Greedy Bad Assays

I'm writing this because one of my positions went down 23% compared to the previous day close.  It got me thinking about what happens on these massive one day drawdowns not in this particular case, but overall. Should I be greedy when others are fearful of bad assays or mine shutdowns?

Summary

Arithmatic math is a liar, but some insights can still be gained.

After a massive drawdown of greater than 20% in a single day stocks that bounce back do so hard. If you can pick the survivors from the wreckage it could be a good opportunity.

Timing your entry point is an art and not a science, but the 3 day rule isn't supported by this data set. If you are looking to go long and see a big single day price drop as a buying opportunity I'd consider buying earlier in the day to get some of the bounce at close or buying at close that day.

Methodolgy and Warnings

You can drown crossing a river that is six inches deep on average.  These numbers are going to be misleading in ways, inaccurate in others, biased as well.

I went and downloaded the daily historical trading data for all of the current constituents of the $GDXJ. This is my universe for this study, a bunch of terrible companies, but a bunch of terrible companies that had a good enough run to get included in an index and not yet removed from it. The data in this universe outperformed the index itself, but it's what I could get without too much trouble.  

I looked at days that the low was at least 20% below the previous day close. I got more than 2000 of these days across the 80 stocks I could get data for.

I then filtered for trading volume in a day of $50,000 in whatever currency the exchange uses (Australian Dollars, Canadian Dollars, and US dollars mostly) to reduce the noise from when they were shell companies or similar.  This surpisingly halved the number of data points. The final number of data points was 1050, which seems large enough to make the results pretty smooth.

It's worth noting these numbers won't add up due to the nature of arithmatic vs geometric means. If a stock goes -50% one day and then +50% the next day that is an arithmatic average gain of 0%, but in reality the stock is still -25% from where it started.  But I don't know how to geometrically average across different stocks other than backtesting a trading strategy, which I'll leave for future work and a follow up post.  So consider this blog post an indiactor that this is a promising area to do more research, but please don't go out and lose a bunch of money trading this until you've done your own work on it.

The Day

In reality nobody is going to bottom tick the trade on a massive down day. But what if you could?  Or what if you waited for the end of trading that day?

How much can you lose on a stock that is -90% on the day? The answer is you can lose another 100%. Think Bre-X.

However, there is often a rebound for stocks that are down more than 20%. There's a rush for the exit door. During the selloff there aren't a lot of buyers lining up for the freefall, but there are a lot of shorts covering and investors selling on the news. I think some long term investors once things seem to have stopped freefall come in and see opportunity for a good entry price.

But there's a reason nobody gets 74,217% returns or is richer than Warren Buffet. Timing trading is very very very hard. You'll be up against math geniuses with super computers. But if you think it's a quality company and the market is overreacting to the news the data is on your side to jump in before the trading day is over, just know you can also get your face ripped off.  But you are investing in junior mining companies, you can get your face ripped off any random day of the week.

Week After

Days are in trading days

These numbers are all kinds of wrong. Nobody is making 35-38% investing in the $GDXJ. The time I did this also matters as the $GDXJ is +26.62% in the last year, and this is an index I'd typically short not be long. If you can pick companies that will go on to be in the $GDXJ and then not fall out of it and choose your endpoint as a big gold price rally you are already in an unachievable universe of stocks. Converting arithmatic average daily returns to annual returns is also sus.

But wow, I think it's pretty clear that among these 80 stocks you'd really want to go back in time and invest after a huge drawdown, even if your holding period was 1 day to 1 week.

Multi-Week After

Days are in trading days

0.7996 * 1.26359 = 1.01.

So among our universe of stocks that didn't die they are on average back to where they started a month later after a seriously bad day.

Do As I Say Not As I Do

For the company that inspired me to go write some code to test what happens during big single day drawdowns in mining stocks the logical thing according to the data in this study would have been to add during the day and trim a month later.

Instead, I exited completely during the day, crystalizing a gain overall but a big paper loss from the day before. It was down a little more at close from my exit and halfway through the next trading day is -8% from that close. It is not doing what the data says these stocks should do on average. It turns out these stocks represent real businesses with real business developments.

I'm not publishing the code for this one because it was quick and dirty and I'm embarrased for anyone to see it, but if you'd like a copy just dm me in the usual places and I'll tar up the .csv daily data and the source and send it to you.