes. If it doesn't, the rule 4 will catch the failure.
- No comment on this rule - it should be obvious
- This is a major sell signal -- when both are changing to a downtrend on the same day. This proves to be a very powerful indicator of a new downtrend and buying stock when both of these are pointing downward is not a great idea, especially if you are selling puts.
- This was a bit counter intuitive to me, as I wanted to stay in the trade as long as possible and wait for the weekly to confirm. Rule 4 states IF the daily fails, get out. Bank your profit / cut your loss and simply get out. The weekly may confirm, or it may not. In looking at the trading logs, this rule has a greater number of whipsaws, but in general, is the better long-term performer.
So, applying the Fisher Transform to stock entry on the long side, if given the SP 500 basket, you can make money over time. How much money? Backtesting suggests beating the market in a consistent manner, but there are many other considerations (position size, number of positions, scaled entry/exit, etc.) that need to be considered. That will be for another series of blog entries.
~~~~~~~~~~
Remember, the problem that I'm trying to solve is "How do I reduce the number of stock assignments when selling put options?"
So far, I've produced an indicator that is optimized over multiple 10 year periods that enters and exits long positions in equities based upon daily and weekly behavior, but this doesn't really help me for options.
Or does it?
I note when I look at the individual trades, the average winning trade length is 17 days, and the average losing trade length is 16 days. This is between 2-3 weeks of holding the position, and is within my historical "sweet spot" of selling put option premium.
The premise is this: if the trade is profitable, then we know it is above the entry price at the time of exit (close of market on the Friday), and if we sold puts when the signal indicated to go long in the stock, the settings that produce the greatest number of winning trades should provide an edge to my trading.
Note, I really don't care about magnitude of those trades, only that they were profitable relative to the entry date.
I modified the selling rules of the system to only sell on a Friday at the close. I also decided to move to just the SPY instead of all of the stocks of the SP 500, simply because processing 500 stocks across multiple 10-year periods, 10-days and 5-weeks at a time is really time consuming. While the use of the SPY is not perfect, it gets me in the ballpark (I was able to duplicate relative behavior as shown above using the SPY so it's a valid proxy). I'll apply to the SP 500 or DJ Composite to fine tune once I get the SPY results nailed.
After considerable testing, the weekly/daily "windows" are unchanged:
- 10-day window for daily bars,
- 5-week windows for weekly bars.
The "best" setups, not measured for portfolio returns but measured as a percentage of winning trades, are a bit different than stocks:
- weekly in a NEW downtrend, daily already in a downtrend, enter on transition to a new weekly downtrend
- daily in a NEW downtrend, weekly already in a downtrend, enter on a transition to a new daily downtrend
- SELL on the 3rd Friday after entry.
- The average number of trades, with the SPY, is 128 +/- 5 trades across multiple 10-year sliding windows.
- The average number of winning trades, again with the SPY, is 63% (about 81 trades), with 7 consecutive wins and averaging 17 days in duration.
- The average number of losing trades, again with the SPY, is 37% (about 47 trades), with 4 consecutive wins and averaging 16 days in duration.
Until further notice, I'm going to use these settings.
In addition to the above, I'll require the following of the stocks:
- the underlying stock is above its 200d, 150d, and 50d moving average
- the underlying stock is optionable,
- the 13-week average volume is at least 150,000 shares.
Other GGT criteria that I use, such as revenues, EPS, and free cash flow cannot be backtested, as that data is not available, but I will use GGT base criteria for going forward.
Stay tuned.
~~~~~~~~~~~
As with all my ramblings, you are responsible for your own investment/trading decisions and I am not. Please do your own diligence, and please take ownership for your actions. Please read and acknowledge the disclaimer that is listed on the left on the web site page.
You can reach me most times at the following site: https://discord.gg/4QAUqyd
Regards,
Paul
Posts
