An attempt to launch a startup in the European market of sports events

An attempt to launch a startup in the European market of sports events

If the attempt had been successful, then obviously the article would have been called a little differently. On the other hand, I can’t call this idea a failure either, since the process is still ongoing. Today I’d like to tell you about the interesting challenges and challenges we face when trying to launch something of our own.

I will note right away that this article is a kind of continuation of mine of the first articleTherefore, all the events described here begin after the publication of that one. This time, it’s just a story of what happened after I shared my hypothesis and plans for it with you. Did the team gather then? Why is Europe here? What tasks have been solved? What is the result? Let’s figure it out, but about everything in order.


Received feedback

So, the article was published, I began to wait for messages from people who were interested. In the first week, about 15 people wrote to me, and what surprised me then was that most of them were from different countries, although the people themselves were Russian-speaking. I had the most interesting experience working with guys who also had a similar background in teamwork and research/development in this field. For example, a programmer from the US who developed a data analysis platform with his team. Their software analyzed large amounts of information received from a variety of offices and found an advantage in values ​​by comparing all quotes with the Pinnacle reference line. Moreover, in the creation of their mat. the doctor of mathematical sciences participated in the apparatus. Or, for example, a young guy from Cyprus who used an algorithm based on the delay of updating quotes. His team even had its own office in Moscow.

In general, we had a good conversation, exchanged experience, expertise, told each other about various problems and ways to solve them, shared views and vision of the market.

Among the others who wrote to me were other interesting people: a young programmer from Belarus and a Russian-speaking entrepreneur from Belgium. They were the most interested in the idea, and it was with them that we decided to work together on the project.

Formation of the team

First, I suggest getting to know the guys better. Ruslan is a programmer from Belarus who works in an international logistics company that develops a product for users from the CIS and Europe. Kyrylo is an entrepreneur from Belgium who works on various projects (from analytics to retail) in Europe with his team.

Ruslan wrote to me first. We quickly held negotiations, during which we realized that it would be interesting and very useful for us to work in the same team from the point of view of knowledge exchange, both in the subject area and programming.

Kyrylo responded to the article a little later. The three of us already participated in the negotiations, after which Kyrylo made it clear that he agreed to work on the project together with us. He was ready to connect his European team to work and invest in us, but we will stop on that a little later.

Concept and goals

Obviously, we did not plan to work on the Russian market right away took the vector to the European one. The main idea was to eventually work not in online offices, but in their offline locations on the territory of Europe. In this way, we planned to avoid all bureaucratic procedures with verifications, cutting limits, as well as leave a large potential for scaling, both in terms of team size and in terms of the funds that are rotated by this team.

The stages of the project can be summarized as follows:

  1. Development of analytical software

At this stage, it was necessary to develop a data collection and analysis system for the online office, as well as a system for event notifications (telegram bot). Ruslan and I are both programmers, so we decided that our efforts and time for development would be quite enough – the system was designed to be relatively simple.

  1. Data collection and analysis

Here we had to collect a sufficient amount of data, analyze it and simply understand whether the hypothesis works (Confirm or deny it).

  1. Preparation for launch

This step involved starting all the processes, but on a smaller scale. We planned that a part of the European team under the leadership of Kirill would start spending time in offline points, placing bets in small amounts based on the signals of our robot. It was necessary to prepare the boys for the work in order to get acquainted with its specifics in advance.

  1. Launching

This phase included the investment of funds for the work and the full launch of all processes with a team distributed across offline locations in several European countries. Among other things, here we considered the development automated system of accounting for rates and finances within the project to eliminate collisions and routine operations in the team.

  1. Scaling

With the successful implementation of the previous points, it was possible to place people in literally every country in Europe and even touch part of the countries of North America. Instead, there were plans to start providing consulting services in the field of software development to analyze and work with data on the market of sports events, but already in Russia.

We drew up a detailed plan, assigned roles and, having tuned in to the result, got down to work.

Development of analytical software

As it should be, you should start with preparation. We created a project management board in Trello, a repository on Bitbucket and found servers for the software. The next step was to develop the terms of reference. I have described all the hypothesis conditions and system requirements in one place so that you can only work with this document in the future. It was decided to collect data from one Russian office with approximately the highest coefficientsBecause before that, all my models were built while working with it.

Now it was necessary to decide on a stack. I wanted to implement the logic and data processing module myself in C# (since it is my main language), and Ruslan would take care of the data receiving and sending module, implementing all this in PHP. However, he proposed to do everything in one place to reduce the complexity and speed of development – so they did. Final stack looked simple: PHP+Yii2+MySQL. At that time, I was already familiar with PHP and even had a small developed project on it, so C# or PHP – there was no particular difference for me, and we started.

Time for our project could only be allocated in the evenings, as both of us also had a main job, so the pace of development was moderate. They closed current tasks, set new ones, overcame emerging difficulties, and just enjoyed the process. The task board then looked something like this.

Early stage project board in Trello

Completed all implementations, tested the work, fixed bugs and inaccuracies, placed on the server and launched. Data collection has begun. Now it is necessary to collect the necessary volume of events in order to start making some initial conclusions and, depending on them, make decisions about our further actions.

Data collection and analysis

We waited, and the system worked. The first collected data had the following format.

The first collected data and their structure

Periodically, I entered the database, making simple SQL queries, to see all the necessary indicators and their changes. Already in a month I refuted his first hypothesis as it wasbut saw something new.

Analyzing the events, I noticed an interesting dependence of the result on one of the indicators. Based on this observation, I formulated new additional rules that looked very logical and organically fit into the existing model.

So, within two months, we went from a total sample of about 1,900 events with an increase of just over zero to a rarefied sample of 200 events with an increase of 20 denominations, ie. ROI (return on investment) was equal to 10%. The schedule is given below.

The growth graph of the optimized algorithm

After consulting with the whole team, we decided to proceed to the third stage of the project.

Preparation for launch

As I mentioned earlier, now it was necessary to run all processes “on minimums”. At this stage, Kirill singled out initial investment 2,000 for the total deposit – exactly €100 was set aside for one bet. They laid the budget for the next stage as well, planning more turnover. The countries decided to start with Germany and Belgium – one person per country. So, everything was ready and we made a test run.

The process went as expected, without any excesses. The European offices provided almost all the events that our bot sent, so the guys put everything in place with confidence. The bank began to grow, and some days the increase reached 200€. It would seem that you can rejoice, but no.

After working in this mode for about a month, we realized that our the bank does not grow, but simply fluctuates around the initial valueAlthough according to the initial data from the base there was a small increase. Here we realized that it’s a matter of coefficients, or rather, their difference. The coefficient for a specific event at the Russian office is by no means equal to the coefficient for this event at the European one. As a rule, it is always smaller. The entire increase according to the algorithm was equalized by the difference in coefficients – increased margin at offices in Europe. A direct consequence of this is the unsuitability of the algorithm with an ROI of 10% (in Russian) for work on the European market.

Searching for a new algorithm

After holding a meeting with the team, we came to the conclusion that a new algorithm is needed. Obviously, it had to be more profitable than ours to cover the negative difference in ratios.

This time we decided not to “reinvent the wheel”because there was neither desire nor time for that. It was pointless to spend months collecting and analyzing data again when there was an opportunity to find a strong analyst with a ready-made, more cost-effective algorithm.

It is no secret that there are many verifiers – special sites with analyst ratings, where each of them makes their forecasts, keeping open statistics for a long period of time. Moreover, such forecasts are always the result of the work of the analyst himself, it can also be the result of the work of his system (bot), which sends signals within the limits of a certain algorithm. This is exactly what we needed.

I found one very interesting Russian verifier with such bots, began to observe his rating and the chat in which the participants of the site communicated. After analyzing the situation, I decided to write an analysis to Oleksandr from Tomsk – the owner of one of the best robots of this verifier.

We quickly conducted negotiations and agreed on cooperation. His algorithm suited us in all respects, the ROI was 20%. Actually, here is the schedule with an increase of 400 denominations for 2,000 events in one and a half years of work.

Work schedule of the robot according to Alexander’s algorithm

Difference of coefficients

This time, before starting the process, we decided to install the exact difference in coefficients. After processing approximately 50 signals, we found out that the deviation was about 10%that is, on average within the framework of the current algorithm the coefficient in the European offices was 10% lower compared to the original signal coefficient.

Here it is worth noting that the deviation for two different Russian offices will not be the same – for a bookmaker with high coefficients, the deviation will be greater, that is obvious. Alexander’s algorithm, like ours, is based on the data of just such an office, which is why we have such a big difference.

With a simple calculation, you can get the ROI of this algorithm in relation to the European market – we have 20% from the starting point about 6%. Taking into account the percentage of the team at points in different countries, commissions for translations, movement between countries and other additional costs, we may not see net profit at all. They decided not to start the process.

Summing up

Now I suggest summarizing everything said above and trying to understand what to do next. At this point we have: a scalable concept; ready team to implement this concept; programmers for the implementation of technical solutions and investments for the future. However, we do not have an analyst with a profitable algorithm on the European market – this is the missing element, which we lack for a full launch.

One way or another, we will continue to look for the right person for us and continue to try to launch our startup. If you think you can be useful to us, have the appropriate competence and desire to join our team, write to me in telegram: @arlol

Otherwise, that’s all. I thank everyone who read the article to the end. I wish everyone success and quick launches! 🙂

Related posts