Data and Forecast Dashboard

Live dashboard with market projections, market and political news sentiment analysis, and a complete suite of market data such as last price, volatility, volume, bid/ask spread, log returns, mean and standard deviations for realized statistics such as volatility and price.
This live dashboard also contains some of our web app metrics such as time to acquire market data and news, model inference time for each forecast and sentiment, time to resolve metrics, and others. All that is to give you a peek at the internals of our algorithms.
Important Notes:

    1) All market data and statistics are for the SPDR S&P 500 ETF Trust (SPY) symbol. We plan to add other stock ticker symbols to our platform in the future, however currently only SPY is supported.

    2) The dashboard should update automatically about every 10 seconds, with a new data point at every 30 seconds. Best viewed on Chromium-based browsers (Chromium, Chrome, Brave, Edge, Opera, etc), not Firefox or Safari.

    3) By continuing on this webpage, you agree with our Terms of Service. All information on this website is for educational purposes and is not an investment recommendation nor to be representative of professional expertise. Individuals are solely responsible for any live trades placed in their own personal accounts. This is not a real company or product. This is a free of charge passion project available to the public.

    4) Please direct any questions or feedback to Vinicius at vgg@cmidas.com.

Checking market hours...

Data last updated: Computing...

Dashboard last updated: Computing...

Market Forecasts and News Sentiment Analysis

The two charts below show the main output of our data-driven models. The leftmost one shows the last price the symbol was traded and our price forecast for 30 minutes in the future. The model is constantly updated as more data comes in. The rightmost plot shows our current estimates for market and political sentiment in the United States based on multiple news sources. Our algorithm crawls the web for news sources and estimates their sentiment using state-of-the-art, deep learning-based, natural language processing models. Sentiment values range from -100% to 100%, depending on how negative or positive the news text is.

But how accurate are our models? The answers lie in these two charts below. When forecasting future prices, or any quantity really, two metrics are important: direction accuracy and the magnitude of the error. In other words, did our model correctly guess the market direction? And how far were our price targets?

To compute the market direction forecast accuracy, shown in the leftmost chart, we wait until our 30 minutes forecast in the future is realized then we check if the market moved in the direction we predicted 30 minutes ago. If so, we get a 100% accuracy for that data sample, if not, 0%. This process is repeated every 30 seconds: acquire new data, update our model, forecast future, check if previous forecasts were correct.

Over time, we can compute how well the model is doing throughout the day by computing a rolling average of these individual accuracy scores, as you can also see in the leftmost chart. While the accuracy for market direction is binary, we either got it right or wrong, the rightmost chart shows the magnitude of the error for the predicted values, which can be useful for some trading strategies that rely on more than just the market direction.

Compiling direction forecast metrics...

Compiling value forecast metrics...

Options Trading Opportunities

Now, given the above market direction forecast and options chain data shown below, our algorithm suggests a few options trading opportunities. For example, if the market is forecasted to move upward, the algorithm will more likely suggest the purchase of calls. If moving downward, to purchase puts. Below are the algorithm's latest suggested options trading opportunities:

Options Chain Statistics

This section monitors the complete option chain for the underlying asset we are interested in. Here we distill all market information and how the option chain is moving with the goal of identifying options trading opportunities with a high probability of success. Below are the calls and puts that changed the most in terms of market value and volatility when compared to their values during the market opening. We also display a histogram of the same changes for all contracts in the option chain.

Call Options: Highest Price Movements (Positive and Negative)
Call Options: Highest Volatility Movements (Positive and Negative)
Put Options: Highest Price Movements (Positive and Negative)
Put Options: Highest Volatility Movements (Positive and Negative)

Realized Statistics

Here are some of the realized market statistics averaged out over rolling windows of 10 minutes. The first row shows the 10-minute rolling mean and its standard deviation for the last price the security was traded. The second row shows the 10-minute rolling mean and its standard deviation for the realized volatility, which is affected by price changes in either direction and can be understood as a metric for market uncertainty: the higher the volatility, the higher are the price oscillations.

Additional Market Statistics

Here are some additional market statistics that are useful for our models and algorithms, which can also give you a better insight into how the market is currently behaving.

The leftmost chart of the first row shows the current bid/ask spread for the ticker symbol, which is the ratio between the best ask and bid prices centered at zero. The rightmost chart is the current volume-weighted average price, which takes into account not only the best bid as ask prices but also their volume.

The charts show the 10-minute market return and its standard deviation. This quantity is computed by dividing the current volume-weighted average price by its value 10 minutes ago. The first row shows the raw values and the second one their logs.

Market Data

This section shows some standard price, volume, and volatility market data that is available to our models and algorithms. We spend a good amount of time transforming all this raw data so it can be useful for forecasting.

Web App Metrics

This whole section monitors technical web app metrics to give us a better understanding of how our models and algorithms are operating in production. This first row monitors the time our algorithm is taking to crawl the web in search of relevant market and political news and how many of them were identified.

Here, in the leftmost chart, are the times our models are taking to update and forecast market direction and value, in addition to the time taken to infer the sentiment of scrapped news stories. The rightmost chart shows the time it takes to compute and update our forecast metrics for direction and value.

These charts monitor the time it takes to write various components of our approach to their respectives databases, as seen in the charts on the leftmost column. The plots in the rightmost columns keep track of the time to acquire the data from the option chain and the latest market data and hours of operation.

This chart monitors if we are meeting our desired time to acquire new data, update our model, forecast the future, and check if previous forecasts were correct. Currently, all that is supposed to happen in under 30 seconds, 24 hours a day. If we do not meet this desired mark, we can often check from the charts above which process is taking longer than expected, which helps us pinpoint the modules that need to be optimized.

Summary Statistics

Computing...