ETF Complete Strategy Insights: Slicing Up the ETF Models (Part 1)

James Kimball | August 18, 2019

The ETF Complete model closed the week down -1.5% compared to the SPY which closed down -0.9%.

Markets retested the prior week's low and closed right in the middle of the range for the last two week, situated between the 50-Day moving average and 200-Day moving average. Concerns about trade and interest rates dominated the conversations about the market activity.

Stay tuned for the daily updates and log into the website to see holdings and additional performance data.

image

This Week's Strategy Lesson: Slicing Up the ETF Models (Part 1)

image

Expectations are very important, especially when you are trading a system. One of the main reasons we like to look at past performance or back-test a trading system is so that we can develop realistic expectations for how the system works and the types of trades we will get into.

Financial markets have a fair degree of randomness. While we can expect that markets will react to certain news events in somewhat predictable ways and that the long-term value of stocks generally follows their fundamental values, over any given period of time, it’s not possible to know with certainty how things will play out. Deviations from “normal” happen and can persist.

That is why in trading we deal a lot with probabilities and expected value. We have covered expected value before but it might be helpful to go over that concept again. On an intuitive level, expected value is the average value or outcome from a high number of repetitions or outcomes. If you rolled a dice a thousand times, you might get a lot of 1’s and 6’s, but the average of all those rolls should be around 3.5.

For the ETF Complete, the expected value of any trade you put on is about 3.22%. While we have never actually closed a trade out in the ETF Complete model for precisely 3.22%, that number represents the average of all the trades over the last 12+ years of the model, including the ones that have reached multiple targets or were stopped out.

We can drill further down and come up with a probability distribution that can tell us a lot more information than just the 3.22%. We know that the range of outcomes runs from about -15% when we hit the initial stop (-17.6% was the worst trade outcome from a blown stop) to a cap of about +62% when we reach all of our targets and scale out at each. Taking all the trades, we can sort them and do a probability-weighted calculation to come up with 10 different buckets (or deciles) of equally-likely trade outcomes ranked by those outcomes.

image

The chart above takes all the ETF complete trades and breaks them down into 10 equally-likely buckets. This means that for any given trade you put on, like the recent new trade in Treasuries or gold miners, you have about a 10% chance of roughly getting one of the outcomes above.

There are few things we can observe from the data above that adds to our understanding over and above the simple 3.22% average outcome. With the first five deciles showing negative outcomes and the second five showing positive outcomes, we could say that each trade we put on has roughly equal odds of ending positive or negative.

However, the positive and negative outcomes, while equally-likely, have very different end values. The distribution of trades is skewed positive with the best average bracket gain being about three times higher than the worst average bracket loss.

We also have a middle section of the chart above with a lot of marginal outcomes that cancel each other out. While we would love for every trade to be a winner and post massive gains, unfortunately, we don’t know which trade signal will be that next big trade. The best we can is that a particular number of them will do great, many will effectively breakeven, and a few will do poorly.

Understanding these probabilities can keep our expectations in line with the model’s history and help us to stick with the rules even if we have a series of lackluster trades in a row or a period of underperformance because we know that in the long run, taking all the trades offers a strong expected value.

Next week we are going to expand our analysis and look at some other stats and how the trade performance skews between the models.