Online sport result predictions built on artificial inteligence, neural networks and other predictive analytics.


We are professionals in coding, mathematics and statistics, financial markets and technology.

Putting our skills, experience and the interest in sport matches together, brought us to the idea of creating the web application for Sport Predictions generated by machine learning mechanisms and artificial intelligence techniques.


In the course of the future, we want to

  • cover all countries
  • cover all sports and e-sports, leagues, tournaments and games
  • provide predictions for as many sort of results as possible
  • continuously improve precision of all algorithms and output
  • improve the customer experience and
  • provide mobile app solutions
  • translation to different languages
  • use data from stadium video surveillance systems


To assure customers of our trustworthiness, we transparently provide

  • historical predicted results and
  • historical graphs for money management (available in the paid Premium Service. Six different models evaluated in terms of win/loss and in financial terms – calculated at average middle Europe odds).


Sport is

  • multibillion-dollar industry (US$ 270 billion in 2017 in U.S.)1
  • immortal
  • since Great Rome Empire
  • insensitive to economic cycles
  • insensitive to demographic trends
  • a huge market for all generations
  • naturally adapted itself to stay in and modern
  • participant of a huge social network, discussion forums, fan clubs
  • hand in hand with technology


1according to research firm Eilers & Krejcik Gaming, LLC.


Watching sports games on TV, as it was 20 years ago, is diminishing today. But the appetite for sport is growing. Today everyone wants to have a sport on the phone, have a link, be interactive and digital. And we definitely want to keep track of modern trends in customer behavior.


Research by McKinsey (2017) concluded that “millennials enjoy sports just as much as members of other generations. It’s the way they consume sports that matters”. The problem that needs solving is that of “declining attention spans”.

Millennials increased interest in short-term things, like stats, tips and quick highlights…


There will be niche segments of the sports wagering market that can be capitalized. Apps that provide up-to-date player data, betting line movements, trends and more all in one place will be valuable.2


There are several hedge funds that promote sports betting as a recession-resistant asset class. As the main reasons for their market advantage, they see:1) bookmakers can be mistaken in their predictions (many of the lost money in English football when Leicester won the Premier League in 2016 against the initial odds of 5000 on 1)

2) the large number of betters can skew the odds (in those cases, the bookmaker’s interest may not be to forecast the result, but to minimize its risk by attracting bets on a low-probability outcome).3



We predict future sport results created by artificial neural networks and advanced mathematics (…so the customer can just enjoy the pleasure of watching a game…)

At this point, predictions are executed at 01.00 a.m. CET (and CEST in the summer). Sport results are updated twice a day. If the customer has some special requirements regarding the run time, he can contact us individually.



In our models, we incorporate past results, head-to-head statistics, standings, home advantage, goals scored, overtimes, penalty shootouts, extra shots, team rankings, seasonality, and other objective variables.

This information is combined with detailed data about the odds available from various bookmakers to evaluate the model performance.



Meanwhile, we do not include tipsters‘ opinions, weather, players’ injuries and other stories whose risk and impact on the predictability has not been measured yet.

At this time, a pure mathematical model without incorporating tipsters gives Scoressi an edge over the average punter. Any combination without the precise explanation would cause an output distortion, and the customer would lose the ability to understand the model and use it responsibly.



In the long run, we will be continuously working on the enhancement of our complex predictive models to deliver the best value for our customers.

In the future, Scoressi will provide the separate outputs:

  • our pure A.I. prediction,
  • odds from betting companies (value bets, dropping odds),

Scoressi is currently absorbing thousands of sporting fixtures to create patterns of failure and success, and the ultimate goal is to create an AI that can track the full range of half a dozen different sports events simultaneously, even in live matches. At the moment, however, Scoressi is starting small and focuses only on a few sports (football and hockey) and a few metrics (1×2, goal chances,..).

We will fit and improve the accuracy of the models, improve customer experience and work on broadening the range of sports, locations, odds types and user devices.

Overall, fantasy sports represents a growing sector of the gambling industry, especially in North America, and more companies are expected to enter the market, which is said to have a US$ 4bn annual economic impact across the sports industry.5  Scoressi plans to be part of this sector as well.

5 www.gbgc.com

Scoressi environment will be optimized for new digital behaviors: convenient access from social media or search, quick navigation, alerts, fast and intuitive social sharing of game highlights and fan chatter.



We have already developed the framework for the predictive analytics and user interface to enable customers to buy our products.

Our system ensures our service is accessed anywhere and anytime.

Precising and updating the result of the bought prediction in the active 30-day period is free of charge.


Description of selected products (see Pricing)


  • 5 credits for 30 days
  • Pay as you use.

(Unused credits can not be rolled over to the next cycle)


  • 1 model
  • Model accuracy. Learn how the model picked the correct game winner.
  • Bar chart. Combine or sum up the probabilities as you like.



  • 30 credits for 30 days
  • Pay as you use.

(Unused credits can not be rolled over to the next cycle)


  • 1 model
  • Multiselection for activating predictions („Select All“ functionality)
  • Model accuracy. Learn how the model picked the correct game winner.
  • Bar chart. Combine or sum up the probabilities as you like.



  • Unlimited number of predictions.
  • 6 models
  • You understand the basics and you want more.
  • Model accuracy for 6 models. Learn how the model picked the correct game winner.
  • Bar chart. Combine or sum up the probabilities as you like.
  • Interactive chart of profitability for all models.


Try it

  • 3 credits for 14 days
  • Simplified Basic
  • Pay as you use.

(Unused credits can not be rolled over to the next cycle)


  • 1 model
  • Do you like it?
  • Yes? You’ll automatically continue in Basic.
  • No? No problem. Unsubscribe.*



* Your Try It will convert to a paid subscription of Basic product and your card will be charged for Basic product after this 14-day cycle of Try It ends.

In order to avoid getting charged for the next billing period for Basic product, you need to cancel at least one day before the end of 14-day cycle of Try It.

We will send you an email reminder giving you the opportunity to cancel 3 days before you’re going to be charged for Basic.

If you have already used a Try It product, you will not be eligible to buy this product again.


  • Model accuracy. Learn how the model picked the correct game winner. Build trust in the models.
  • Bar chart. Combine or sum up the probabilities as you like.



  • Enjoy one free prediction every week.
  • Monitor the trends.
  • Boost your self-confidence.


  • No credit card required. No commitments.
  • Model accuracy.* Learn how the model picked the correct game winner.

*Only for registered users. (User Account required).

  • Bar chart. Combine or sum up the probabilities as you like.




  • Pay with tokens and get 30% discount.
  • Cryptocurrency payments accepted.


  • Enjoy full access
  • Unlimited number of predictions.
  • 6 models


  • Model accuracy for 6 models. Learn how the model picked the correct game winner.
  • Bar chart. Combine or sum up the probabilities as you like.
  • Interactive chart of profitability for all models.



There is an option to get a discount for a referral:

50% off for you and your friend


Refer a friend and spend less.*


*It’s easy, the new customer enters your Referral Coupon in his first order. We will simply notify you whenever your referral is detected.

Discounts can not be accumulated. As an example, for 10 recommended paid customers, you get a 50% discount on the next 10 payment cycles.


Real time

In the current phase of development, predictions are calculated once a day during midnight CET. Teams or individuals do not play more than once a day, so this frequency should be sufficient to some extent for a temporary period of time.

However, to improve our predicted results, we are planning to replace once-a-day calculations with real–time calculations.

Real-time calculations task is incorporated in our Road Map and requires certain capital investments.

IT Architecture

Scoressi is engaged in statistical modeling and complex problem solving, with the aim of predicting future sport results.


The main process is described as the flow of the individual steps: data collection, unification and storage in a defined form, the calculation of statistical characteristics, preparation of statistical models and their use for generating a prediction and finally publishing the calculated values ​​to customers.

Data collect

Data collection is an important part of the main process. If the data is not delivered to the repository in a timely manner, the whole process ends with a poor quality output. The process will not stop, but without the latest input data, customers will not get the best possible result and the most up-to-date predictions. For high quality results, it is necessary to obtain high-quality data with minimal delay and also historical data from a long back period. In predictive analytics, historical data is a necessary input. The key task is to set basic logic for statistical models and to find attributes that affect results of the prediction. Since the impact of attributes on the prediction changes over time, the models must be tuned regularly.Model results are then compared to reality data that is also collected into the database.


Scoressi data collector is designed to collect data from structured data sources and report anomalies in the structure to prevent data loss in the next steps of the main process.


For quick configuration / reconfiguration, the data collector is designed in JSON format. These configurations and the runs of data collection in the test and development environment in the modeling and installation phase can be performed directly by the Statistical Model Developer (“Quant”).

The data collector reports the status of all runs in the HTML report sent by email with indications of problems and instructions for quick solving. For example, it is necessary to add a new mapping of two different data sources to the database, and so on.

It is planned to change this type of failure detection to another, more sophisticated monitoring tool.


Collecting data from different sources and combining them with core database requires transformation. There are differences in identification, date, time, etc., which must be mapped and converted to a common data format in the main database.

Transformation is a part of data collection and uses JSON configuration as well as the error reporting described above.


The result of data collection and transformation step is data that must be stored in the database. Data is stored in a MySQL database in a defined database structure model with the necessary time stamps and identifications. Identifications and time stamps of database items (e.g. a match) are interconnected with real object (e.g. a team) to provide time continuity in the database. Over time, some data sources (e.g. data provider) change for some items and the identification of new data sources must link to the old data source. This simple mechanism ensures the correct behavior of predictions of sports, competition, tournaments that are based on the data collected over time from different data sources.

Storing phase is part of data collection and uses similar JSON configurations and error reporting.


Predictions are calculated whenever new data is collected. This process takes place partly in JAVA and partly in the statistical tool R. JAVA is used to prepare data for statistical models that are subsequently processed in R and the results are stored in the database.

Scoressi predictor reports the status of all runs in the HTML report sent by email with indications of problems.

It is planned to change this type of failure detection to another, more sophisticated monitoring tool.

The design of the prediction model is a process that must be revalidated regularly. The first designed model is not expected to be the best. All models will produce some results, but this results need to be evaluated based on reality. Scoressi has suggested several models based on neutral networks and several machine learning methods for classification, regression, clustering, novelty detection, quantile regression and dimensionality reduction. All models have different inputs, f.e. standings,  results, or ELO values.

Validation of models take place in JAVA. In addition to comparing predicted results with reality, it calculates financial return on the basis of real odds and bets.


Most of data which main process collect, transform, store and predict are visible to customers. At Most data that the main process collects, transforms, stores, and predicts is visible to customers. At this time, only web access is available, but there is also an API add-on plan. The web interface is built as an extension of the WordPress content management system (eg user, user groups, rights, product, order management, …) Some WordPress plugins are customized. Other Web GUI extensions are built in PHP7.

As web server is Apache2 used. Ad database server is MySQL 15.1 used. As mail server is Postfix 2.10 used.

Apache2 is used as a web server, MySQL 15.1 is used as a database server, and Postfix 2.10 is used as the mail server.

Programming languages

For its components Scoressi uses JAVA, Linux shell scripts, R, SQL Procedures, PHP, JavaScript, JQuery, Ajax and other minor technologies.

Development, Engineering

Current functionality covers the entire main Scoressi process. Data is collected, transformed, stored, used for predictions and published to customers. The customer can order Scoressi products and, depending on the product type, the selected products and model performance are available.Scoressi processes recurring payments as defined in Terms and Conditions.


For a more user-friendly and useful graphical user interface (GUI), Scoressi plans to implement many additional functionalities, such as:·         At this point, it is possible to search events by date only, but in the near future we will add filters and sophisticated search functions.·         It will be possible to sort events according to the expected value (by odds)·         An evaluation of the model’s success will be added to help customers choose the optimal model for them·         Customer notifications and alerts


In addition to functionalities for customers, Scoressi will tune its statistical models, implement more sophisticated monitoring and scheduling tools and development environment (SVN, GIT, …).


Scoressi will attend sport exhibitions, sport conferences, affiliate events, will publish articles and posts like “bet of the day” into different discussion forums, customer’s emails, e-newspapers, on the website.



The collected data from our website allows us to conduct detailed analyzes and categorize our customers to better target their specific interests.

Each game has a specific audience, everyone who watches sports has different interests and a different experience of joy.

Not only quantitative data such as click rates, page views, and traffic are important to understanding user preferences. The qualitative user data are also important for the development of so-called user empathy. From our customers, we will continually collect feedback and try to incorporate their requirements so that we reach a balance of satisfaction among all our customers.


We will be turning the simple human interest in sport to the global experience of sport knowledge worldwide.


Since the potential customers are persons from general public and the service is sold directly from the website, marketing will be the main driver in the early phase of the enterprise.

Sales team

We understand, that also for SaaS company with a good, easy-to-use product, and in modern world of try-before-you-buy and freemium-to–premium models, it is a myth that the product should be sold itself. That’s why despite we are currently more technical oriented team, we know having the sales team is highly important.


We are strongly inspired by Andreessen Horowitz and Peter Levine recommendations, that we would like to follow.

For early stages, we will start with a light sales configuration: 2-5 technical sales engineers, with a lot of know-how in procurement, negotiation, discounting, etc.  They should be great with customers, be people-friendly and should have a keen understanding of the market landscape and what differentiate different players.

When we reach the revenue generated from this sales team 1 times higher than their costs, then we will hire additional 2-5 sales reps.

We start with selected European hockey and football matches.

When 80% of our reps are hitting a quota set at 3x their on-target earnings, they are consistently selling, and there is emerging some repeatable model, we may hire the Head of sales to set signals how to scale, determine the strategy for different geographies and verticals, maximize the contract value, manage pipeline and expansion to new regions and new sports. However, this phase might take more than one year.

The long-term target is to reach the revenue generated from sales team that will be 5 times higher than their costs.

The great focus will be put on the product marketing manager. He will be responsible for sales “pitch”- define the product fundamentals (set the reasons for wanting this product badly enough and wanting it now) refined over hundreds of meetings with founders, developers and sales team.

Sales team will have a target (total quota) 20% higher than the company plan.


Sales team, marketing, developers and founders will be involved in regular meetings with the agenda of the product design, functionalities, new releases and feedbacks. Integration of these pieces of a company plays a very important role in the health of the business.