In last week’s post I wrote about a strategy for shorting shares that looked to sell short overbought, ‘supernova-type’ companies. The strategy produced some good results in testing and was also picked up elsewhere in the blogosphere.

Seeing that this supernova-type strategy appeared to perform so well I thought it would make sense to look at the strategy in reverse. In other words, to see what would happen if we were to buy highly oversold stocks instead. Stocks that might exhibit a ‘waterfall‘ type pattern.

We know that the stock market has an upward bias so surely a mean-reversion strategy like this will perform well?

Waterfall Pattern

The waterfall pattern is not one that is commonly talked about but my impression of it is that it’s basically the reverse of the supernova.

Trade example taken from system two results

Waterfall trade example for NAV taken from system two results

Instead of a stock that is going exponentially higher, this is a stock that has dropped sharply lower, with each bar becoming larger than the last so that there is an almost parabolic pattern on the chart, where the stock can seemingly go no lower without heading to zero. Volume should also spike higher as the sell-off takes place.

Putting The Idea To The Test

In order to test this idea, I simply reversed the rules of the original short-selling strategy. Thus, long trades are entered when:

• Stock has dropped more than 80% over the last 5 days.
• Stock price should be more than $2 and less than $20.
• Volume is more than 500,000 shares.
• Holding period 5 days.

In addition, commissions were set at $0.02 per share and starting capital was set at $100,000, with equity split equally between 10 positions. All trades are placed on the next day open.

Running this strategy on the Russell 3000 universe (including historical constituents) between 1/1/2000 and 1/1/2015 produced the following results and equity curve:

test 1 table of results


test 1 equity curve

As you can see, the strategy performed poorly. The compounded annual return was just 1.57% over 15 years and the maximum drawdown was -6.34% giving a CAR/MDD of 0.25.

The drawdown here is small but the system didn’t make many trades. Just 13 in fact. Of course, this isn’t surprising since it’s not often that a stock loses 80% of it’s market value.

Some optimisation of parameters

Because of the small number of trades, I decided to do some optimisation and test various levels of percentage loss (drop threshold) against various levels of holding periods. This way we can see what parameters work best.

The results are shown below:

optimisation results buying oversold stocks

As the results indicate, the best simulation came with the parameters 90 and 5. In other words, buying stocks that had dropped 90% or more over the last 5 days and selling them 5 days later gave the best results, with a compound annual return of 0.6% and a CAR/MDD of 0.57.

However, this simulation only produced one trade! Overall, looking at the table of results, it’s quite clear that, unlike the short-selling strategy, there does not seem to be nearly as much profit potential here.

What is happening?

So it seems that even though stocks have an upward bias, there is not so much profit to be had in buying sharply oversold stocks. I have to say that I found these results quite odd at first as I thought the stock market was much more mean-reverting.

However, it appears that when a stock drops a large amount in a short space of time there’s normally a good reason for it and the stock does not always recover for some time – if at all.

These results also go to prove the point that stocks fall a lot sharper and faster than they do rise, something to always be aware of when trading.

Test Two

Since the results were less than stellar for this system I decided to delve a bit deeper and try out some more ideas.

In the first test, we managed to find plenty of smaller cap stocks that had fallen by a lot but never recovered. So maybe we should be looking at large cap companies instead. Perhaps only stocks that are over $10.

And since we want to test the waterfall pattern, maybe we need to be more clear about this in the rules.

Therefore, for test two, I made the following adjustments to the rules:

• Buy a stock if it has dropped more than 15% in the last five days.
• Stock price should be more than $10.
• Volume is more than 500,000 shares.
• Most recent price bar should be larger than the previous two price bars combined.
• Holding period 5 days.

Hopefully, these rules will allow us to find stocks that more closely exhibit the waterfall pattern and are more likely to revert back to a higher level. (Commissions and other settings were kept the same as test one).

Test Two Results

Running this strategy on the Russell 3000 Universe between 1/1/2000 and 1/1/2015 produced the following results and equity curve:

oversold waterfall stocks table of results

waterfall stocks equity curve

As you can see, the results are still not good. The compounded annualised return was negative at -5.06%, with a maximum drawdown of -56% and a CAR/MDD ratio of -0.09. Interesting…


Reversing the short-selling system sounded like a good idea in theory but the test results above clearly show that this is not a profitable strategy.

There are no doubt much more profitable mean reversion systems out there (such as the overnight trading system which I talked about previously).  However, the evidence here shows that traders should be very weary about buying a stock just because it is oversold.

Often, there will be a good reason for why the stock in question has plunged. And often, those initial losses will turn into even heavier losses.

To sum up… Don’t Just Buy A Stock Because It’s Dropped!

Example of a trade gone wrong, taken from system 2 results

Example of a trade gone wrong, taken from system 2 results. Entry was made on the green arrow and closed on the red arrow.

Thank-You For Reading! You May Also Like:

Overnight Stock Trading System: Make Money In Your Sleep
How To Build Mean Reversion Trading Systems
How to Beat Wall Street: 30+ Trading Systems For Stocks

Tags: ,

2 opinions

  1. Hey JB – what a great article. Thank you!

    With those results from the second test, why not try it as a trend following strategy? i.e. short under those conditions, instead of buy?

    I might try it out too, it could be promising!

    Happy trending – Dave

    • Glad you liked it Dave. I think it’s good to show what doesn’t work as well as what does work.

      Trend following might be a good idea, with longer holding periods? Yes, I might try it. Cheers!

Leave a Reply

%d bloggers like this: