The ABCs of creating a mean reversion strategy

[LeeDanello]

This is a good primer

The ABCs of creating a mean reversion strategy – Part 1

[RobGiles]

Thanks Maurice, I found this very useful.

[RobGiles]

I’m in the process of building a swing trading system. I’ve looked at the Cesar Alverez article posted by Maurice, which has been a great source of ideas, but I’m finding many of them are giving ho-hum results. I’m trading at 50% leverage max 20 positions as I don’t want another 4 x LVR system. To date I’m looking at the Russell 2000 and the following approaches:

Using RSI oversold as an entry / overbought as an exit with a secondary stale close exit
MAR Ratio 0.91
CAR 17.8% (would like to see this at 20%+)
Max DD -19.2% (trying to get this < 15%)
have played around with changing liquidity filters, ATR stretch entry parameters, RSI entry and exit parameters, stock and Index MA filter values, number of stale exit bars

Using RSI oversold as the entry setup, short term MA as the exit, with a secondary stale close exit
MA(3) gives me MAR of 0.98
CAR 17.7%
Max System DD -17.6%

Have also looked at “x” days down as an entry, which is also not that encouraging, but will persist in playing around to see if I can improve it.

FWIW (for what its worth!) I have the following objectives:
CAR >20%
MaxDD <=15% aprrox
Max LVR 50%

Any tips, pointers would be appreciated.

[ScottMcNab]

in no particular order: ROC…(.eg ROC(C,3)<-2)....also price action on set up day (eg close in lower 20% of bar)....adx filter has been posted about previously and useful

[Anonymous]

Hi Rob, I also trade my MRV on R2000. Something to really take into account when back testing would be the unreliability of getting a fill on a lot of those stocks. In real trading I miss a decent number of the big winners you’ll see in a backtest.

The issue quite often is there will be a LOW of the day that has one trade, or is some other special type trade that you won’t get a fill of course – and they always be winners in a back test!

Also not sure if you’re holding overnight or MOC, but if holding overnight be prepared for some pretty decent gaps down when you’re lurking in the R2000 haha.

That said the volatility of it still can produce results.

[RobGiles]

Kerry O’Keefe wrote:

Hi Rob, I also trade my MRV on R2000. Something to really take into account when back testing would be the unreliability of getting a fill on a lot of those stocks. In real trading I miss a decent number of the big winners you’ll see in a backtest.

The issue quite often is there will be a LOW of the day that has one trade, or is some other special type trade that you won’t get a fill of course – and they always be winners in a back test!

Also not sure if you’re holding overnight or MOC, but if holding overnight be prepared for some pretty decent gaps down when you’re lurking in the R2000 haha.

That said the volatility of it still can produce results.

Cheers Kerry
Yep, I’m looking to hold positions for up to 7 days, so I would be getting nasty surprises on large down days. Have you come to any conclusions about how much lower real time results are likely to be relative to back-tested metrics as a result of the fill issues you describe above (thanks for the heads up on this, I wouldn’t have factored that in)?

Have you explored other indexes to trade the same / similar systems on (like NASDAQ for e.g.)?

[RobGiles]

my objectives for building my MR / swing Trading system have been:
CAR in high 20%+
Maxx DD better than -20%
CAR/MaxDD Ratio > 1.5

but after reading the article by Cesar Alvarez “The ABCs on creating a mean reversion strategy” he’s made me think maybe I’m being too precious about the maxDD and The CAR Ratio:

Max DD’s 20% – 35% “I know a lot of people want CAR to be 2 times more than MDD. When you do find those strategies understand that everyone else is looking for the same thing. By not caring about having MDD greater than CAR, I am looking for strategies that others are avoiding. Which is good.:”

Has anyone else come to the same conclusion?

[SaidBitar]

It is tricky thing honestly i think the answer depends on many factors
1- Are you planning to use margin
2- Is it the main strategy or complementary strategy for a longer term strategy
3- Risk tolerance
4- Account size
5- Trading universe
personally i will trade one strategy that is more aligned with your objectives because for me it will be fulfilling its role of smoothing the equity curve of longer term strategies

[ScottMcNab]

I find that after minimsing selection bias, from 2010 onwards my cagr is in low to mid 20’s (I use 2x….40×5 )…I am happy if maxDD does not exceed cagr

I think the idea of looking for different systems is a very good one Rob…particularly if it is one of multiple systems and sufficiently different for the months of high return and maxDD to be uncorrelated and so maxDD not as big a concern

[RobGiles]

Scott McNab wrote:

I find that after minimsing selection bias, from 2010 onwards my cagr is in low to mid 20’s (I use 2x….40×5 )…I am happy if maxDD does not exceed cagr

I think the idea of looking for different systems is a very good one Rob…particularly if it is one of multiple systems and sufficiently different for the months of high return and maxDD to be uncorrelated and so maxDD not as big a concern

Thanks Scott,
The more I read about MR systems, the more I’d be happy with return stats like yours (i.e. CAR in the 20’s and MxDD about the same). I’m jumping the gun a little here, but once I do build a few more MR systems, I’d love to know how to easily measure the correlation between them and get to a place where I can actively allocate capital weightings on the strategies that I have in my portfolio based on current market conds.

[RobGiles]

Said Bitar wrote:

It is tricky thing honestly i think the answer depends on many factors
1- Are you planning to use margin
2- Is it the main strategy or complementary strategy for a longer term strategy
3- Risk tolerance
4- Account size
5- Trading universe
personally i will trade one strategy that is more aligned with your objectives because for me it will be fulfilling its role of smoothing the equity curve of longer term strategies

Thanks Said. In response to your points that you make:
1) yes using 50% LVR or 2 times leverage
2) complimentary to a couple of longer term systems. I ultimately would like 3 or 4 different MR systems that make up about 25% of my invest-able capital
3) Risk tolerance – max 30% DD or thereabouts with a draw-down period less than 6 months (not sure what is realistic here)
4) account size would mean i would have to be keeping out of the Russell 2000 due to liquidity issues I feel
5) Trading universe – been guided by Nick away from Russell 2000 into Russell 1000 or any other US mid to large cap index. Hearing liquidity and commission drag issues within this forum re the Australian market has put me off trading the ASX (but happy to be proven wrong).

[RobGiles]

I’ve recently listened to the BST podcast 062 with PJ Sutherland about MR trading…very interesting. One thing he recommended looking at was ROC normalised over a longer time frame (last few days vs past 200 for e.g.). has anyone looked at this approach?

[Author]

Posts