How do we manage to stay ahead of the curve? Through the process of thorough and methodical backtesting. And we’re not talking about just punching some numbers in Amibroker or some similar program. All the data we use is entered manually into Excel spreadsheets. The formulas are painstakingly checked and cross referenced. To avoid any kind of user bias or entry error, the process is stretched out over weeks and months. In such a long process the temptation to “overfit” is kept in check by keeping the algorithms concise and the system rules to a bare minimum. The spreadsheet design we use is derived from the fabulous book “How to Backtest a Trading Strategy Using Excel” by Mark Ursell, a must own for any serious system trader.
Before setting out to backtest a system, it’s a good idea to have a thesis. Not just “I’m doing this to make money,” but rather something a little more ideological that you can put to the test. In my years as a trader I’ve discovered that 2 rules hold true to my core philosophy:
Never short a bull market
Whenever there’s a correction, refer to rule #1
Okay, so now on to the technical stuff…
The system trades SPY options, but first the tests were run using the SPY ETF. While price data for SPY is readily available, historical options prices are a bit more elusive. End of the day data is available back to 2000 (for a price from sites like ivolatility.com), but intraday only goes back to ’11. Since many of our entries are at the open, EOD prices can’t fully suit our needs, so to get an accurate picture of options over the 16 year backtesting period (and save quite a bit of $$) we use the Black-Scholes method.
Using Black-Scholes for historical prices raises the issue of volatility. Implied volatility inherently looks toward the future, so for the calculation of past prices, historical volatility was used. Factors like time decay (causing the win rate to drop slightly) and leverage (increasing the payout ratio substantially) are all reflected in the model. How do we know we got it right? The returns of the backtesting results using options are comparable to the ETF results (minus the leverage), and they also hold up when checked against the last 3 years of live trading.
Here are some of the highlights:
1. We made a 1.83 million percent return
That’s not a typo. Over the 16 year period from Jan 2000 to December 2015, the TVO System (which includes TVO, HG and IO) turned $10,000 into $183 million. Here’s the equity curve compared to SPY. It’s a pretty straight line even without a regression trend line to compare it to, but in case some of you quants are wondering, R-squared equals about 95%.
You don’t need to be right 80-90% of the time to accumulate long-term wealth, in fact it’s often the high rollers, who boast about their one or two-year long winning streak, that end up blowing up their account when their luck suddenly runs out. A slow and steady 60-70% success rate has a much better chance of making it over the long haul… just think of the tortoise and the hare.
And speaking of winning streaks, the model produced a maximum of 11 consecutive wins, close to double the amount of straight losses. With more winners more times in a row, along with the average profit almost 50% higher than the average loss (1.48 payout ratio), it’s not hard to comprehend how the system is able to stay firmly to the upside.
2. The largest account drawdown was -30%
Markets don’t go up all the time and neither do account balances. But how often do they go down and how long do they stay there? This chart shows how many times and to what extent the account went “underwater” before recovering.
Drawdown reached -30% only twice (in ’02 and ’08). In all the other cases, DD rarely exceeded 15% and the sharp spikes in the graph show it didn’t stay down there for very long. Having your account take any kind of hit is not a great feeling, but what eventually saves the day is the time it takes to get back in black. Even with 16 years filled with bear market corrections, the TVO System managed to finish positive (and then some) every single year.
3. In the 4 worst years for SPY, we made money
Not only has the system never had a losing year, but we also showed a substantial profit in ’00, ’01, ’02 and ’08. In the years when the market was falling below -10%, we were making an average of +20%. Even in the good years for the market we still outdid SPY (in many cases by 10 times or more), which is not an easy task as most price systems tend to under perform in a super bull market. Here are the TVO Systems annual returns vs. SPY.
Oh, and did I mention that the TVO System ignores all types of price data? All of our algorithms are based only on volume and market internals. Price can be manipulated, so we ignore it… plain and simple.
4. We were in the market less than a 3rd of the time
Keeping market exposure under 30% greatly reduces risk by leaving most of your capital intact out of harm’s way. Losses are also decreased by limiting the number of options contracts per trade. You can’t lose more than the premium paid, so the position size of each trade is kept in a range close to 10% of total account value. Because of this, the system does not require the use of stops… That’s right, no stops. In fact when you add things like stops and targets to the backtest you may see a higher win rate, but overall returns sink dramatically.
Just about anything that could cause risk to become unmanageable was placed under intense scrutiny. For that reason, methods like averaging in to a position 2 times or more were excluded from the system. And (drum roll please) going short or buying puts also didn’t make the cut.
Wait a minute… Does that mean that all of these results are LONG ONLY?
We tested previous versions of the system that had a short side, but no matter how much tweaking was done (and remember over-tweaking is generally not a good idea) the results on going short always fell far below the 60% win rate benchmark set for a successful strategy. The returns could marginally end up greater, but adding more trades ultimately places more risk on the table, leading to overtrading, which opens the door on another factor… Trading psychology.
The general consensus out there is that in bear markets you need to go short to make money. One of the true benefits of doing this backtesting study is being able to say with confidence that the opposite is true.
So what allures folks to do so much shorting then? Thanks to all the outlier stories about Jesse Livermore, Martin Zweig, and more recently Michael Burry in The Big Short, the romantic myth of making a killing while the rest of the world goes down in flames will be perpetuated by traders for generations to come. What’s often not discussed is the mental pain many of these souls endured from guilt, loss of relationships and, in many cases, persecution from their peers and loved ones for their actions.
Perhaps it’s time that traders start to see short trading for what it really is… a device invented by brokers as a way to continue to rake in fees in the absence of buyers when markets are going down. Stay ahead of the curve by paying less fees and buying the market when it’s down and selling when it goes up. It’s not a new idea, but we put it to the test and found out it still works.
Here are the complete stats:
|TVO System Backtesting Results '00 - '15||SPY call options||SPY ETF long|
|Net Profit %||1830313.74%||1507.07%|
|Market Exposure %||28.76%||66.11%|
|% Profit p/yr (CAGR)||84.87%||18.99%|
|Risk Adj Return||295.08%||28.73%|
|Avg P/L (Expectancy $)||$404,936.67||$333.42|
|Avg P/L (Expectancy %)||2.39%||0.64%|
|Avg Bars Held||8.8||8.8|
|Winning Trades||298 (65.93%)||322 (71.24%)|
|Avg Profit %||5.58%||1.59%|
|Max Winning Trade||$26,990,713.78||$7,181.48|
|Max Winning Trade %||47.47%||20.13%|
|Max Winning Streak||11||19|
|Losing Trades||154 (34.07%)||130 (28.76%)|
|Avg Loss %||-3.77%||-1.71%|
|Max Losing Trade||-$9,599,050.29||-$4,259.17|
|Max Losing Trade%||-12.58%||-9.05%|
|Max Losing Streak||7||6|
|Max Drawdown $||-$15,243,241.69||-$8,435.11|
To test drive the TVO System for yourself subscribe today. The first month is free.
To view past positions check out our Trade History.
Performance results on this website dated prior to September 2014 for TVO (prior to May 2015 for HG, and prior to May 2016 for IO), including backtesting and trade history, are simulated. Please read our full disclaimer.