Loading...

Information about Modeling

What is a model?

Backward Testing

A simple example of a market model is comparing today's price to the historical average of price over, for example, the 50 past days. The rule would be to BUY when today's price is above the past average, and SELL if below. This rule could be easily backtested and other rules could be applied to produce win/loss ratio and profit/loss delta. Other rules could be applied and tested, but not so many that the rules only work for the test data. Known as curve-fitting, too many of today’s rules are created to fit just one set of past data, which will not work for new data. In simple terms, the rules were so stringent that they could only work once.

Forward Testing

The next method for a model is to forward test, which measures the results against actual forecasts. The goal is to gather real-time trades, capture their results, and compare against the benchmark backtest. Since 2013, the results were consistent with the historical testing, falling within test parameters including win/loss ratio, average drawdown, average ROR, as measured in 3 mo, 6 mo, 9 mo, and 12 mo periods. The forward test is key in determining if the backtest was uniquely unrepeatable or whether the methods applied will continue to work.

Publishing Trades

By 2012, Michael’s confidence in his own MarketModel led him to establish a futures account and commit to trading the model signals, testing different position sizes and sharing trades on Twitter. Michael's trading diary moved to private subscription service in 2013-2014 via protected Twitter, where the nightly model signals were posted, along with trading results. More recently, daily and weekly model values are delivered via membership site at http://www.macrospx.com

The model uses SP500 futures instead of the SPX cash because of the use of leverage provided by futures. Use of leverage is typically associated with high risk, however in our case, we are simply choosing to not to tie up $1M is cash buying and selling shares of the SPX. Leverage allows for the scaling from underweight SP500 to up to 1x leveraged (100% long or short).

The model's use of futures allows for scaling into very large account balances. For example, a $1B account could invest alongside the model within the SP futures market easily within the average volume and open interest of today's market.  The investment decisions are almost entirely systematic, based on the model's signals.

No comments yet, be the first one who will leave the comment.