Skip to content Skip to sidebar Skip to footer

Source: Le temps as of 30-03-2020

Raphael Steigmeier, managing director at VT Wealth Management,  reports  that 0.5% of shares  contribute three quarters of the performance. Artificial intelligence outperformed by 50% during  the  coronavirus crash

Raphael Steigmeier:”We  know  that unexpected shocks  remain  possible, whether it  is  an OPEC  decision  or  a  pandemic. Stress  scenarios,   like  the last  few  days,  show  that  our  model  is solid.”  

Raphael Steigmeier, managing director  at VT Wealth Management,  reports  that 0.5% of shares  contribute  three quarters of the performance. Artificial intelligence       outperformed  by 50%  during  the  coronavirus crash.

Raphael Steigmeier   has scheduled his first game  at the age of 12. Today,in  his  role as  manager  and architect of artificial intelligence  at VT Wealth Management  in Zurich, he  combines the traditional management  of a  private  bank  with  new technologies.

VT Wealth Management is  a wealth management company, with more than one billion francs under management. It was  created  in  2008 by Thomas  Fedier,  now Chairman  of the  Board  of Directors,and since  2013  it  has been  run  by the  latter’s son, Sacha  Fedier..

How have you  weathered  the  2020  crash so far? The  current situation and the  economic consequences  are  worrying..  But  we remained  rather  relaxed in  the face of falling   titres markets,  under  the  resilience  of  our stock selection. .

VT Wealth Management’s  strategy combines  traditional  wealth management  with  the use of  new  technologies. Customer management is  individualized and  personalized..  The use of  artificial intelligence  and  algorithms is subtle and non-bureaucratic.  That’s  why  we  fortunately don’t have stocks that we wouldn’t  really like to hold  in our  portfolios. Our  securities  have  become  more affordable because of the  decline  in  the market,without them becoming  really  cheap. If they  had  become,  it  would mean that they do not have  the  quality  they are looking for..

While the global  equity index fell  by  32%,  our investment fund gave up only 26% and  our  long/short  strategy even  gained  18%,  achieving    an  out performance  of  almost  50% compared to the n’a index..

Is this strong  out performance   the result of chance? No, we can rule it out. Our approach,and the    resulting  model, has  been extensively  tested  and  designed generically.  This  means  that he  has  been  trained  to recognize   general trends, but  that he  is not troubled  by  low  nuances. In  short,  our system  is  not  designed  to  recognize  a poodle and distinguish  it  from a  German shepherd. .  But  he  is able to  see  that a poodle and a  German shepherd  are  both  dogs..  From  the moment it was designed, we gave  up optimizing it  down  to  the  smallest detail..  I don’t know why. Critics of  our type of  model  accuse  it of an  excess of conditions to  be fulfilled  so  that it  fits exactly with the  historical behavior  of the markets. The grievance  is  justified  to the extent  that    many  systems  completely lose  their  relevance as soon as  a new event,  a black swan,   occurs. We  preferred  to build  it in  such a way that  it  behaves well in  all  situations, including  unforeseen    and  unprecedented events. .

Is it  a  system  that can  improve without human intervention? ?  Yes,  yes. It incorporates  a lot of artificial intelligence. .

What do you mean by  that??   For example,  humans  can easily  recognize  the  difference between a  dog and a cat. But  if  I  show these  two  animals to my computer,  he  won’t make the distinction and  he  won’t be able  to  identify them. By making  it available 100 million  images  of  dogs  and cats, we  can train the computer  so  that it  understands  the difference. We  have  done the same with equities  so  that  our  model  distinguishes between good and  bad  securities..

For decades,  hedge funds have been trying to build such models,  but  accidents follow one another,  like  LTCM  in  1998. Aren’t  black swans  inherently  unpredictable??  Absolutely.  Events  such  as the coronavirus  outbreak were  not  predictable..  The key  is  to  understand how    the   model  is  built..  If it  is generic in   nature,   it will  behave  better in the event of an unexpected  event  than a  specific  model such as the  LTCM model. The latter,  based  on  the assumption of a convergence of  interest rates  in  Europe, was  not  bad  in  itself..  But  the  profit margin was  so  modest that the fund  needed   huge  debt  to  get a reasonable profit.  It only takes   a little  for  such a  system  to be destabilized..  That is    what   happened.  What’s  your  approach?  From  a  particularly large universe of  data,  it  is enough  to identify  a small  number  of actions  that are,with  a  high  probability,  better  than the  others..  Bessenbilder  and Al,  in  August  2019,  presented research work over the  period from 1990 to 2018 which  proved  that  only  40.5% of the 61,100  securities  analysed  offered  a  return higher than government bonds. .  So  0.5% of the total companies  are  responsible for three quarters of the performance. The   Swiss  equity investor can  have  a passive strategy,  with an ETF,  or  active and  select for  example  five  securities  without knowing if they  are  the  best. Our  model  aims to  recognize  these five”best” values.

We know that unexpected shocks remain possible, whether it  is  an OPEC  decision  or a  pandemic. Stress  scenarios,   like  the last few  days,  show  that  our  model  is  solid..

How are  your  models built?  Our  models are based  on traditional risk factors  (criteria based   on  dividends, momentum,  size, value,  quality,  growth). We  add artificial intelligence  so  that the computer  discovers for  itself  the  best possible combination  of titles.  A  company that pays a  high  dividend  may not be as  attractive  as  another company that  offers  a  lower  return for the  same level of  profit,  but  makes more investments  to increase  its  long-term growth.  Classical theory   would lead me  to  favor  the company with a  high  dividend,  but  the computer  could  choose  the  second if it considers  that it  offers  the  best opportunities.

In addition, our  approach  learns  to distinguish  between industries. We carried  out monthly   analyses  for fifteen  years for more than 60  data (profitmargin,   taxes, profit growth, etc.) within 5000  companies. We collected an   estimated  70  million data  to  deduce  a  company’s    likely  future behaviour  compared to  that of other companies..  

Do other  managers  work  in the  same  way? To my  knowledge,  very  little..  Some  hedge funds used  a  similar  approach in the 1980s, such as  Rebellion Research,  but in Switzerland the  subject is just  beginning  to emerge..  Having  said that,  I stress that we are  not a hedge fund,  but   traditional managers  who  use   new  technologies.

You   are based  on  data  on  fifteen  years of the life of  each  company..  Don’t you omit  much of  its culture and  history? We don’t integrate   older  data, but  the computer learns  from  its   errors. The  current data are the most  important. In the  1990s, for example, banks and insurance companies traded at a price  equivalent  to three  times their  intrinsic value. .  That  has  change dat the latest  since  the financial crisis.   These securities  are  now trading at  their  intrinsic value.  The  system  notices  these  changes.

Your model  is based on the  Fama Nobel Prize  theory  and   market efficiency,   but behavioural  finance has en  challenged  this  work. How  do you take that  into account??  Yes, I do.  That’s why,  pour if someone tells me they can predict  everything  100%,  I’ll tell them  with a 100%  probability  that they’re  wrong. Behavioural finance  shows  that  investors  can  over-react or  misjudge  events. If  a  company makes  a  profit,  the stock  may  go up. But     this information may  not be fully  integrated  from  day one.    

The evolution of a  share price contains information  on  what  we do not know, for example  the  trends of the  oil market. .  Since last    fall,  we  have been  negative about oil  stocks  because of  the signals  sent  by  our  system..

VT Wealth Management, with 21 employees,does not rely  on  a research team  the  size of a  large  bank..  But  if  I  receive  the  recommendations  that  accompany  the studies of   these  institutions,I  remain  quite skeptical,because I do not know  how  the analyst came to  his  conclusion. The tone  of the employee  sometimes  differs  from  his final recommendation.  I sit    because of the bank’s business relationship?  I’m  wary  of  this type of  filter..

Is   your  use  of the  term n’est “artificial intelligence” not a marketing exercise? No, it’s not. It  is only  with the use  of  artificial intelligence, big data and conventional algorithms    that we  can  simultaneously analyze thousands  of actions. We  then manage to prioritize them,  like   bank analysts,   by “buy,””conserve” and”sell.” Unlike the bank analyst,  however, we  have no  potential conflict  of interest.  Our computer  has  no  emotion, so  our  ratings are  consistent..  These   allow  en     us, as asset managers,   to put  into  practice active management and  to have  a clear  opinion on equities, which  sets us apart from  our  competitors..  

What are  the most advantageous stocks today? In the  United States,we  avoid oil companies like  Schlumberger, Chevron, Occidental Petroleum,  big  banks  like JP Morgan, Morgan Stanley. On the other hand,we  like United Health, Microsoft, Visa, Mastercard,  so  fintech rather than  investment banks. .  In  Switzerland, Nestlé, Schindler, Bucher,  Sonova,Logitech,  Conzetta  or  Straumann  are  for  purchase..

Hedge funds suffer from   their high  fees  and poor performance.  Do  they still have  a future? We  are  not a hedge fund,  but this branch of finance has been around  for  almost a century and  has been going through  several crises. The goal of  hedge funds  is  to contribute to the  stability of performance. Their  interest lies  not  in  their  possible annual  outperformance  but  in  their  ability not to  suffer  sharp  declines..  Suppose   I only have    two shares,  one bullish  position  (long) and  one for sale (short). I can’t win  if both go up the same..  Better    like today’s  highly  volatile markets. The performance-related fee structure is only  practiced   if the objective  is  achieved. Our  long/short product outperforms  by 50%. No one  is obliged  to buy it..

How do you integrate the GSS? I  consider environmental,  social and governance (ESG)  criteria  from  a performance perspective.  There  is  a link between a company’s   data count   and  its  performance. In emerging markets,the contribution of the ESG  approach comes mainly from the G (governance)and  less  from  environmental  and  social criteria  because  it    avoids scandals. Our  selection  does  not  include a  durability ▅ filter,   but  we  can  take this  into  account  at the request of a customer. .

“It is only  with  the use  of   artificial intelligence, big data and conventional algorithms    that we  can  simultaneously  analyze thousands  of actions”»

Show CommentsClose Comments

Leave a comment

News ORS © 2020. All Rights Reserved.