Some readers have posed good questions about the Power Dip, in terms of how the system would perform during various alternative external conditions. For example, what would happen if gains were not compounded, or we bought one day late, or trade size was reduced as not to move stocks without enough liquidity to handle the required position-size. This post compiles a summary of system results during various not-optimal external conditions.
Keep in mind that the factors in the Power Dip that determine when stocks are bought or sold have not been optimized to any extent, except in the case of one factor, which was loosened to provide more opportunity.
However, we should not ignore what I consider to be “external conditions” as they too may be optimized. In many ways it is harder to curve-fit external conditions as the trader may have little choice as to how much slippage occurs, or whether or not he is forced to trade only at certain times of the day. What the results seem to suggest is that external conditions are primarily measuring exposure, be it due to increased position-sizing (compounding) or more/less time in the trade.
One reader suggested that I discard the top x% of trades and then report the system returns. That would take some extra work in excel, and I think the profit distribution graph below shows that gains are not coming from a few large trades. I therefore did not run that specific test.
Summary
The results are of course predicated upon the risk per trade and stop level used.
As I mentioned, the primary difference in most of the external conditions is the adjustment in exposure levels as well as amount of opportunity. Keeping with the line of thinking that returns are affected by exposure, we can get out-sized returns similar to the returns provided by not using a stop. This is accomplished rather simply, by adjusting our risk to 2% per trade. And, we can keep using a 10% stop.
In the end, monitoring slippage, monitoring the overnight vs. daytime risk premium, and making sure to capture every opportunity are the most important considerations when trying to replicate historical, backtested results with real-time trading results.
Wood,
Being outside the US, it is difficult for me to understand the chances for the introduction of a financial transaction tax. But if it gets introduced, especially with the lately heard levels (0.25% of each side of a trade) that could be a serious blow to your system – and most other systems with short term trades. While Power Dip may still stay profitable, CAGR would be seriously hurt.
What do you think the likelihood for such a tax is?
Would you be able adjust your system by narrowing down the number of trades to the ones most likely to produce a higher average gain?
I guess a tax like that could also considerably impact short term market patterns, possibly increasing volatility and creating new opportunities for Power Dip ..(?!). Just some (hopefully philosophical) questions…
Andras, excellent question.
A tax such as the .25% Tobin Tax would significantly impact a strategy like the Power Dip.
Here is what I am hearing about the tax. i. That it won’t pass. ii. That it won’t be applied to retail traders.
You are correct that it would impact short term patterns, and there will be new opportunities. I could work on narrowing the criteria to the one most likely to achieve a higher average gain.
Great stuff – thanks for running my request!
It certainly seems like no stops is the way to go, with more wins and higher returns with only a slightly higher max drawdown. Is there some problem with it that I’m not seeing?
Without stops, are there times when the system would stay in a losing trade for a substantially longer time, or does it remain pretty close to the 5.5 day average?
Michael,
Running this (or rather, any) system without stops would probably require a high degree of intestinal fortitude. Look at the large red bar on the graph at the end – those are positions that have stopped out at the 10% stop. With no stop in place, chances are a significant number of those positions move further into the red before the system sells. Based upon prior posts by Wood, running no stops has a significant opportunity cost associated with it. The no stop system made 5632 trades, which is the lowest number of trades among any of the situations presented (and the regular system, IIRC) and in turn had the lowest opportunity. The no stop system was probably stuck waiting for the 850 or so losers to trigger an exit and missing opening new positions where its ~70% win rate and positive expectancy could make some $$$.
@Wood: Nice work!
Michael, largest loss with no stop was ~-57%.
I don’t have a lot of time right now, so I’ll just jot down a quick thought, but the only way I’d trade this with no stops is if I was using 25% of my total capital, and then using 1% risk/trade. That would look like this, assuming 100K total equity.
25K allocated to the system-
250 risked per trade (1% risk)
10 max positions
25K/10 = 2,500
So one could theoretically lose his entire investment, 2,500, if there was an adverse event. What would be devastating would be several adverse events, which could result in the loss of 7.5% or more of one’s total capital.
Again, best way to increase returns is to double risk per trade, and keep using stops.
I believe at 2% risk the system does near 70% CAGR.
Ouch though, even granting a low fee/cost account (let’s say $4 a trade), that would put you 1.6% in the whole before you even got started.
Or is that already accounted for?
I’m not sure what you are referring to Mr. Thaler. Are you talking about the tax?
No transaction cost per unit purchased in $250 baskets.
Kill the Banks,
Thanks for the reply.
That makes sense, but even having made fewer trades, the CAGR of the no-stop version was about 17% higher, so wouldn’t that already account for the opportunity cost?
Very impressed with PDS results, since using it.
Thank you Dr. Fly.
Kill the Banks, Michael,
You can always use “time stops” instead of “price stops”. Time stops will make sure you do not tie up capital in bad positions. The problem with price stops in case of a mean reversal system like the Power Dip is that they often activate in an extremely stretched position, just before a major bounce takes place. I suppose Woodshedder’s “No Stop” strategy actually has a time stop in place, but I may be wrong.
I’ll run time stops tonight on the version with no price stops, and update the results in the comments section here. My problem with time stops is that they can be easily curve fit. Still, not a bad idea, and probably preferable to no stops.
A 10 day exit worked the best, but it did nothing to protect you from a sudden loss.
CAGR: 61.94
Max % Drawdown: -15.81%
Profit Factor of 1.93
Sharpe Ratio of 1.78
Honestly, pretty much any time exit longer than 3 days worked well. I tested up to 20 days. That is pretty robust if you ask me.
A 4 day time exit has a CAGR greater than 50%, but the drawdown is a little larger at 21.91%
I tested time stops on this system almost a year ago. I think after the launch I’ll see if I can’t include that option, for subscribers who might be interested in such an option. It looks like a time stop might be robust. The tradeoff will be to see if it actually increases opportunity without a decrease in the average trade.
Just another thought: as a new system, you could take positions in those stocks which get stopped out of Power Dip, and sell x days later or sell at a certain profit target. There is a good chance the average gain per trade would be significantly higher than within PD. A drawback of this sytem, however, would be the reduced number of trades and possibly increased volatility.
Another interesting idea.
Excellent and interesting work, Wood. I think that you are overstating the commissions cost at .01, especially for a guy like me with a large account and understating them for a small account. But you probably already know this. My broker is Siebert, and I usually buy and sell 5,000 share lots which cost $14.95 which comes out to .003 a share (and I guess a buck or two for exchange fees?). I would think that over 5000-6000 trades, this would greatly increase the system profit?
Siebert. Wow, a blast from the past.
Siebert has pros and cons. The commission rate for large accounts is good. As I said, .003 cents per share @ 5,000 share trade. Hard to beat that.
The margin rate (I’m paying 2% right now) is good, but you have to negotiate this rate if you have a big account (over a million $ in margin debt). I think Interactive brokers margin rate is even less.
Money market rates are comparably good, Siebert uses Fidelity’s MM accounts and you can even get some Fidelity MM funds not listed on the Siebert site if you ask.
On the con side, Siebert trading platform is bad, and the tools are bad, (and their old website is actually better than their new one! I still use the old one, they let you) , but my IRA account is at Etrade, so I use their Power Etrade Pro platform for charts and quotes and tools. It gets the job done.
I have 7 diffferent accounts at Siebert, so the paperwork nightmare involved, keeps me from switching to another broker. Siebert used to be top ranked, but now they have fallen far down the list. For my particular situation though, they are still one of the best, because of the reasons stated above.
Thanks Dave. A Tradestation commission structure would be about .0074/share, or at least that is what I am running.
Interactive Brokers would be .005/share.
I used .01 to try and find a reasonable medium.
I’m sure this question has been answered before but what’s the minimum allocation of capital you recommend for this system to produce decent results. Assume IB’s commision structure.
I highly recommend a FAQ
I have an FAQ already written on the subscription site, and an entry for commissions…but not for minimum allocation. I’ll add it…thanks.
Well, you probably do not want to mess with pattern daytrading rule, so I’d say 25K.
Also, I would not recommend anyone putting 100% of their capital into one strategy.
Another consideration is that with NYSE issues, anything less than 100 shares is an oddlot and you will not get as good pricing. Having enough capital to avoid oddlots most of the time would be a bonus.
Thanks
I look forward to hearing what the subscription pricing looks like and an official launch date
Makes sense – thanks for the replies.
So to make sure I understand you correctly regarding the 2% vs 1% risk, are you saying that the ideal CAGR is attained with a stop of 20%?
Mike, keep the stop the same. (I will go into more detail soon about the relationship between the stop and percent risk).
So 1% risk, 10% stop = 1K risked on a 10K position
So 2% risk, 10% stop = 2K risked on a 20K position
What will change is the number of positions you hold. If you hold 10 max positions with 1% risk, you will hold 5 max positions with 2% risk. That is because your 2% risk positions will be twice as big as your 1% risk.