My test for SPY.

This would be because SSO got nailed harder in the market crash, and had a slower recovery. The leveraged funds usually get hit harder because they have to sell into losses to maintain constant leverage. Also by using order_target_percent, you are actually having to sell into losses as well, since the drawdown is so high for SSO, it never really fully recovers.

I'm sure there is more to it because SSO uses derivatives to leverage, but the fact that the returns are different should be the underlying reason for the difference in the results.

Here's a plot of the cumulative returns for SSO and SPY since Jun '06.

This one is the cumulative returns of SSO Vs. 2*SPY

HI Sébastien,

Note that SSO isn't meant for long-term holdings, a discussion of many articles:
http://www.investopedia.com/articles/exchangetradedfunds/09/broken-leveraged-etfs.asp
http://www.investopedia.com/articles/exchangetradedfunds/07/leveraged-etf.asp
http://beta.fool.com/acardenal/2012/04/11/right-way-use-leveraged-etfs/3563/

(Then again, I have done this in past nicely, or SDS during bears. Just be aware of what you actually have.)

Basically, twice the holdings of SPY does not equate to 1 holding of SSO, since the percentage relationship is not over time, but is only maintained on a per day basis. As the market goes up and down, the long term trend you expected will not be intact, as noted in David's plots above; and, if you try other random returns you can see a similar effect.

As the motley fool article states: "The problem is not that these instruments fail to deliver their promise, the big issue is that double or triple the daily return of an index doesn't mean double or triple that return over a month or a year. In fact, the higher the leverage level and the longer the holding period, the most likely it is that the leveraged ETF will become a losing proposition."

HTH.

Note that if he is rebalancing his SPY portfolio daily, that should negate the problem of the monthly+ compounding.

2xSPY net yield = (2 x 1.81% div yield) - (2 x 0.11% MER) = 3.40%
SSO net yield = 0.22% div yield - (2 x 0.9% MER) = -1.58%

Net net yield diff = ~5% P.A.

So, there's that. But the majority of this is probably still from the volatility drag of the leveraged ETF rebalancing, which you are not adequately replicating in your backtest. http://www.math.nyu.edu/faculty/avellane/LeveragedETF20090515.pdf has some VERY good information on how to analyse these ETFs.

Ok thanks everyone for your answers. As Simon Thornington noticed, I am rebalancing my 2x SPY portfolio daily (the idea was to replicate what I had read on Investopedia), so each day, if the return of SPY is 1%, my portfolio will yield 2%, and if the return is -1%, it will yield -2%. I will definitely read the paper on Leveraged ETFs to get a better understanding.

I know leveraged ETFs are designed for short term. If these backtest demonstrate something, it is that you get better returns by using the margin (assuming interests are lower than 4-5% per year, but I know that in Interactive Brokers they are much less than that) than using a leveraged ETFs in the long term. I know though there are upsides as downsides in both.

Thanks again,