Moving average is notoriously unreliable as a guide to making investment decision. Many signals generated by a moving average are false breaks that are reversed immediately, resulting in loses and causing much frustration. In this paper, I propose an unsupervised adaptive algorithm based on the concept of moving average that is provably better than any moving average system. Instead of a fixed averaging signal, Adma is a parametric model that learns from the stochastic process and follows the underlying trend.
Learning is formulated as a constraint optimization problem that minimizes the strategy's P/L. Since the objective function is non-convex and non-differentiable, the gradient does not exist and thus much of the current optimization techniques would fail. To tackle this unconventional and challenging problem, I borrow techniques from statistical physics and microbiology, using coordinated and evolving search agents to seek out local optimum.
Applied to the benchmark S&P 500 index, Adma outperforms all the popular moving average systems not just in terms of profitability; it shows significantly better risk metrics such as volatility, maximum drawdown, and frequency of trades.
On an out-of-sample basis, Adma with a L0-regularizer shows consistent profitable performance well into the third year after the training set ends.
========================================================
The information in this presentation is provided for information purposes only. It is not intended to be and does not constitute financial advice or any other advice, is general in nature and not specific to you. Before using this information to make an investment decision, you should seek the advice of a qualified and registered securities professional and undertake your own due diligence. None of the information here is intended as investment advice, as an offer or solicitation of an offer to buy or sell, or as a recommendation, endorsement, or sponsorship of any security, Company, or fund. I am not responsible for any investment decision made by you. You are responsible for your own investment research and investment decisions.
The information in this presentation is based on financial models, and trading signals are generated mathematically. All of the signals, timing systems, and forecasts are the result of backtesting, and are therefore merely hypothetical. Trading signals or forecasts used to produce our results were derived from equations which were developed through hypothetical reasoning based on a variety of factors. Theoretical buy and sell methods were tested against the past to prove the profitability of those methods in the past. Performance generated through back testing has many and possibly serious limitations. I do not claim that the historical performance of the timing systems, signals or forecasts will be indicative of future results. There will be substantial and possibly extreme differences between historical performance and future performance. Past performance is no guarantee of future performance. There is no guarantee that out-of-sample performance will match that of prior in-sample performance. I provide absolutely no guarantee that the trading signals, forecasts, opinions or analyses presented here are consistent, logical or free from hindsight or other bias or that the data used to generate signals in the backtests was available to investors on the dates for which theoretical signals were generated.