Forum Replies Created
- Posts
Thank you. I would like to hear the rationale behind removing MYRS from the top 3 strategy. I have found trading strategies like Nasdaq 100 and Dow 30 – that hold individual stocks – to be less than ideal – leads to too many transactions, slippage, susceptibility to risks of individual corporate actions, de-listing, mergers etc. These strategies look great on paper, but it is difficult to match the returns as a small independent trader.
Do you have any suggestions for someone who uses the top 3 strategy but doesn’t want to trade individual names? Because my guess is that going forward Nasdaq 100 and Dow 30 will feature heavily in top 3.
I would also be interested in a reply to these posts. I like the idea of trading a “strategy of strategies” or “top 3”, but the allocation history on these strategies seem to change more frequently and drastically when the software or data is updated. These small changes are significant because they can cause a domino effect that impacts a full third (or 2/3) of the portfolio for a past month. For example on QT, the strategy of strategies shows it was invested in World top 4 in August, but this was not the case. At some point in September I believe the allocation in the past changed from BRS to World Top 4. This kind of change has a significant effect on the reliability of the long term statistics on these strategies. The theory behind the strategy is sound but its probably good to understand the limitations of the backtests and keep one’s expectations in check.
Certain strategies and portfolios create a higher percentage of short term gains – for example NAS 100, DOW 30, and top 3 – because they buy and sell more often and switch between strategies and funds frequently. But in theory if you created a static portfolio of strategies like UIS, MYRS, BRS, GLD, you would more or less be re balancing the same group of holdings every month allowing for a bit higher percentage long term gains. Although in certain more volatile years like 2018, this static portfolio mix still creates a lot of short term gains (and losses).
My portfolio YTD is -5.4% I was in MYRS and UIS3x in Jan. and Feb., so the poor performance is to be expected. Beginning in March I moved to LI strategy of strategies. However understanding the allocations for it and getting used to using QT has been confusing. Day to day on QT the 3 strategies it chooses and the hedges it uses change quite a bit. The consolidated allocations function on QT either isn’t reliable or I’m not understanding it correctly. I always compare allocations to my manual calculations using the emailed signals at the end of the month just to be sure then rebalance about 2-3 days into the new month. I think some of this and just overall increase in market volatility can explain discrepancies between backtest results and actual results. I also short TMV and so far this year going long TMF would have been the better choice based on borrow costs and how those funds have moved month to month. Someone smarter than me could explain why that is.
In 2017 I outperformed the backtest portfolio on the online portfolio builder by about 7%, in Jan.-Feb. 2018 I underperformed it by about 2%. The same thing happened to me in Aug-Sept 2015, the days I chose to execute the signals caused me to take a much larger hit than the backtests indicated. But I’ve been trading long enough and enough strategies to know that backtests really just give you a general picture of a strategy, you shouldn’t expect it to be exact. That being said I do think it is helpful to drill down into the detail of your trades to understand how the discrepancy occurred. Doing that and learning from my experience in Aug. 2015 really helped me in February.
I just downloaded the latest QuantTrader with updated leveraged strategies that now use UGLD. The 3x gold ETN only goes back to 2012 or so. Has anyone found or generated synthetic data to extend the backtests to 2004? Thanks.
Alexander I read your comment about ZIV in 2007/2008 losing 83% and was curious so I downloaded the synthetic data you linked. What I wanted to know was the largest loss in a rolling 2 week period (since that’s the re-balance period for MYRS). As you can guess the last 2 weeks of Oct. 2008 ZIV lost 27% according to the synthetic data. However the 3 preceding 2 week periods it lost about 7%, 5.5%, and 15%, so it was a large drop but somewhat gradual. My understanding is that if this strategy works the way its designed you might take a lot of those first 3 losses in September/October, but when it drops 27% the last 2 weeks of Oct. you would have a very small allocation to ZIV. While this would be difficult to trade through it isn’t anywhere near the same as losing 83% of course. But here is the interesting part for me at least. I was trading MYRS the end of August 2015 and apparently the rolling 2 week loss on ZIV was slightly worse than Oct. 2008 at about 28-30%. ZIV was at 70% allocation at the time because the drop was actually more sudden than in 2008 (the preceding periods were positive). MYRS got whipsawed and took some time to recover from the 2015 shock, but I’m happy I was able to stick with the strategy. It’s not so much a problem if ZIV has large losses, but matters more how suddenly they occur. Just thought this might be some helpful context for your comment. Would be interested in your thoughts. Thanks. -Derrick
You are correct Ivan, thanks for explaining that. I did not state that correctly. You are borrowing the shares, not cash. I guess the cash in your account increases by the amount you are short because you are essentially cancelling out your long positions. (a 50K account with 20K in long positions, 10K in short positions will have 40K in cash). I assume this means you can now use that cash without having to use margin. This would be relevant if you trade a leveraged portfolio or something like BUG leveraged because you won’t need to borrow as much to get that leverage. Is that correct?
I use IB. My understanding is that when you short you are essentially borrowing the cash to be able to cover the position. This is why it increases the cash position in your portfolio. You are then paying interest on that borrowed cash while short. Whether you already have enough cash is irrelevant, you always must borrow when shorting. Currently TMV is about 3.8% to borrow I believe. This is a good deal to go short TMV because the edge over going long TMF is much stronger than 3.8% per year. On a side note, I did notice that TMF costs much more to borrow. Another option would be to go long TMF and tell IB you are willing to loan your shares to collect the interest paid by short sellers. I still think shorting comes out ahead, but if you can’t short, loaning your shares could be a great option.
I don’t use MOC or LOC for my LI portfolio. I use limit orders intraday or near the close. I do use MOC for another strategy and it works well. My understanding is that you must place the order at least 15 minutes prior to close and you can’t modify it within the final 15 minutes.
Scott, I will be anxious to see the results of your more accurate backtests for this strategy. While you are correct that the most recent months for are accurate I think the most egregious errors in your data are the outsized returns from NFLX and TSLA in first half of 2013 before they were included in the index in June and July 2013. Which just goes to illustrate my point that momentum is often strongest just prior to entering the index. 2013 would not have been 100+% returns without those 2 components.
Also, does this strategy use the most recent list of Nasdaq 100 stocks each month the scan is run? I ask because EA was included in the index Dec. 12, 2014 and a position was taken Dec. 31, 2014. Would this trade have been possible given the way you run the scan?
Thank you.
Thank you Scott, that is helpful.
I do have another question about this strategy if that is okay. In your article detailing this strategy you said subscribers should keep in mind the fact that only the current list of Nasdaq 100 stocks were used. Survivorship bias is not the issue. The real concern is that the strongest candidates for a strategy like this would have been companies in the months and years just prior to inclusion in the index (perhaps this is also called survivorship bias?). It seems logical that companies just about to be included in the top 100 would very likely show the strongest momentum. These companies were growing rapidly, hence the reason they made it to the top 100.
Just one example: ADSK was included in the Nasdaq 100 on 12/20/2004 (according to historical data found on this site: http://www.nasdaq.com/indexshares/historical_data.stm). Your backtests show this strategy was invested in ADSK for 3 months in 2004.
I don’t know a lot about the Nasdaq 100, but I think they made less changes after 2011, so it is possible the most recent years are accurate. However, the strategy statistics (maxDD, CAGR, etc.) need to be thrown out since I’m sure there are at least a few other examples like the one I found.
Would it be so difficult to run your backtest with the current list of Nasdaq 100 stocks at the time of the scan? The data is not hard to find, it can be purchased here: http://marketcapitalizations.com/historical-data/historical-components-nasdaq/
I think this could be a very good strategy, but I would need to see more accurate backtesting to be sure this issue isn’t significantly affecting the results.
Thank you for taking the time to address my questions.
Makes sense Thank you. Are there any other factors to be aware of when making this substitution? What are the long term effects of holding TMF over 3x TLT? Also does TMF have a dividend? I assume the backtest results include the TLT returns with dividends.
I am curious how you came up with a 68 (or 66?) day look back period? How do different look back periods affect the results? I would be interested in seeing the stats on different look back periods you tested just to get a feel for the robustness of this strategy. Also, how does the strategy do using only the largest volume funds, say the top 20 or so? Thanks.
More specifically, your bar chart at the top right corner of the page uses cumulative return numbers for 2010 – 2012 instead of annual return numbers. Were the wrong numbers also used for other calculations? Thanks.
Maybe I am looking at something wrong, but the numbers from 2012 and earlier on the table of trades don’t match up with the bar chart and stats in your article. Which numbers are correct?
I am a new subscriber and I had the exact same question as Ronald. Hopefully this was a particularly bad month for the rotation because the allocation changed by such a large percent (50/50 to 80/20 in UIS unleveraged) coincidentally with a large move in the ETFs between the signal and execution. Frank, regarding your answer, I wonder why you don’t just use the opening price for the first day of the month on the website? If it doesn’t affect the results much overall, it would certainly make the percentages line up more closely to actual execution prices month to month. I’m sure as you say, long term it makes little difference but it would help me to see my percentages more closely match those on the signals pages month to month. Just out of curiosity, do you personally execute your trades at the close on the last day of the month or after the open on the 1st day? Thanks.
Thanks Alexander, this is what I thought. I’m just trying to wrap my mind around the implications. So in essence when I run a portfolio of strategies this way the individual performance of each strategy will be different than it would have been if run independently. By taking funds from the winning strategies and adding funds to the losing strategies each month an extended time of over-performance by one strategy along with under-performance of another strategy will be more detrimental to the portfolio than if they were run independently. However if the strategies tend to mean revert at different times then the blending of strategies could produce superior results. Am I understanding this dynamic correctly? Do you have any books or articles you would recommend to further understand these concepts?
Also, if MYRS is in the portfolio it’s not practical to rebalance just that strategy in relation to the other strategies bi-monthly. Though I can’t imagine this affects the results too significantly except in periods of extreme volatility. For the portfolio tool you created do you assume MYRS only rebalances with the rest of the portfolio monthly?
Thank you for your patience in answering my questions.
I have a basic question when it comes to building a portfolio of strategies. If I were to allocated 25% to 4 different strategies, then a month later say 2 were up 10% and 2 were down 10%. Would I rebalance so that I begin each month with exactly 25% of the total in each of the strategies or run them as 4 independent strategies? For example in a 100K portfolio rebalancing monthly would make each 25K in month 2 but running independently it would be 27.5K in the 2 winners and 22.5K in the 2 losing strategies. Those 2 scenarios would produce very different results by the end of a year and especially at the end of multiple years.
Which scenario does the portfolio builder assume when looking at all the statistics like MaxDD, Sharpe, CAGR of my combined portfolio of strategies?
Alexander, The max drawdown for UIS-SPXL-TMF says -16.95%, but I assume it was probably 3x that amount? That DD looks like it is from the March 2009 date when data from the normal UIS was used. Does this then affect all the other data for UIS-SPXL-TMF – sharpe, volatility, etc.? When incorporated into a portfolio it will look like your are getting the returns of UIS-SPXL-TMF but with the risks of normal UIS at least in regards to 2008 type DD. If max DD were in the -50% range that would change things quite a bit. Am I understanding this correctly?
Thanks for all your work on this! It looks really good. I’m looking forward to the meta-strategy.
