by cwwmbm
So, I’ve been doing some thinking that it would be nice to see some stats behind BTFD mentality. For that, the formalized metrics would have to be established of:
- What is the “dip”?
- When do you exit to avoid being in the future dip?
- What’s the performance like?
Obviously, the stats would have to go back enough to make sure we account for at least a couple of major recessions.
First, some generic considerations:
- I would personally like to make 7-15 roundtrip trades per year.
- No day trading. Buy and sell at market close.
- I’d like to time them so that I’m keeping cash for majority of the time.
- I don’t aim to beat the market, but I do aim to outperform the market in bear and flat markets on average.
So, one of the key metrics in my definition of “dip” is RSI. Your may vary. We all heard of undersold/overbought conditions. However the default RSI period is 14 days and that’s just way too much. It’s very slow moving and a lot of the dips I’d be trying to capture will be missed. We’re trying to deal on the scale of 5-15 days per trade. In other trading strategies I’ve considered over the past 2 years I’ve been heavily dealing with RSI(2) and RSI(5), so let’s stick with them.
We’ll define the “dip” as RSI(2)<=15 and RSI(5)<=35 (obviously, these numbers aren’t just random, they came from doing an analysis of the last 2 years of SPX performance).
Next, we need to define when do we exit. Criteria 1 will be “overbought” condition of RSI(2)>=95 and RSI(5)>=85. However it might take months to get there, especially in the bear market, so we also need a Plan B. A lot of trend reversal and technical analysis based strategies rely on crossing EMA(8), so out plan B is exit when we cross below EMA(8). Let’s also cap our gains at 10% per trade.
Next, we’ll need to create some exceptions to the cases when the massive sell off occurs:
- Don’t enter a position for a at least a day after 4% or more daily decline
- Stop loss of 3%.
Here are the backtest results:
| Year | SPY performance | BTFD SPY performance | # of trades |
|---|---|---|---|
| 2000 | -9.74% | 5.90% | 15 |
| 2001 | -11.76% | -5.53% | 15 |
| 2002 | -21.58% | 8.34% | 17 |
| 2003 | 28.18% | 26.78% | 11 |
| 2004 | 10.70% | 15.91% | 12 |
| 2005 | 4.83% | -1.79% | 11 |
| 2006 | 15.85% | 16.02% | 13 |
| 2007 | 5.15% | 17.24% | 14 |
| 2008 | -36.80% | 9.16% | 17 |
| 2009 | 26.35% | 22.28% | 14 |
| 2010 | 15.06% | 1.35% | 9 |
| 2011 | 1.89% | 18.24% | 17 |
| 2012 | 15.99% | 16.76% | 10 |
| 2013 | 32.31% | 24.80% | 10 |
| 2014 | 13.46% | 14.54% | 11 |
| 2015 | 1.23% | 21.16% | 18 |
| 2016 | 12.00% | 17.44% | 12 |
| 2017 | 11.90% | 5.80% | 5 |
Some stats:
- Overall trades: 231
- Winning trades: 148 (64%)
- Average win: ~3%
- Average loss: ~2.6%
- Average days in trade: 9
- Average trades per year: 11
| Parameter | SPY | BTFD SPY |
|---|---|---|
| Annualized return | 5.13% (~7% with dividends) | 13.46% |
| Worst year | -36% (2008) | -5.33% (2001) |
| Best year | +32% (2013) | +26.78% (2003) |
| Number of years with negative returns | 4 | 2 |
| Number of years with >10% | 10 | 11 |
| Largest drawdown | 55.19% (2009) | 26.96% (2002) |
Instead of conclusion, few additional fun facts:
- All backtesting was done in Excel based on the historical data from Yahoo Finance. It only goes back to 1994. 2001 was the first negative year since then using this strategy.
- I think this shows that BTFD works.
- Backtesting the same thing on QQQ yielded same-ish results.
- When backtesting on several different ETFs you need to account for higher volatility of some of them. Example – for ETFs like QQQ, XLK, XBI the maximum allowed daily decline should be more than 4% as they are more likely to occur routinely. Same goes for stop losses for them, as well as profit caps.
- If you go further and analyze the behavior of volatility and volume during the crashes you can time your entrance a bit better via additional conditions on Volume EMA and Volatility. I put the following condition: do not enter the trade is the Volume exceeds VolumeEMA by 35% and Volatility exceeds 10% at the same time. It brought the annualized return up to 14.59%, and decreased the largest drawdown from 27 to 21%. This was done after an observation that a lot of big dips (i.e. the ones where it goes much lower than our pre-set “dip” definition) start with the massive increase in volume. This condition makes sure we don’t enter the trade in the first 1-2-3 days of the “big dip”.
- When the stop loss is triggered it doesn’t mean the system doesn’t buy the same dip, it defaults to the same rules of entering the trade, often entering the trade the next day after the stop loss. The philosophy here is that the larger the dip the better the recovery. Absence of stop loss worsens the performance slightly – to an annualized return of 12.39%.
- If instead of buying SPY at the dip you buy a 3x leveraged ETF your annualized return will be 37.37% over the same period. Your wors year will be -22.49%, your best year 94.81%, and your largest drawdown -68.34%
Disclaimer: Consult your financial professional before making any investment decision.