2018 - The Year the Quants Got Destroyed

Jason Leavitt  |

The market moved up 3 of 4 days last week, and despite the one big down day, the weekly candle closed with a solid gain and near its high.

This is the first time since late October the S&P has posted back-to-back up weeks. If the mini move up continues a few more days, quants will be out in full force arguing the positive end of one year and good start to the next year has bullish implications for the entire year.

Some argue the Santa Claus rally (last 5 days of a year and first 2 days of next year) is predictive.

Others say the January Effect (first week of January? first two weeks of January? entire month of January?) is a better predictor of what's in store for the year.

But how much trust should be put into quant studies, given 2018 was the year the quants got destroyed?

Let's go through a check list of "studies" that suggested very high odds the market would do well in 2018 but ended up being very wrong. Anyone who blindly went long and just held, assuming historical tendencies would win out in the end, lost big time.

In January, the S&P traded far enough above its December high to suggest posting a loss over the following 3, 6 and 12 months would be a rarity. Not 2018. The S&P posted a loss off its January high 3 and 6 months out, and absent a gigantic rally, it'll post a big loss 12 months out too.

Since 1970, when the S&P has sported a year-to-date gain in May (any time in May, not just at the end of the month), it has posted a gain for the entire year 35 of 36 years. A random occurrence is one thing, but 35 of 36 years is pretty convincing. 2018 was the second miss. The market did indeed post a ytd gain in May - at one point it was up 7%, and although the index moved above and below the unchanged level, it closed the month with a gain – but for only the second time, it didn't post a gain for the year.

Since 1950, when the S&P gained more than 15% in a year - as it did in 2017 - the first 5 months of the following year posted a gain (check), and the final 7 months of the year have averaged a 5.6% gain (miss). Not last year. The market posted a 7.3% loss June-December.

The S&P closed two standard deviations above its 50-day MA in June for the first time in 6 months. History says there are high odds of posting a gain the following six months. Not last year - a 5% loss was recorded.

When the Nasdaq outperforms the S&P by more than 6% during January-May and both the Nas and S&P are positive during that time, the S&P's return the rest of the year is relatively big, with a 90% win rate. 2018 recorded an 8% loss.

When the Russell 2000 outperforms the S&P by more than 5% during January-May and both the Russell and S&P are positive during that time, the S&P has posted a solid gain 5 of 5 times the rest of the year. Again, not 2018. The S&P closed down for the first time (albeit, the sample set is small).

Since 1989, when the Russell 2000 has posted a 7-week win streak for the first time in 6 months, the S&P has posted a gain 6 and 12 months out 100% of the time, with a median return of 15.5%. In 2018, six months out the S&P was not only nowhere near the median return, it was down a healthy amount. Time will tell for the 12-month period, which is currently posting a 7.3% loss.

Also, when this happened, (7-week win streak for the Russell), RUT itself posted a gain 6 and 12 months out 9 of 10 times, with a median 12-month gain of 22%. In 2018, the Russell posted a double-digit loss six months out. Time will tell with the 12-month period, which currently sports an 18% loss. So much for the 90% win rate.

Since 1950, when the S&P was up 3% heading into the summer solstice, it posted a full-year gain 35 of 35 times. The S&P posted a 6.7% loss in 2018, the first time the stat didn't fulfil.

The S&P posted a gain for April, May, June and July. Since 1950, when this has happened, the index has never posted a loss 4, 5, 6 and 12 months out. Until 2018. The 4 and 5-month time frames posted losses. The 6 and 12-month periods are still "open" and sporting a 10% loss.

Since 1950, when the S&P closed higher at the end of July than it did at the end of February, March, April, May and June, the index has closed up on the year 23 of 24 times. 2018 was the second miss.

When the Nas hits an all-time high in July, as it did in 2018 for the 11th time, the index has posted a gain 6 months later 9 of 10 times and has never posted a loss 12 months out. As of now, with 3 weeks to go since the July 25 ATH, the Nas is down about 14.5%.

Since 1950 there have only been 5 occurrences when the S&P posted a gain April, May, June, July and August. In each instance, the market closed higher the final four months of the year. The average and median gains were 10.9% and 10.3%, respectively. This compares to average and median gains of 3.7% and 3.6% for "all years." In 2018, the S&P dropped 13.4%, the only instance the study didn't fulfill.

The stats are just as bullish when considering all 5-month win streaks, not just April-August. Since 1950, there have been 25 times the S&P has posted a 5-month win streak (not double-counting longer win streaks that have more than one 5-month win streak within the bigger streak). Only once has the market not posted a gain 12 months later. The average and median gains were 13.2 and 11.4%.

The best 5-month block during the 48-month presidential cycle is the October preceding midterm elections, along with November-February after. Not right now - at least not yet.

Since 1950, when the S&P has posted a Q3 gain greater than 7% (as it did in 2018), the index has posted a Q4 gain 13 of 14 times, with an average gain of 5.9%. In 2018, the Q4 loss was 14%.

Over the last 20 years, October has been the best month of the year. Not October 2018.

One more (and there are others).

Since 1950, the S&P had never posted a loss between its October low and the end of the year during midterm election years. The index was 18-for-18. The average and median gains were in the mid 10% area, with four years posting gains less than 5% and five years gaining more than 13%. In 2018, the S&P dropped 97 points, or about 3.7%, making it the first time the S&P has posted a loss.

These are not random, hand-picked studies. The underlying tone is that when the market is able to accomplish certain things at certain times of the year, odds heavily favor a continuation of the positive vibes. But 2018 was the year everything got blown to bits. None of these studies fulfilled. Studies that that had 100% or nearly-100% win rates didn't fulfill - many by a wide margin.

Either 2018 was just a one-off aberration that will go down in history as the year the market went off the rails and made no sense, or it's a sign of very bad stuff in the pipeline.

There's another possible explanation...

These statistics used to only be available to big hedge funds with multi-million dollar research budgets. But today, the data is available for very cheap, and decent knowledge of Excel is all one needs to unearth some extremely interesting tendencies.

What used to be available to a select few is now available to everyone. Even if you don't know Excel or Python or R, quants are more than willing to share on Twitter. It makes me wonder if 2018 was the start of a trend. Perhaps now that everyone has access to this stuff, there'll be many more instances when the studies don't work. After all, if everyone is sitting around waiting for everyone else to buy, how is the market to move up and fulfill the odds?

It's something to consider. I wonder if the same thing will happen to the quants. What used to work because very few had access will stop working because everyone is looking at the same stuff. It's on my mind as we exit a year where historical tendencies did not play out.

What did work?

The internals. At the January and September tops, the market's breadth was in decline. Even though the indexes were trending up and numerous studies pointed toward a continuation of the gains, the internals didn't support a continuation of the moves. We didn't know if the market would correct for a week or month. We didn't know if the S&P would drop 3% or 10%. But we were told a pullback was coming.

One more comment.

Per Michael Batnick (@michaelbatnick), since 1990...

During bull markets, the market has been positive 54.4% of the time and negative 45.6% of the time, so a slight advantage for the bulls, but not much. The average up day was +0.67% while the average down day was -0.66%, so essentially the same.

During bear markets, the market has been positive 46.7% of the time and negative 53.3% of the time, so a slight advantage for the bears, but again, not much. But the average up day was +1.19% while the average down day was -1.32%. Close, but slight advantage to the bears.

The point is this: during bull markets, the market moves up a little more than it moves down, and during bear markets, it moves down a little more than it moves up. But there's a big difference in the average daily change. During bull markets, the S&P moves ~ 0.66% each day while during bear markets it moves twice as much.

This is the animal we're now dealing with. Instead of calm days that come with little movement, we're getting volatile days with lots of movement. Things happen fast. What used to take weeks to play out can take place in a couple days.

Jason Leavitt

Sources: @bullmarketsco, @RyanDetrick, @SJD10304

DISCLOSURE: It is not recommended a trade or investment be initiated simply because stats say the trade worked in the past.

The views and opinions expressed in this article are those of the authors, and do not represent the views of equities.com. Readers should not consider statements made by the author as formal recommendations and should consult their financial advisor before making any investment decisions. To read our full disclosure, please go to: http://www.equities.com/disclaimer



Symbol Last Price Change % Change










Blockchain in Fintech - Discussion at the EU Parliament

From the recent Blockchain For Europe Summit in Brussels: Panel on Financial Market Infrastructure