Introducing TMI Data Science: Delivering Alpha With Custom NLP-Based Leading Indicators

Managing Director

Elan Gore

TMI Data Science

TMI Data Science utilizes Natural Language Processing to build custom leading indicators using unstructured data sourced from the global financial, trade & traditional media.  Our proprietary software, TMI Machine, is a text-mining platform designed with the buy-side & capital markets professional in mind, text-mining the global media to discern and quantify nuanced qualitative shifts in press coverage as they apply to macroeconomics, equity indexes/ETFs, commodities, currencies, fixed income & individual equities.  TMI provides custom leading indicators and systematic trading strategies, operating at the nexus of linguistics & statistics drawing on vast experience in investment banking and hedge fund settings in senior quantitative and fundamental roles.  

TMI Data Science was founded by Elan Gore, an award-winning Senior Global Basic Materials Equity Analyst & China Economist with 15 years of experience in North American hedge funds & institutional managers; and Alessandro Porrino, Ph.D., an Algorithmic Trading Analyst and Programmer with 10 years of front-office experience ranging from complex FX option pricing to auto-hedging strategies at Bank of America Merrill Lynch and Standard Chartered.  Mr. Gore holds an MBA in Finance from New York University Stern School of Business, and Mr. Porrino holds a Ph.D. in Systems Engineering from Brunel University London.

Cross Asset Strategy

<ul>
<li>The strength of US housing has been an unassailable consensus theme coming out of the March-April shock, aided by record-low rates and signs of deurbanization</li>
<li>We text-mine across a range of housing vectors to find greater nuance to the story than headlines imply</li>
<li>Extreme sentiment readings are seen in recent reporting on housing data releases (i.e. existing home sales, housing starts, NAHB); but our text-mined indicators of homebuyer intentions, deurbanization pressure and sales reported above asking price have turned softer</li>
<li>This is partially confirmed by Zillow app store data, which peaked on a Y/Y basis in July and on an absolute basis in August, recently turning roughly flat Y/Y; and by Redfin median square footage comps, which peaked in August</li>
<li>Similarly, our indicator of wood product shortages peaked in September and has stabilized since, in-line with forest products ETF relative performance and wood product pricing</li>
<li>We suspect US housing demand has been losing momentum, but remains at elevated absolute levels</li>
</ul><div class="content-highlight"><p>TMI Data Science utilizes Natural Language Processing to build custom leading indicators using unstructured data sourced from the global financial, trade &amp; traditional media. Our proprietary software text-mines the global media to discern and quantify nuanced qualitative shifts in press coverage as they apply to macroeconomics, equity indexes/ETFs, commodities, currencies, fixed income &amp; individual equities. </p></div>


The strength of US housing has been an unassailable consensus theme coming out of the March-April shock, aided by record-low rates and signs of deurbanization
We text-mine across a range of housing vectors to find greater nuance to the story than headlines imply
Extreme sentiment readings are seen in recent reporting on housing data releases (i.e. existing home sales, housing starts, NAHB); but our text-mined indicators of homebuyer intentions, deurbanization pressure and sales reported above asking price have turned softer
This is partially confirmed by Zillow app store data, which peaked on a Y/Y basis in July and on an absolute basis in August, recently turning roughly flat Y/Y; and by Redfin median square footage comps, which peaked in August
Similarly, our indicator of wood product shortages peaked in September and has stabilized since, in-line with forest products ETF relative performance and wood product pricing
We suspect US housing demand has been losing momentum, but remains at elevated absolute levels
TMI Data Science utilizes Natural Language Processing to build custom leading indicators using unstructured data sourced from the global financial, trade &amp; traditional media. Our proprietary software text-mines the global media to discern and quantify nuanced qualitative shifts in press coverage as they apply to macroeconomics, equity indexes/ETFs, commodities, currencies, fixed income &amp; individual equities. 

TMI Snapshot: Some US Housing Indicators Lose Momentum as Deurbanization Theme Appears Transitory

United States

TMI Snapshot: Some US Housing Indicators Lose Momentum as Deurbanization Theme Appears Transitory