Friday, July 26, 2013

Motley Fool CAPS - Analysis Through Data - Part 1

So a little known feature of Motley Fool is that they have an "API" that includes most of the data from their user based research / stock rating game called Motley Fool CAPS. I say API here lightly because in actuality it is a data-dump of ticker history and CAPS ratings throughout the years.

I stumbled upon the CAPS API because I started getting into coding about a year ago through CodeCademy. In March of 2013, CodeCademy started a whole group of lessons based upon APIs. Being a fan of investing and the Motley Fool Caps platform, I did some Googling and after a few conversations with their excellent CAPS support team, I was up and running within just a few hours.

For those of you interested in doing your own research with CAPS historic data, I'll be writing an entire blog post on how you can quickly get started with the CAPS API and how you can go about parsing through the data.

Here is What I've learned From the Data


Stock ratings early in the beginning of CAPS had better success at predicting which stocks would outperform and underperform the market.

The median star rating for the best 200 stocks of the market in 2006 was 3 (due to some surprises), the median for the next top performing 600 stocks was 4. The average rating for the best 1000 stocks in 2006 was > 3. Likewise, the bottom 1000 stocks had an average star rating < 3 and all five quintiles of the worst 100 stocks in 2006 had a star rating of 3 or less.

Meanwhile, in both 2010 and 2011, the median and average ratings for both the top 1000 and bottom 1000 stock performances for both years have centralized around a star rating of 3. 

This makes sense, because aggregated top quality user-generated content tends to get drowned out by the masses as a site continues to gain more users. Additionally, CAPS continues to maintain the picks of IDs that have long since abandoned the game, as well as users that have maintained a pick in their portfolio for years, though the CAPS wiki does say that they factor time frames into their algorithm. What was true for a stock then may not be true today, so while many may have a rated a stock well in the past, newer users may not like the current stock price and have thumbed it down -  leading to a regression to the mean.

CAPS is better at predicting losers than it is at predicting winners.

The only thing that has remained consistent over the years is that the average star rating continues to decline with the worse a stock gets in the bottom 1000 performing stocks. For example, the worst quintile in the bottom 1000 of 2010 had an average rating of 2.01, the second worst quintile had an average of 2.75, the third 2.98, and so on. 

Unfortunately, the same cannot be said for the best 1000 stocks. The average rating of the best 200 stocks in 2010 was 2.42, the next 200 had a rating of 2.83, and the third a rating of 2.90 - which seems counter intuitive and can make picking the actual losers much more difficult.

However, there does seem to be an under-representation of 5 star rated stocks in the bottom 1000 worst performers compared to the other ratings, indicating that sticking to just 5 star stocks may help you mitigate some steep losses.

With the exception of stocks unrated by the CAPS system, all 5 star rating levels are almost evenly distributed.

This was a surprise to me, essentially all 5 star rating levels, 1 through 5 have just about the same number of stocks in each bucket. I was expecting a whole lot more stocks rated as 1s and 5s, due to convictions or a spike of 3s due to uncertainty, but it turns out that CAPS stock picks are just about evenly distributed. I assume that this can be attested to some sort of smoothing algorithm within the CAPS code.

Lastly, here is a sample of the summarized data. 

If anyone is interested in doing their own research, I'm willing to upload all years of data (2006 - 2011) to Google Docs. You can also download it and do your own analysis by going to the Motley Fool Developer website. And stay tuned because I have plenty more information to share in the following weeks coming from this excellent resource of data!


2006 Count % of Total Average Return Median Return Average Star Rating Median Star Rating
All
No Rating 1674 59.3% 17.8% 14.3% N/A N/A
1 Star 538 19.1% 25.6% 12.3% 1 1
2 Star 558 19.8% 23.5% 14.4% 2 2
3 Star 578 20.5% 25.0% 17.3% 3 3
4 Star 578 20.5% 29.3% 21.0% 4 4
5 Star 570 20.2% 31.5% 25.9% 5 5
Top Quintile (Best) 899 89.6% 63.6% 3.25 3.00
Second Quintile 899 30.7% 30.1% 3.24 3.00
Third Quintile 899 30.7% 30.1% 3.24 3.00
Fourth Quintile 900 4.2% 4.7% 2.98 3.00
Bottom Quintile (Worst) 900 -22.4% -17.8% 2.70 3.00
Average Return 23.6%
Median Return 16.0%
Average Star Rating 3.03
Standard Deviation 0.61
Best 1000
No Rating 226 29.2% 72.6% 51.1% N/A N/A
1 Star 126 16.3% 114.8% 65.0% 1 1
2 Star 121 15.6% 97.3% 59.6% 2 2
3 Star 156 20.2% 82.0% 62.9% 3 3
4 Star 176 22.7% 85.7% 67.6% 4 4
5 Star 195 25.2% 72.1% 60.7% 5 5
Top Quintile (Best) 200 191.0% 142.2% 2.91 3.00
Second Quintile 200 82.6% 81.4% 3.50 4.00
Third Quintile 200 59.7% 59.5% 3.29 4.00
Fourth Quintile 200 48.3% 48.1% 3.29 4.00
Bottom Quintile (Worst) 200 41.3% 41.4% 3.28 3.00
Average Return 84.6%
Median Return 59.5%
Average Star Rating 3.25
Standard Deviation 1.02
Worst 1000
No Rating 291 41.0% -18.0% -12.2% N/A N/A
1 Star 179 25.2% -24.7% -20.5% 1 1
2 Star 156 22.0% -21.3% -13.6% 2 2
3 Star 143 20.2% -21.8% -18.1% 3 3
4 Star 145 20.5% -19.6% -15.5% 4 4
5 Star 86 12.1% -16.7% -12.7% 5 5
Top Quintile (Best) 200 -4.0% -4.0% 2.90 3.00
Second Quintile 200 -9.1% -9.0% 2.72 2.00
Third Quintile 200 -15.7% -15.5% 2.82 3.00
Fourth Quintile 200 -25.9% -25.9% 2.83 3.00
Bottom Quintile (Worst) 200 -47.3% -44.1% 2.38 2.00
Average Return -20.4%
Median Return -15.5%
Average Star Rating 2.72
Standard Deviation 0.16




2010 Count % of Total Average Return Median Return Average Star Rating Median Star Rating
All
No Rating 822 17.3% 38.4% 6.0% N/A N/A
1 Star 823 17.3% 52.6% 9.6% 1 1
2 Star 957 20.1% 28.0% 18.6% 2 2
3 Star 1016 21.3% 28.5% 20.0% 3 3
4 Star 1036 21.8% 24.8% 19.2% 4 4
5 Star 929 19.5% 19.8% 18.1% 5 5
Top Quintile (Best) 1116 145.7% 74.4% 2.93 3.00
Second Quintile 1117 33.2% 32.4% 3.24 3.00
Third Quintile 1117 16.1% 16.1% 3.25 3.00
Fourth Quintile 1117 1.5% 2.3% 2.96 3.00
Bottom Quintile (Worst) 1116 -39.9% -33.3% 2.90 3.00
Average Return 31.3%
Median Return 16.1%
Average Star Rating 3.06
Standard Deviation 4.71
Best 1000
No Rating 91 10.0% 433.3% 97.1% N/A N/A
1 Star 171 18.8% 265.5% 89.0% 1 1
2 Star 193 21.2% 107.3% 84.9% 2 2
3 Star 205 22.6% 106.7% 82.5% 3 3
4 Star 210 23.1% 90.6% 71.0% 4 4
5 Star 130 14.3% 80.8% 67.3% 5 5
Top Quintile (Best) 200 475.4% 191.5% 2.42 2.00
Second Quintile 200 109.6% 106.8% 2.83 3.00
Third Quintile 200 80.2% 80.6% 2.90 3.00
Fourth Quintile 200 64.6% 64.2% 3.25 3.50
Bottom Quintile (Worst) 200 54.9% 54.8% 3.17 3.00
Average Return 157.0%
Median Return 80.6%
Average Star Rating 2.93
Standard Deviation 11.02
Worst 1000
No Rating 253 33.9% -57.4% -56.0% N/A N/A
1 Star 189 25.3% -47.2% -42.1% 1 1
2 Star 129 17.3% -40.6% -35.9% 2 2
3 Star 151 20.2% -34.5% -30.1% 3 3
4 Star 150 20.1% -35.1% -31.9% 4 4
5 Star 128 17.1% -32.5% -27.8% 5 5
Top Quintile (Best) 200 -16.1% -16.1% 3.17 3.00
Second Quintile 200 -25.2% -25.1% 3.04 3.00
Third Quintile 200 -37.4% -37.5% 2.98 3.00
Fourth Quintile 200 -55.1% -55.0% 2.75 3.00
Bottom Quintile (Worst) 200 -82.9% -83.3% 2.01 1.00
Average Return -43.3%
Median Return -37.5%
Average Star Rating 2.86
Standard Deviation 0.25


No comments:

Post a Comment