How realistic is it to achieve anything near a 2 MAR in real-time trend trading?
I’m relatively new to mechanical trading and Ive been looking at CTA results @ www.traderview.com and a 1 MAR seems exceptional. Even the likes of ecklhard, dunn, and J henry don’t achieve this.
I’d be interested to hear what some of the seasoned system traders have achieved real-time as opposed to what back testing has produced. What should a new trader expect long term? Is 0.5 – 1 MAR the best we can hope for?
(Perhaps this should be in the newbie section… Feel free to move it)
Realistic Returns?
-
Jake Carriker
- Site Admin
- Posts: 1493
- Joined: Fri Sep 12, 2003 10:32 am
- Location: Austin, Texas
Hi JD,
You have hit on the weakness of MAR as a performance metric. Any system or combination of systems is statistically gauranteed to have a worse MAR in the future than it does right now. This is the reason you see traders with long track records sometimes "unfairly" penalized by MAR. Imagine a CTA that has realized a 30% (after fee) CAGR for the last thirty years. If he had a 30% drawdown during October 1987 or some other unfortunate time, his MAR ratio is stuck at 1 or worse until he can raise his average annual returns by quite a bit without suffering another drawdown. UNLIKELY.
Therefore, I suggest you take MAR for what it is worth, a good measure of performance, but only if taken in context. The flipside of our previous example is the hot new manager this year that has a MAR of 10 or better.
The MAR of 10 is just as meaningless for real world evaluation purposes as the long time manager's MAR of 1. The moral of the story is, look at the equity curve. Figure out what your particular strategy's performance looks like under a wide array of circumstances. Determine if you can live with the worst case results, and if you are happy with the "average" results.
I view long term (20 year plus) MAR ratios of 3, 4 and more as evidence of possible curve fitting. However, if you find an exception to this rule, please let me know.
BTW, other performance measures such as a MAR with a shorter lookback period, or a MAR that substitutes average drawdown for max drawdown all exist. However, each metric has its own drawbacks. You are the best ultimate judge of what sits right with your own gut.
Jake
You have hit on the weakness of MAR as a performance metric. Any system or combination of systems is statistically gauranteed to have a worse MAR in the future than it does right now. This is the reason you see traders with long track records sometimes "unfairly" penalized by MAR. Imagine a CTA that has realized a 30% (after fee) CAGR for the last thirty years. If he had a 30% drawdown during October 1987 or some other unfortunate time, his MAR ratio is stuck at 1 or worse until he can raise his average annual returns by quite a bit without suffering another drawdown. UNLIKELY.
Therefore, I suggest you take MAR for what it is worth, a good measure of performance, but only if taken in context. The flipside of our previous example is the hot new manager this year that has a MAR of 10 or better.
The MAR of 10 is just as meaningless for real world evaluation purposes as the long time manager's MAR of 1. The moral of the story is, look at the equity curve. Figure out what your particular strategy's performance looks like under a wide array of circumstances. Determine if you can live with the worst case results, and if you are happy with the "average" results.
I view long term (20 year plus) MAR ratios of 3, 4 and more as evidence of possible curve fitting. However, if you find an exception to this rule, please let me know.
BTW, other performance measures such as a MAR with a shorter lookback period, or a MAR that substitutes average drawdown for max drawdown all exist. However, each metric has its own drawbacks. You are the best ultimate judge of what sits right with your own gut.
Jake