# Building the Inverse Jim Cramer Index

### So, yeah, we built a risk-adjusted inverse Jim Cramer Index. Launching soon. We go through the entire process.

**The Strategy**

A long-only portfolio mimicking an “Inverse Cramer” strategy through positions in stocks described as “sells” on Jim Cramer’s Mad Money. Lightning round picks omitted, rebalanced monthly. Risk-adjusted pre-rebalance through volatility + Sharpe-hedging, as well as anti-momentum hedging based on previous monthly returns. Check it, and other portfolios out at Quantbase.

**The Data**

The data is pulled from The Street, a financial news and literacy site. 23,000 rows of data following every Mad Money episode, every call on every segment on the show. Segments divided into Lightning Round, Game Plan, Sell Block, Am I Diversified?, Mad Mail, and 10 others. Calls ranked from 1 to 5 (inclusive) — 1 being a highly negative, 5 being a highly positive, with less than 1% at neutral (3). This was all sourced on a pretty untransferable HTML. Tried a bunch of things, like copy-pasting to Excel, downloading the source, etc. What ended up working was saving the webpage to a 644-page PDF, copying the data to an HTML stripping site, then pasting it to Excel finally.

Data cleanup consisted of sorting the data by ticker, date, segment of the show (Lightning Round, etc), Cramer’s call (our primary source of alpha) whether positive or negative, historical and current prices, as well as by rebalance period. After that, we did cleaned out names that were no longer investable at the point of investment in the backtest — aka if a rebalance was to occur in December 31, 2016 of stocks between October and December that year, and a stock mentioned by Cramer in October was no longer tradable in December, it had to go. Decided not to do any statistical or call/period-based cleaning until later on in the process (as this entire process was iterative) As we are looking at an aggregation of data from a set past period to then make the picks that we’ll hold onto for the next period, a backtest looks at each rebalance period as if the only information it knows is the information presented up until that date. Effectively, it means that we can’t use 2021 data to “predict” stock prices in 2016. Duh, right?!

**The Testing**

Backtesting, again, is iterative. We start off with a set of data, and come up with strategies that do well with that data in backtests, and then hope that this process holds over the following periods. A good investment strategy works until it doesn’t, meaning you have to tweak it again to create any sort of alpha. We expect to have a team tweaking this index from time to time for that reason.

To set a baseline, we did a “naive” test that doesn’t actually make any kind of strategic (allocation) or tactical (timing) decision-making happen. Cramer does about 150–200 “1” picks (aka “sell immediately”) every quarter. We picked a random 10 per quarter, for all 24 quarters that data existed for, then did the backtest on it. We ran this naive testing algorithm 500 times — here are the average results:

Correlation with the S&P — 66.69%

Daily Volatility — 1.32%

Sharpe Ratio — 0.18

Max Drawdown — 53.73%

Annual Return since 2016 — 1.27%

Ouch! Clearly, just naively picking Cramer’s inverse stocks doesn’t work. We need to be more clever here. Deep discussions of each of these stats will take another blog post. For this one, we focus on the Sharpe and the Annual Return. The Sharpe ratio is a measure of risk-adjusted return, calculated by taking the mean return of a portfolio over a period and dividing by the standard deviation of the portfolio, then adjusted based on the frequency of the return data. A Sharpe above 1 is considered great for a portfolio (“I’m getting a return without having to take on a similar amount of risk”). A Sharpe between 0 and 1 is considered not so great. Below 0 means the portfolio is just losing money, and you’d be better off investing in bonds.

We also ran the naive algorithm multiple times with Cramer’s “2” calls.

Cramer’s 2’s:

Correlation with the S&P — 71.12%

Daily Volatility — 1.39%

Sharpe Ratio — 0.42

Max Drawdown — 46.04%

Annual Return since 2016 — 8.00%

This is…a little funny. Maybe Cramer does have some predictive power? Stocks he likes a bit better than absolute hatred, have a 3x Sharpe ratio, and the returns are starting to speak for themselves.

The S&P, by comparison, has had a stellar last decade, abut 15% annualized return. An inverse Cramer isn’t a panacea, unfortunately, for market-beating returns.

We tried a few other iterations. Naive but full test (investing in an allocation of all his picks, not just a random sample), weighting by volume, by market cap, by sentiment at date of rebalance, by sentiment at the date of the call, what it looked like without taking into account his rapid Lightning Round picks (accounting for 40% of all his calls), with different sets of his calls (maybe some 4s and 5s in there to hedge against the inverse-ness, but that’s no fun) and with different rebalance periods.

What we ended up with was a portfolio that rebalanced monthly, with just Cramer’s “2” picks, without any of Cramer’s Lightning Round picks. Test results below:

Correlation with the S&P — 81.31%

Daily Volatility — 1.05%

Sharpe Ratio — 0.51

Max Drawdown — 39.84%

Annual Return since 2016 — 8.54%

A Sharpe of 0.51 still leaves room for improvement — we’re not done yet. We did some portfolio optimization for weighting, then looked to decrease our risk by getting rid of the top n worst performers (in terms of highest volatility) pre-rebalance of the previous rebalance period. The n we arrived at was around 90 (thus making a much more concentrated portfolio than earlier). Final tests:

Correlation with the S&P — 77.13%

Daily Volatility — 1.15%

Sharpe Ratio — 0.61

Max Drawdown — 40.21%

Annual Return since 2016 — 10.94%

The last quarter has been negative, for the market as a whole but also for Cramer haters who look at all of his picks systematically (and not just the ones that get the most attention, like Coinbase, Robinhood, etc etc). If we take rolling returns, then without taking into account the last year (upswings in early 2021 included with the downturn of the last quarter), we get a better Sharpe of 0.89. Still not above 1.0, though, meaning the portfolio isn’t as optimal as it could be, on a traditional risk-adjusted basis. But that’s something that can be said for all portfolios at some point in their horizon, even the benchmark. SPY didn’t hit above a 0.63 rolling 12-month Sharpe for almost the entirety of 2018 and 2019. Some food for thought, and probably not the direction you might have expected an Inverse Cramer blog post to go. The index is still undergoing testing on our end, but if you’d like to see similar portfolios (Wall Street Bets, Pelosi Tracker, Twitter Sentiment) in action, check us out at Quantbase.

oh no, if it isn't the consequences of the efficient market hypothesis