Back to Community
1500% returns, -52% drawdown. Thoughts?

So I'm experimenting with an algo and it's starting to look promising:

Back test

I'm getting 1500% returns from a 2002-present backtest. I made some mods and now it's 1300% but much lower beta (not at contest standards) and a little less drawdown.

I wanted to get people's thoughts on whether they'd run this with real money? What sort of risk calculations should I be looking at. I'm 29 y/o, would you put a % of your savings into something like this?

The algo definitely thrives in a bull market and I have some safeguards in place to help it survive a crash like 08, but in neutral markets, the s&p will slowly creep ahead.

5 responses

@Alexander

It's too much of a roller coaster ride for me. I'd put some 'fun money' in it (ie money that I could lose or easily add a 50% loss to). Generally my preference is to have drawdown below 18% (below 25% if it's during the 2008 crash) and volatility below 15%. That about mimics the volatility of the S&P500 so if you can do better than the 12% or so average annual return of the S&P500 you are doing good. An 'acceptable' investment would usually carry a risk adjusted return (ie the Sharpe ratio) greater than 1. Your .81 value would be unacceptable to most investors. I would expect something greater than 1.3 for a solid algo to start investing real money.

What is the typical and max leverage of your algo? Can you simply dial back the leverage and reduce the volatility?

Thanks for the insight into what an investor would look for. My leverage spikes to 3.5, averages below 1.

It's tempting to be enamored by curves that jettison to the upper right-hand corner, so it's healthy that you're skeptical of the results. To piggy-back off of Dan's questions, I'd take a look at:

  • What securities is the strategy trading? Are they achievable? Or are they typically hard to borrow or illiquid tickers that may be harder to trade with a broker? Try testing the strategy on the Q1500 or Q500, easy built-in universes to plug into your strategy. Keep in mind that short borrowing costs are not modeled in the backtester.
  • What transaction costs is your simulation modeling? How sensitive is your strategy to these costs at varying account sizes?
  • Is the universe static or dynamic? If it's static, is it possible it suffered from lookahead bias by selecting securities that had success over this period?
  • Is there a specific security, or sector, driving the results?

I'd also suggest to run a tearsheet to understand the underlying performance and risk attributions in the strategy: https://www.quantopian.com/posts/new-feature-comprehensive-backtest-analysis-pyfolio. Good luck!

Disclaimer

The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by Quantopian. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. No information contained herein should be regarded as a suggestion to engage in or refrain from any investment-related course of action as none of Quantopian nor any of its affiliates is undertaking to provide investment advice, act as an adviser to any plan or entity subject to the Employee Retirement Income Security Act of 1974, as amended, individual retirement account or individual retirement annuity, or give advice in a fiduciary capacity with respect to the materials presented herein. If you are an individual retirement or other investor, contact your financial advisor or other fiduciary unrelated to Quantopian about whether any given investment idea, strategy, product or service described herein may be appropriate for your circumstances. All investments involve risk, including loss of principal. Quantopian makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances.

I'm new to Quantopian so I didn't quite understand everything on the images you shared in your "backtest" link. It was also pretty hard for me to see.

First and foremost, I'd recommend asking yourself "why" the strategy works in the first place. Then ask yourself why 3 or 4 more times after that to gain an even deeper understanding. Once you've gone through that exercise, ask yourself why you'd expect the strategy to work in the future. For years, possibly decades, to come? Because that's what you're really after, right? Future returns. Not historical, simulated returns.

You see, there's a difference between data mining and building a winning strategy. It's easy to data mine above average historical results. But the real challenge -- and what you're really trying to solve for -- is to build a strategy with above average future returns.

And that's why I suggested asking yourself why you'd expect your strategy to work in the future. Because winning strategies typically capture risk premiums in the markets and they typically have an economic justification behind them. If you can't walk away with some common sense answers to your "5 whys" exercise, then you might want to think deeper about what your strategy is really doing.

I wrote an article on this if you'd like to read about it. The article was written for discretionary traders, but the model and the concepts apply to quantitative trading too. In trading (and in life), winning Strategies are born from Insight. And insight is the reason why the Strategy was developed in the first place. After reading the article and considering the mental model, I'd encourage you to consider, "What's the Insight behind your strategy?"

I hope this helps. I know it isn't a "green light thumbs up" type of response, but I do believe it'll provoke a line of thinking that'll be highly productive for you.

Have a great night.

https://alphaunlocked.com/?p=9

EDIT: One last thought. Your leverage spikes to 3.5? Meaning your exposure exceeds cash by a factor of 3.5 at some point? And it's trading securities? If all of that is true, I'd recommend asking yourself whether or not you can even implement a strategy like that in the real world.

My unexpert opinion: Watch out for leverage. Back testing, from what I understand allows for infinity leverage. Real world is limited to what the laws / regulations / brokers will allow. Piggybacking a little on what Jason said.