Today I will tie up a couple loose ends regarding NP Study 2.
First I present overall trade statistics stratified by stop level:
Cells that show trends from low (tight) to high (loose) stop level are highlighted:
- The looser (tighter) the stop, the higher the winning (losing) percentage
- The looser the stop, the higher the average loss and standard deviation of losses (more extreme outliers)
- The looser the stop, the higher the overall standard deviation of trades (more extreme outliers)
- The looser the stop, the longer the average losing trade and overall trade (stops shorten what would otherwise be trades taken to expiration)
Average trade and profit factor both peak at the 4X stop level. I find this interesting because net return / maximum drawdown (MDD) peaks at the 2X stop level. To me, that was a pretty compelling case to use 2X.
Which statistics are more meaningful in this comparison brings us to the second loose end: my previous mention about Monte Carlo simulation. The trade statistics shown in the table above are based on 3,906 trades. The MDD is based on just one possible selection of trades. If I stick with the percentages and run multiple simulations with replacement then I can take averages of the MDD and other portfolio parameters. I think this would be very insightful.
I go back and forth as to whether the Monte Carlo simulation should be done with or without replacement. To give me the most confidence in trading the system, I want to select the method that is more likely to underestimate returns. In thinking back over the last 15+ years, we have seen some unlikely (multiple standard deviations) market crashes. Simulating with replacement could result in a higher or lower number of extreme losses. Simulating without replacement will result in the same number of losses in random order. This means the MDD will vary depending on the extent to which losing trades become grouped together.
I doubt any blanket answer to this question exists. I should probably do 10,000 simulations each way and see how they compare.
Time to brush up on MS Excel…Categories: Backtesting | Comments (0) | Permalink
I left off analyzing stop-loss data for naked put study 2.
In particular, I noted the importance of the net-return-to-max-drawdown (DD) ratio. In looking at the 2x vs. 5x stop levels, for example, the difference in net return is not that big at 27%. The difference in DDs is big and combining the two as ratios reveals a 2.11-fold difference.
By proportionally changing position size to normalize DDs, we can better see DD impact as risk-adjusted return:
The last column sums it up. Even the loosest stop (5x) led to a 180% performance improvement. The tightest stop led to improvement far better than that.
Graphically, backtested performance looks like this:
Here is the same graph normalized for maximum DD (risk-adjusted):
The worst DD occurred during the 2008 financial crisis. If you look closely at that part of the graph then you will see comparable losses.
Outperformance can result from trading a system with greater net returns. Outperformance can also be achieved by selecting a system with lower DDs. Lower DDs allow for larger position sizing, which also leads to greater net returns.
Just out of curiosity, I ran some numbers to see if max DD was proportional to stop levels. In other words, did the max DD at 3x exceed that at 2x by 50%? Was the max DD at 4x 20% less than the max DD at 5x? Six comparisons were done: 2x vs. 3x, 2x vs. 4x, 2x vs. 5x, 3x vs. 4x, 3x vs. 5x, and 4x vs. 5x. The 5x max DD was just barely less than 25% more than the 4x max DD. In the other five cases, though, the max DD was more than proportionally greater for the higher stop level.
I really have no meaningful conclusions to draw from these comparisons. I thought it might be a bonus benefit if the higher stop levels were accompanied by less-than-directly-proportional max DDs but this was not the case. These comparisons were based on one data point (max DD). What might be more useful would be to look at average loss (mark-to-sale multiple) for each stop level. This would represent a larger sample size.
Along these lines of small sample size, I also regard the equity curve comparison to be limited. While the graph looks impressive as summary data of a 15-year backtest, it is only one possible equity curve generated by this system. At the least, in other words, these equity curves represent a single ordering of trades. In the future I would be interested in doing a Monte Carlo simulation to study the distribution of net return and max DD values. Whether to do the simulation with or without replacement of daily equity changes is something for further consideration.Categories: Backtesting | Comments (0) | Permalink
What better way is there to break in the New Year than with new data? Today I continue analysis of naked put (NP) study 2 by studying the efficacy of stops.
I previously presented data comparing the naked put (NP) trade with long shares. This suggested outperformance by NPs. A drawdown analysis added complementary evidence. From there I went on to study the equity moving average as a potential filter for when to be in the trade. That was not encouraging.
Stop-losses are another potential means to improve the trade. Stops help avoid the largest losses at the cost of more frequent smaller ones. Stops result in trade exit, though, which shuts the window of opportunity for trade recovery (“whipsaw”). In theory, avoiding the most catastrophic of losses offers a better chance of long-term survival.
I therefore repeated NP Study 2 using end-of-day (EOD) stop-losses. Stop levels tested were multiples of the original credit. EOD implies the possibility of realizing a 4x loss even under a 2x stop condition (good future blog topic).
Let’s start with some vital statistics:
Return numbers are shown in the first two rows (not including header). The stops seemed to help. Net return without stops was the lowest. Looser stops (higher stop level) then resulted in greater returns.
Maximum drawdown (DD) is shown in the third row. No stops resulted in the largest max DD. Tighter stops resulted in the lowest max DD. This is precisely the reason stops are advertised.
Remember that DD helps to determine position sizing. I have written about this in terms of psychic pain. For these reasons, a system with lower DD is worth considering. I have a greater chance of sticking with such a system through tough times because I may not lose as much.
For these reasons, I included row 4 as a ratio of net return to maximum DD. It turned out the tighter (lower) the stop, the larger the ratio. Even the highest stop level has a ratio that is 2.8 times larger than using no stops at all.
I think this is huge.Categories: Backtesting | Comments (1) | Permalink
Chris Concannon is CEO of Bats Global Markets. Today I will conclude analysis of his article “Stock Splits for the Middle Class,” which was recently published in Modern Trader magazine.
Concannon suggests targeting “buy and hold” investors (for which he provided no evidence) is somewhat classist:
> While it’s nice for a company to say it
> wants to attract long-term buy and hold
> investors, failing to split a high-price
> stock encourages small investors to avoid
> a stock altogether or causes them to pay
> a substantial penalty when they trade
> such a security… One could argue that a
> company’s refusal to split its stock is
> a refusal to embrace middle class retail
> investors with less investable wealth.
Concannon qualifies himself nicely here by saying “one could argue.” The argument would be stronger if he could provide some evidence that higher-priced stocks are eschewed by a significant number of retail investors. Maybe it doesn’t make a meaningful difference given the dominance of institutional traders but I would at least like to see some effort to offer more than empty claims, which I consider speculative and meaningless.
Concannon closes with the following:
> When debating… the goal of making stock trade
> better, we should also focus on… stock splits.
> While this is self-serving to a degree, given
> our per-share revenue model, the evidence is
> clear that many investors would save money and
> more investors could participate.
This, along with many ideas in the article, makes logical sense. Unfortunately many logical ideas in finance do not bear fruit. I am baffled as to why he says the evidence is clear because he provided no evidence in the article. If there is clear evidence then please show us! As written these are nothing more than hollow claims.
What is clear from this final paragraph is Concannon’s agenda, which I believe helps to categorize this article. As the CEO of a major stock exchange, his corporate revenue is proportional to share volume. I consider this article a marketing piece to encourage greater use of the stock split. I also consider this an editorial since it includes no supporting evidence. A table showing number of stock splits per year seems like a no-brainer to include. I probably walk away more skeptical since such simple data is not presented.
I believe we should always read critically for the most complete understanding. Reading critically includes understanding the writer’s agenda. Writers aim to persuade, which is in their best interest. Being persuaded is not always in the best interest of the audience, however. Critical thinking is a very useful tool to defend against being persuaded by lower-quality [sometimes false] information.Categories: Financial Literacy | Comments (0) | Permalink
Chris Concannon is CEO of Bats Global Markets. He wrote an interesting article recently in Modern Trader magazine called “Stock Splits for the Middle Class.”
Concannon begins by presenting some data to suggest current stock trading is weak and depressed:
- Average share volume for the 12-month period ending April
30, 2016, is down ~9% from the same period four years ago
- Notional volume during this period is up 15%
- Share volume in January 2016 is up 1.92% while
notional volume is up 42% compared to January 2010
Weak/depressed trading is seen by a lack of significant share-volume growth. More notional value traded on equal or lower share volume suggests stock appreciation is outpacing stock splitting. I would have liked to see how these numbers compare to other historical periods for a stronger case, though.
Concannon believes the refusal of corporations to split stocks is one limiting factor on share volume:
> In an ill-conceived effort to attract “long-term
> investors” and detract “speculators” from trading
> their stocks, a number of popular U.S. companies
> have kept their nominal stock prices above $200,
> $500 or even $1,000 per share.
This is an interesting claim but no evidence is given to support it. If he is referencing a conversation(s) or quote by directors or C-level execs then he should most assuredly cite them! The use of quotation marks implies credibility but a critical reader cannot assume these to be actual words spoken by others.
Concannon goes on to explain the psychological impact of higher stock prices. He begins:
> An average individual investor may avoid a $500
> stock as being too richly priced but change his or
> her view when that security is split 10-for-1 and
> then trades at $50 [emphasis mine].
This makes sense but the actual impact is unspecified and I’m not sure it could really be measured. Concannon qualifies the statement well.
He then goes on to suggest the wider bid/ask spread is a significant drawback when trading more expensive stocks. He argues stock splits, while leading to higher per-share commissions, would more than make up for this by improved liquidity. In theory this makes sense but it is just that: theory. Why not interview some big institutional traders to see if they agree? This would be compelling evidence. Without any evidence it remains an empty claim.
I will conclude with my next post.Categories: Financial Literacy | Comments (1) | Permalink
> And then, take it one step further and pretend that
> you are not the investor, and instead are an advisor
> for a relatively unsophisticated (in the field of
> finance) client (or 200-300+ people as is usually
> the case). Which would you rather be responsible for
> advising a client to do… or, would you choose some
> other option that you have available? Keep in mind
> that advisors bear responsibility, not just legally,
> but personally (emotionally) for knowing that they
> may lose their clients’ retirement, education, or trust
Having so many clients is no excuse for using sub-optimal investment methods—methods that any individual could execute for him/herself given a healthy dose of education.
With regard to options, I have heard advisers complain about the excessive time/cost it would take to get regulatory clearance. Perhaps regulators are somewhat behind the curve but I’m not convinced. Some investment advisers do employ option strategies for their clients. Given my belief that many advisers don’t know options I wonder if those complaining have even tried? And if the compliance firm is to blame then find one that is option-friendly!
> money. And, remember that they have to implement their
> investment strategy for not just one or two families, but
> hundreds. Given these conditions, do you think advisors
> will be overly risk-averse (compared to the purely
> logical choice they might otherwise make based on
> economics) in their decision making? Statistical inference
> matters less to investors/clients who have lost any
> percentage of wealth, especially if they don’t fully
> understand why the risks were taken/decisions made.
To me, the omission of statistical significance is like off-label drug usage. Physicians occasionally prescribe medication for reasons not mentioned in the package labeling. If this is not standard of practice and if peer-reviewed data do not support said off-label indication then the physician could be susceptible to legal action should adverse events occur.
Given that evidence-based medicine has done well to advance medical practice in this country, why aren’t financial advisers held to the same standard? Do good statistical research that can be authenticated and replicated and use those methods to manage money for the public. Without this, I struggle to view the financial industry as delivering anything more than quackery to unsuspecting retail clients.Categories: Financial Literacy | Comments (0) | Permalink