简介人工智能在基金界的应用 (一)
原创 2017-10-24 袁玉玮 WaterWisdoms
要点:
◦人工智能ETF,如果真是收割机,你付0.75%的管理费能买到吗?
◦LHC全球对冲基金Top20,8家人工宏观,5加量化,5家事件驱动。14家不涉及量化
◦BlackRock的人工智能/量化部门SAE刷新了历史亏损
◦各大银行用算法交易取代的是报单员(execution trader),不是自营交易员(prop trader)
◦智能投顾取代的是理财销售员,不是基金经理。智能投顾和ETF近年相辅相成,同步泡沫化,我们预测它们未来会同时受到打击
先声明一下态度:我不排斥机器;我只排斥认为机器排斥人的人。
1
人工智能ETF?!
最近一只人工智能ETF在美国上市交易,AI Powered Equity ETF(AIEQ)。媒体上轩然大波,高呼人类基金经理/分析员终结者来了。
AI Powered Equity ETF由EquBot管理,后者是IBM旗下的IBM Global Entrepreneur的投资分部,专门负责在投资分析应用人工智能技术。跟据其官网公开信息,EquBot基于IBM的人工智能平台Watson, 分析经济和新闻数据,对美国上市的股票和REITs进行基本面分析,构建一个含30-70只股票的组合。这个组合每天对公众公开。
由于AIEQ上市时间太短,我们对业绩无法置评。我只有2个问题:
◦如果AIEQ的真的是收割机,你相信它会作为一个只收取每年0.75%的管理费的ETF卖给公众吗?
◦如果AIEQ的真的是收割机,投研团队会每天给你公开持仓信息吗?
2
LHC全球对冲基金TOP20
2016年Rothschild公布的全球史上最赚钱的20家基金,被国内媒体以讹传讹,误传他们全部是量化基金。
事实上,其中
◦只有5家或5家半(#2)以量化著称:BW, DE Shaw, Citadel, Millennium, Two Sigma。SAC Capital虽然有系统化交易部门,但不是主策略(Steven Cohen以类似alpha capture的投顾信号池为主,搭配类似MOM的多策略),所以最多只能算半家。
◦#1 多的反而是宏观策略:其中8家人工宏观:Soros, Apaloosa, Lone Pine, Viking, Moore Capital, Brevan Howard, Caxton, Tudor;2家系统宏观(与量化交叉):BW, Two SIgma。
◦#3 多的是5家事件驱动:Och Ziff, Elliott, Farallon, Paulson, King Street,大多集中于Distressed策略,固定收益,房地产,能源基金,PE,与量化可以说绝缘。
◦#4 最少的是价值投资的Baupost,只有1家。不过由于8家人工宏观(尤其3家选股型宏观)和5家事件驱动都必须使用基本面分析,我们不能说价值投资是弱者。
◦2家多策略(类似MOM):SAC,Millennium
◦2家属于Julian Robertson的老虎基金家族(Tiger Cubs): Lone Pine, Viking
◦9家的绝对收益超过了目前的自身管理规模:Soros, Citadel, Apaloosa, SAC, Paulson, Moore, Brevan Howard, Caxton, Tudor。这其中只有1家主策略是量化,当然这与量化起步较晚有关。
◦有2家开始培育人工智能/量化团队:SAC, Tudor。对Tudor来说,量化还没带来任何收益。
◦George Soros并没有投资量化,新近投资量化的是他的儿子Robert Soros。后者已经从SFM卸任,新成立自己的Soros Capital。
◦需要指出的是,Two Sigma的合伙人之一David Sigel曾经是Tudor的CTO。我想这也是Two Sigma把宏观作为旗舰产品的渊源之一。
请见附表:LHC2016全球对冲基金Top20
来源:LHC, WaterWisdom
3
BlackRock的人工智能/量化部门刷新亏损
媒体盛传全球最大的资产管理公司BlackRock(黑岩)进军人工智能/量化;但媒体只告诉了你一半事实,另一半是这个人工智能/量化部门SAE刷新了历史亏损。
Bloomberg报道,数据显示2016年BlackRock的2/3-4/5的量化策略跑输大盘。
BlackRock’s Robot Stock-Pickers Post Record Losses
by Sabrina Willmer, Bloomberg 2017 M01 9 18:00 GMT+8
Like so many fund titans these days, Laurence D. Fink is betting on machines to turn around BlackRock Inc.’s beleaguered stock-picking business. Trouble is, they just might have made things worse.
BlackRock’s main quantitative hedge-fund strategies – which use computer models to sort through vast amounts of data to pick out patterns – were on track for losses in 2016, according to a monthly client update sent out in late December. Of the five included, four were set for their worst returns on record, data through November showed. A separate investor presentation with a broader quant lineup showed that almost two-thirds underperformed.
A separate investor presentation showed 12 of 19 quant strategies trailed their benchmarks in the 12 months ended November, before management fees were deducted. Five of the seven that outperformed were part of their regional lineup. BlackRock declined to say the total number of quant strategies it runs.
At least three of the quant strategies used by BlackRock’s global hedge fund platform have suffered losses greater than 10 percent in the year through November, according to the client update, a copy of which was seen by Bloomberg. That compares with an average return of 3.6 percent for quant funds, Hedge Fund Research Inc.’s directional quant index shows.
To make matters worse, SAE has lost some of its top talent. The departures included Bill MacCartney, a former Google scientist that BlackRock hired in 2015 to help build out machine-learning, and Ryan LaFond, a head researcher and one of the brains behind the firm’s socially responsible funds.