According to Reuters, GS is considering to exit quantitative hedge fund strategies altogether after suffering persistent losses and declining AUM. At the same time, similar funds (Simons' Renaissance & IKOS for example) are again profitable (risk metrics not available for either naturally). Assuming that GS has unconstrained resources (infrastructure + smart people that constantly devise, test, implement & improve), the situation seems paradoxical. Is the personal element that important and so rare?Thanks in advance for your answers.

I don't know much about GSAM. Here is a general background article from January: Bbg article.

So you're saying it doesn't make sense that Global Alpha is losing money, but other quant funds (which is a huge category covering tons of different strategies) are making money?Highbridge Capital Long/Short Equities is down 11% YTD and Citadel Global Equities Fund is up 14% YTD. Would you say that doesn't make sense, because they're both fundamental long/short funds? There are winners and losers and ultimately there's only so much alpha out there. Just because two funds are in the same strategy class doesn't mean they're doing exactly the same thing. In many ways they are competing for the same fixed size pie. Global Alpha was once the best stat arb trading shop in the world. In the past 5 years though the landscape has radically shifted in quant trading. The direction GSAM went in was not the right one for where the market was going. Make no mistake the people who work there are absolutely brilliant, some of the brightest I've ever met. However, like anyone could, they made serious strategic missteps.

QuoteOriginally posted by: xlnap after suffering persistent losses and declining AUM.Just to put some numbers on this:QuoteThe fund, which managed $11 billion of assets in 2007, had less than $1.7 billion at the end of June, according to a person familiar with the matter who spoke on condition of anonymity because the numbers aren?t public. Quote... has been shrinking Global Alpha since 2007 when it lost 40 percent because of bad bets on currencies, equities and bonds worldwide.

Here was a headline that caught my eye a few months ago: Goldman Had Just One Losing Day Last QuarterDoes anyone understand the accounting of just what constitutes trading revenue for a firm like Goldman?(All the winning trades from the Global Alpha fund? ) Seriously, this "trading revenue" category has always seemed mysterious to me. If it is explainedin a filing or accounting guideline somewhere, somebody please post a link.

QuoteOriginally posted by: undergrad86 but other quant funds (which is a huge category covering tons of different strategies)True, but is it not puzzling that they apparently did not match strategies with market conditions for a number of years?QuoteOriginally posted by: undergrad86 Highbridge Capital Long/Short Equities is down 11% YTD and Citadel Global Equities Fund is up 14% YTD. Again, YTD it is expected to have a range within a category, Global A however consistently underperformed for sufficient years to get out of business. Not doing well due to circumstance is one thing, not adjusting/reacting at all is another.

QuoteOriginally posted by: AlanHere was a headline that caught my eye a few months ago: Goldman Had Just One Losing Day Last QuarterDoes anyone understand the accounting of just what constitutes trading revenue for a firm like Goldman?(All the winning trades from the Global Alpha fund? ) Seriously, this "trading revenue" category has always seemed mysterious to me. If it is explainedin a filing or accounting guideline somewhere, somebody please post a link.Take a look at section 2.2 of this paper: http://www.nber.org/chapters/c9606.pdfIt is primarily made up of daily mark-to-market changes of trading book positions along with associated fees and net interest income.

QuoteOriginally posted by: bearishQuoteOriginally posted by: AlanHere was a headline that caught my eye a few months ago: Goldman Had Just One Losing Day Last QuarterDoes anyone understand the accounting of just what constitutes trading revenue for a firm like Goldman?(All the winning trades from the Global Alpha fund? ) Seriously, this "trading revenue" category has always seemed mysterious to me. If it is explainedin a filing or accounting guideline somewhere, somebody please post a link.Take a look at section 2.2 of this paper: http://www.nber.org/chapters/c9606.pdfIt is primarily made up of daily mark-to-market changes of trading book positions along with associated fees and net interest income.Good link -- thanks. Sounds like a better name for it would be brokerage-and-trading revenue, which would explain the small loss probs.

QuoteOriginally posted by: undergrad86Global Alpha was once the best stat arb trading shop in the world. In the past 5 years though the landscape has radically shifted in quant trading. The direction GSAM went in was not the right one for where the market was going.Interesting!Could you please elaborate on what you think they got wrong? Thanks!

Well my interaction with Global Alpha has only been tangental, so take what I say with a grain of salt. But...Global Alpha's stat arb models were very macro-ish focused. A lot of their alpha traditionally came from country allocation in addition to security selection. Even on the security selection side they had a very fundamental focus, e.g. balance sheet, earnings, industry to generate their factor models. Starting in 2007 and especially in 2008 these types of models broke down. The stat arb that survived had a much more "numerical" focus. The main difference between the two is that GSAM style stat-arb is more longer term oriented, after all macro and fundamental data gets updated on the order of months. Numerical alpha models can more easily be adjusted at any horizon as long as you have market data at that resolution. There are a couple of reasons for why their style did so poorly. The first would probably be the high-frequency revolution. A lot of the stat-arb alpha simply moved to shorter horizons. Global Alpha's models just didn't work at this level. The second would be the rapidly changing relationships during the period. Market factors rapidly shifted in short periods. In 2008 the second largest factor after market beta was pretty much financials vs non-financials, in summer 2010 and 2011 it quickly became euro-zone vs non euro-zone, at other periods momentum became a huge factor, etc. Correlation rapidly went to 1, then came down just as fast. Global Alpha has always done poorly during regime changes, and between 2008 and today there have been a lot. With numerical alpha you can more quickly invalidate your models, detect regime changes and refit, because you have more data points. The question would be why then didn't they see the writing on the wall and adapt. I think a number of reasons. First would be because they had a large asset base. The problem with moving to shorter term models is that it lowers your trading capacity. At one point they were $10bn, and I think their business goal was always to get back to that asset level. The fundamental/macro stuff takes a lot more researchers, so they couldn't support their headcount with a smaller asset base. One could argue they should have shrank their operations and asset base and possibly converted to prop trading where they could net all their pnl instead of just 2/20. But hindsight is 20/20.The second thing is that Global Alpha was a very academic environment. They had a long history of publishing their results (which arguably was the beginning of their downfall as people replicated them and took their alpha). You don't see any published papers coming out of RenTech. The numerical alpha stuff is simply much less interesting from an academic viewpoint, and most people there had backgrounds more in fitting macroeconomic relationships, rather than crunching huge datasets in a HPC. They probably figured their comparative advantage laid in that style of stat arb.

very interesting. Thanks for that. As an ex-strategist i'd concur with your observations on the death of the global macro trade - once it was possible to gain an edge through fundamental analysis, but that edge has diminished (in addition to fast trading, i would add as related reasons; data dissemination and increased inter-asset class, and geographic, correlations). How many truly diversified global macro trades can managers concurrently implement these days? Despite the prominence of global macro themes in driving asset class returns in recent years, the reality is that it's become a crowded trade with high stakes - long risk or short risk.

Thank you. Very insightful, especially your suggestions on why they didn't adapt (inflexible strategy vs regime shifts). Do you believe that simultaneously using uncorrelated models might do the trick for a kind of auto-adaptation?