09.24.2022

## This article takes a quantitative look at how the S&P 500 behaves after Wilder's 14-day Relative Strength Index (RSI) flashes oversold readings.

#### KEY TAKEAWAYS

Where the S&P 500 will be 21 days after the RSI crosses below 30, with price below its 200-day moving average, has the same odds as a coin toss.

Where the S&P 500 will be 21 days after the RSI crosses below 20, with price below its 200-day moving average, has a 60% probability of closing higher.

#### INTRODUCTION

To start this past week, the S&P 500 (SPX) was sitting right at its volume-weighted average price (AVWAP), anchored to the March 2020 Covid-low. The AVWAP is an important level because it tells us the price at which the average share, since the Covid low, was transacted. In this context, I add it as a gauge of sentiment. The further below price dips and the longer price holds below it, the more the average market participant will endure unrealized losses.

Today is Saturday, so the last 5 candles show last week's price action. After chopping around the AVWAP Monday and Tuesday, Wednesday was also flat until the 2pm thrashing as Chair Powell confirmed another 75-basis point rate hike. After the hour of volatility, Wednesday’s last hour finished with a flush lower. Thursday was flat and closed at its session's lows. Friday gapped lower, trended lower all-day, and then buyers stepped in during last hour of the week and pushed price more than halfway up into the day’s range. While the candle is hammer-like, the handle is not technically 2/3rds of the body. We see the volume pressing higher during the final 3 days of the week, expressing increased emotions with price falling towards the June lows and under the AVWAP. The RSI become oversold with a reading below 30, and this is the focus of this article.

##### SPX Daily. Click to enlarge.

#### THE TESTS

__Test 1__

Now that SPX is oversold, using Wilder’s 14 period Relative Strength Index (RSI), can a signal test tell us anything helpful? The first test I ran asked: how did SPX act 21 days from the RSI crossing below 30, if the closing price is below its 200-day moving average?

Price was higher after 21 days 54% of the time and lower 46% of the time.

The average return was +0.59%. The median return was +0.79%.

While the standard deviation looks innocent enough at 5.87%, that deviation only speaks to the deviation around its mean return at 1 standard deviation, that is roughly 68% of the time, on the distribution of returns. The largest return was +23.5% and the largest loss was -16.25%.

The average winner was almost 4.75%, while the average loser was only -4.18%.

The kurtosis was 4.42. Being >3 implies the distribution of returns is a bit thinner, with a higher peak than a normal distribution.

The skewness was 0.14, which is close to 0, meaning the left and right tails of the distribution are nearly symmetrical.

For the full table of signals,

__click here.__

##### Distribution of Returns. Click to enlarge.

##### 21-Day Path of Mean Returns. Click to enlarge.

While the average and median return was slightly positive, with the odds close to a coin toss, this test did not reveal much about SPX 21 days from RSI 14 crossing below 30.

__Test 2 __

Then I remembered Perry Kaufman. Perry is a literal rocket scientist turned market participant. He worked on the Hubble Space Telescope’s predecessor, as well as navigation systems which where eventually used by NASA on the Apollo missions to the moon. A chapter from one of his books, titled *Trading Systems and Methods, *came to mind. He discussed systemic trade-offs and explained: when mean-reverting trades are entered at more extreme overbought or oversold conditions, the risk is reduced, the number of trades is reduced, and the profits are larger. This led to me the next test which asked: how did SPX act 21 days from the RSI crossing below 20, if the closing price is below its 200-day moving average?

Price was higher after 21 days more than 60.5% of time and lower 39.5% of the time.

The average return was +1.01%. The median return was +1.66%.

The standard deviation was 6.20%. The largest return was +14.25% and the largest loss was -13%.

The average winner was almost 4.75%, while the average loser was -4.75%.

The kurtosis was 3.41. Again, being >3 implies the distribution of returns is a bit thinner with a higher peak than a normal distribution.

The skewness was -0.37 is close to 0, meaning the left and right tails of the distribution are nearly symmetrical.

This test turned out as Kaufman suggested it would. There were less signals, with more wins, and a greater average return. The standard deviation was higher, but the maximum and minimum return values were less than the first test. With a 60% historical win rate, perhaps I would be more interested in a trade if the 14-day RSI dropped further to a reading of 20. That said, the last time it happened was 02/28/2020 which resulted in an 11.25% loss 21-days afterwards. The time before it however, on 12/24/2018, resulted in a 13.25% gain. __Click here__ to see the full list of signals.

##### Distribution of Returns. Click to enlarge.

##### 21-Day Path of Mean Returns. Click to enlarge.

__The side-by-side results__

__ __

##### Side-by-Side Signal Results. Click to enlarge.

##### 21 days after a cross below RSI 30 on left. 21 days after a cross below RSI 20 on right.

#### A SIMPLE TRADING SYSTEM

Since the signal tests above showed that we have some reason to be interested in a mean-reversion trade when the 14-day RSI crosses below 20, I wanted to test a very simple system to see what the results would look like.

The entry trigger was when Wilder’s 14-day RSI crossed below 20 when price was below its 200-day simple average.

Per Perry’s suggestion, the exits were when the RSI crossed above 50 (T1) and a second test exiting when RSI crossed above 70 (T2).

Both T1 and T2 returned interesting results. I will highlight a few of the interesting statistics. For a full breakdown of the results, __click here.__

Percentage of winning trades (PWT): T1 had 19 wins and 5 losses making the PWT an almost unbelievable 79%. T2 had 13 wins and 10 losses making the PWT 56.5%.

The percentage of profit on T1 was 56% and T2 was 36.75%.

The maximum adverse excursion, so the amount the trade went into an unrealized loss during the holding time, during any trade on T1 was -6.5% and T2 was -23%.

The average trade duration of T1 was 31 days and T2 was 183 days.

The profit factor, average $s per win/average $s per loss, for T1 was an unacceptable .83 while T2 had a profit factor of 3.2.

The biggest issue with these systems is the 13 year and 11 months they both took to recover from losses sustained during the 2000/2001 crash. That’s just not realistic or acceptable.

The RSI and 200-day moving average were not enough by themselves to act as reasonable trade triggers. I tried to help the system out by adding stops, but that did not improve the results at all. I tried an 8%, 10%, and several Average True Range trailing stops. I even tried adding Bollinger Bands to test some form of a standard deviation stop. These choked off the profits without meaningfully reducing the drawdowns of the system.

#### CONCLUSION

Anticipating if SPX will be higher or lower 21 days after an oversold reading of less than RSI 30 is not better than a coin toss.

Anticipating if SPX will be higher or lower 21 days after an oversold reading of less than RSI 20 increases the odds of price being higher 21 days later to roughly 60%.

All in all, in this downtrend, if Wilder's RSI can drop below 20, we might see a 21-day mean reversion.

As always, thank you for reading. If anybody reading this has any ideas or feedback, I encourage you to reach out.

This article is for educational and informational purposes only. Trade at your own risk. The author may or may not have a position in the securities mentioned. __Read our full disclaimer here.__ Please reach out with any feedback or comments. I would love to know if you agree, disagree, or don't care at all. __Louis@eastcoastcharts.com__

## Comentários