Back to Community
Q Fund Down 3% Since June

Paywall article WSJ:

I don't have a subscription so I can't see the whole thing. Do they reveal anything else of interest besides the wanting performance of the fund and the departure of the CIO?

3% loss over six months for a fund of low volatility algorithms? That's fairly significant. I bet a good bunch of these algos only hope to make 3-5% annually. So this is a significant loss. You'd think even if some algos were performing poorly, it would be compensated for by algos that perform well, and at least cancel out the losses.

So How is this even possible that the fund is performing so poorly?

My thought is that the selection process so emphasizes low volatility and market, sector, and dollar neutrality that probably according to those criteria the fund is a raging success. Probably a straight downward sloping line for six months.

Harder to control for alpha factor exposure? Perhaps too many of the algos depend on certain market conditions for their alpha, which the current extreme low volatility regime doesn't provide.

Another thought is that backtesting can't account for how other market participants will react to your algorithms' trades. This is perhaps a significant variable.

Another idea, just because an algorithm continues to perform out-of-sample, doesn't necessarily mean this is statistically significant. With thousands of algorithms you'd expect a bunch to continue to perform out of sample by chance.

Still a bit baffling nonetheless. The selection process is rigorous.

7 responses

Conversely, maybe this is expected behavior. Maybe QFund meanders during a raging bull market but makes a killing during a crash and ensuing high volatility.

More and more people are pointing out, that algorithm evaluation process is inadequate, but Q refuses to take that into account (as positive criticism). That can become a crucial problem in the future and can have a big impact on Quantopian.

Topic about contest algorithms evaluation:

The 3% drawdown is explained by leverage. 3x-6x if i'm not mistaken

3% down isn't bad. Bridgewater has a fund down about the same so far this year. These algos are also leveraged up so if you're making 5% annually it's more like 25-30% after being leveraged. It's far too early to judge Quantopian's performance. Many great hedge funds struggled for years at their start... like Bridgewater or Millennium before catching their stride. The WSJ was just looking for some clickbait.

Ok but it's a gigantic endeavor and they've performed some engineering miracles. I hear Quantopian is a lot faster than QuantConnect for example.

Rely on us author-quants more, we need a focused way to be heard, we're kind of fragmented despite noble intentions with this forum.
Don't wait for another loss to take some more drastic measures.
Step 1: Let a quant volunteer to set up a suggestions-and-upvoting page using the Q stylesheet.
Now that I think about it, why don't one of you members just go ahead and do that on any domain of your own, a single page to start where we can submit our suggestions (500 character limit) and upvote (sorry I'm too busy), or Q, offer a $500 reward for it.

I am ever more convinced that markets, being complex adaptive systems, are inherently unpredictable in the short term. In the long term there has been an upward bias whose continuation depends on continued global economic prosperity.

The big question for me is whether prediction is even possible in markets for the timescale on which Q operates or whether they would be better to "invent" or follow some sort of quantitative index following strategy.

Failing which to attempt to compete on the market anomaly front on the high frequency timescale.