Ivontu is a recently launched portal that provides automated financial news analysis and enables its users to understand the impact and effect of breaking news on the prices of the stocks that users are interested in. (currently targeting most US stocks in the S&P 500).
How does Ivontu work?
Imagine, as part of your periodic investing activity, that you track the reports of an analyst for some specific stock, say the classic Acme Industries. Your favorite analyst writes reports that lay out, in excruciating detail, the various factors underlying upcoming changes in the stock price. In all of this, you still haven’t been able to understand if the analyst means that Acme will go up like a rocket or down a la the Red Baron. You need help.
Ivontu enables you to create news models for most stocks in the S&P 500. These news models can be visualized as brand-new virtual analysts that need to be trained. Training is as simple as feeding news such as the analyst reports mentioned above, or news from other sources such as Bloomberg or the WSJ.
Once your model is well-trained, you can use it to analyze the latest news reports and predict its effect on the stock that it is tracking, Acme Industries in this case.
For each model, Ivontu calculates the robustness (model’s predictions behavior)and accuracy (quality of the news that you have used to teach the model) and depending on these scores, one can train the model for certain accuracy.
I did a quick QnA with Anshuman (whose earlier venture was BanKaro):
More info on modeling aspects of Ivontu:
I use my own proprietary model, which is powered by a mixture of statistical analysis of the text and stock price data. Normally such analysis is usually delivered via the standard black box mechanisms of neural networks and/or evolutionary computing. However, under most circumstances I normally avoid using models that I wouldn’t be able to understand by myself; hence, there are no black box models being used in Ivontu. A whole host of data (including statistical significance values) are generated and tracked on a periodic basis for every user model; this data may or may not be made available to the user later on (if I can figure out how best to package the data visually). The results (likely direction of stock price movement) are generated numerically; however, only the direction of the movement is shown to the user, primarily to reduce accuracy variation.
There are tons of other predictive tools available in the market..
Most other predictive tools for the stock market derive their “power” from a cursory analysis of price and/or returns data alone, which serves as a proxy for both stock fundamentals as well as the sentiments of market participants (I refer to both fundamental and technical analysis). On the other hand, Ivontu provides the seasoned investor a shot at separating the two effects. For instance, by creating two separate models
to track a single stock, one for fundamentally-oriented news items, and one for technically-oriented news items, a composite picture of likely stock movements can be drawn. Further actions are dependent of course on what the user learns from the model predictions. In short, Ivontu can only provide analysis; action is dependent on the user.
Also, Ivontu models can be used independent of time axes, e.g. models can be trained on what the user believes to be news having longer-term effects. These models can then be used to understand price action 6 months or a year down the line. Again, such variation in modeling is dependent on user discretion.
If you do have a specific predictive tool in mind, do let me know and I can then perhaps draw out a better comparison between the two.
I’d like to think of Ivontu as a potentially disruptive innovation. As most of us know, computing power has expanded dramatically over the last decade; the usage of such power in analytics has increased proportionately. However, in regards to normal human activities such as investing, we are still stuck in the paradigm of the 50s – we pay too much heed to measures of central tendencies such as the mean, ignoring the fact that the reason it was introduced was so as to be able to manipulate statistics effectively in the absence of computers.
This situation has been rectified to a large extent inside large financial institutions or nimble hedge funds, which use their own proprietary models to track anything from the amount of snowfall in New York to footfall patterns in Bloomingdale’s (I kid you not – Jim Simons’ fund uses this apparently to correlate with retail stock price movements). However, the power of these models lie locked inside the same institutions, thus effectively stopping the average investor from ever managing to grasp an edge in the market.
I believe Ivontu is disruptive primarily because it unlocks the same power of proprietary models, distributing the fruits of its labor to everybody. Why am I doing this? Because computing power has become too cheap, and the knowledge required to develop such models is becoming more widely distributed; it is difficult to keep these things caged. In short, Ivontu is positioned as a low-end disruption aiming to uproot traditional business models that are hinged upon selling analytical results that are generated by humans.
Ivontu is thus disruptive to those institutions that employ analysts to perform analytics that can be more easily automated and/or replaced by computational modeling.
What Ivontu offers is a singularly unique proposition: the ability to get financial information analysed by a source that is inherently unbiased, coupled with the sheer novelty of doing so liberated from transaction costs, which translates to a potentially much larger learning and (hopefully) richer experience.
Give Ivontu a spin and share your comments.