Esports Transfers - How Do We Calculate Player Transfer Values?

Our vision on CS economics

We are currently witnessing the end of the CS:GO era, and soon, we'll be in CS2. With thousands of tournaments and millions in prize pools, you might wonder about the CS economy. The truth is, we know very little because most contract details and prices are hidden. We only see the tip of the iceberg. But let me ask you this: What's the prize pool of the Major? If your answer is $1 million, then you're quite wrong.

One year ago, Valve reported that they paid around $70 million in one year to professional organizations.

“It’s a big number, yet CS:GO is stronger than ever. Over the past 12 months alone, we’ve seen more players than ever before (averaging over 20 million monthly unique players), record viewership for Majors (2.7m concurrent viewers), and massive community support with over $70 million raised for professional organizations, teams, and players” - Source

Let me remind you that it's just a revenue share, so Valve also earns a big chunk of money from the Majors.

So basically, the top 50 teams (or 250 players) generate $70 million in revenue just from the Majors (or $35 million per Major). But what about the other tournaments?

A Ukrainian company, Maincast, bought the rights to cast all ESL tournaments in 2019 for 3 years for $11 million. So ESL earned $11 million just for those rights. Source.

In 2022, FACEIT was sold for $500 million, and ESL for $1 billion (well, ESL is not just a CS:GO brand and they handle multiple events and have a huge infrastructure. However, FACEIT was mainly focused on CS:GO stuff, and more than 70% of its players are CS:GO players).

Summary. The top 50 teams or 250 top players are generating about $70 million in value for themselves and their organizations just from Majors, not including different partnerships, sponsorship deals, or the prize money. So, what's the total market value of those players? Well, that is the tough question, and I will be sharing more details and thoughts below, but in our opinion, a $100 million market value for those players is quite a fair number, isn't it?

I want you to keep those numbers in mind until the end of this text, so we will avoid comments like "LMAO, $3.9 million for ZywOo. Who's gonna pay that?"

How Do We Calculate Transfer Values?

What kind of data do we use?

To calculate transfer values for esports players, we rely on a comprehensive set of data collected from HLTV and FACEIT for players over four different time periods: 3 months, 6 months, 1 year, and all-time. We track player performance separately for online competitions, LAN events, and Majors, resulting in 12 distinct data sets.

Full image

It’s just a small example. At present, we monitor more than 300 HLTV players and over 1000 FACEIT players. While we possess over 300 statistical parameters per player, we do not use all of them for player valuations, as many of these parameters overlap.

We also gather media statistics, and short notes about the players.

Pentagon (Skills, Consistency, Experience, Media and Leadership)

For player evaluations, we consider five key parameters: Skills, Consistency, Experience, Media, and Leadership. We have experimented with various combinations of data to arrive at the desired valuation numbers. The screenshot you see displays the metrics we employ for assessing Consistency, for instance.

Although these metrics are not perfect, we continue to fine-tune them. However, with the transition to CS2 on the horizon, we have mainly focused on other tasks since the shift from MR16 to MR12 will significantly impact the data we have.

Here is another example related to the Media parameter:

We recently excluded FPL, Twitch activity, and Twitter activity from this parameter, as they were found to be less useful, and their numbers fluctuated within the margin of error. For the Media parameter, we now primarily rely on Twitch, Twitter, and Instagram followers. Additionally, we have specific settings for individuals who have followers on Twitter and Twitch but lack an Instagram presence.

At the end of the day, we utilize five parameters (note that we plan to replace Leadership with Teamplay soon, as the screenshot data indicates). These parameters are calculated automatically. While we welcome constructive criticism and aspire to involve more experts in our community to enhance our methods, our vision is to maintain automatic calculation of these parameters.

How do we estimate the prices?

Demystifying the Data Challenge: Making Sense of Complex Information

Imagine you're trying to understand a puzzle with 370 pieces, each unique and wildly different from the next. That's what we faced when diving into a vast dataset with approximately 370 features, or individual data points. But, as with any complex puzzle, not every piece is crucial to seeing the big picture.

Initially, our team attempted to make sense of this data using basic tools, much like trying to solve our puzzle using just intuition. We turned to techniques like linear regression, a standard go-to in the world of data analysis. However, given the vastness and high variance of our data (that is, the wild differences between our puzzle pieces), these methods simply got overwhelmed. The result? Overfitting. It's like forcing puzzle pieces together that don't truly fit – the end image doesn't quite look right.

To avoid this, we employed some clever preprocessing techniques. Imagine being able to reduce the number of puzzle pieces by merging some of them together without losing the core essence of the image. Among used methods we utilized PCA (Principal Component Analysis) tool that help in diminishing data dimensions while preserving its core information. From our initial 370 puzzle pieces, we managed to maintain 95% of the original image's integrity using just 30 main components.

With our refined puzzle pieces, we ventured beyond the basic tools. Consider this: instead of a single individual trying to solve a puzzle, imagine an ensemble of experts each with their own special methods. We used a combination of machine learning algorithms.

But how to make the best use of each expert's puzzle-solving method? Here, we implemented a Voting Regressor. Think of it as a council of experts, where each one shares their solution, and a collective decision is made, combining the strengths of all the methods.

In conclusion, dealing with vast and varied data is akin to solving a complex puzzle. While initial methods may falter, with the right tools and a combined approach, clarity emerges from the chaos.

Prices, prices, prices

We gathered known prices from the news, thanks to the latest transfer window, which provided us with 20 more relevant numbers. However, the issue is that we need more official prices to make our model more accurate. Then, we matched the relevant stats with the prices and instructed our model that when it sees a certain range of data, it corresponds to a specific market value. We asked it to help us determine other numbers.

The first problem we encountered is that sometimes prices seemed random due to media activity. For instance, FalleN would constantly be in the top 10, even if he decides to end his career. Age and experience also played significant roles in these valuations.

We also faced the challenge of explaining to the model why m0nesy was considered expensive, even though he hadn't participated in any Major or LAN games, only in the WePlay Academy League. This evaluation was somewhat subjective, making it difficult to assess certain player types automatically.

As a temporary and stable solution, we decided not to ask the model for specific prices. Instead, we requested it to provide us with a price range. So, instead of saying 'this player's value is $1.2 million,' the model could suggest, 'I'm not sure, but it's somewhere between $1.2-1.5 million.'

To make our model results easier to understand, we simplified them and organized them in a spreadsheet format.

So, the predicted rating falls within a certain range of numbers, which helps us estimate a player's value. Once we have the model's expectations, we review the player manually and gather information about them. Our Pentagon rating system is also useful in this process. Let's take a closer look at the evaluation of Insani as an example.

Insani has a Skills rating of 93 and a Consistency rating of 75. This raises a flag for us to check his stats more carefully.

We refer to HLTV and find that his stats are indeed impressive, even when facing top-30 teams.

However, when it comes to consistency, we notice that he may not perform consistently, which is essential information when considering a player for a future team.

Which might be important to know when you will be able to research players for your future team.

Despite our model suggesting a specific range for Insani, we ultimately value him at $110.000. The prediction can vary since Insani’s price might be influenced by his current team’s (MIBR) position, which is currently 25th in the HLTV rating. However, we take into account that he lacks the experience, consistency, and media presence to justify a higher valuation at this moment.

For some players, particularly those aged 24-27 from the top-30 teams, the model provides a wider valuation range because their parameters align more closely.

In this update, the total market value of the top 250 players is approximately $108 million, which we consider to be fairly accurate, especially when considering the $70 million revenue generated in 2022 just from the Majors.

We acknowledge that our model can still be improved, and some of our valuations may be inaccurate. We welcome your feedback and suggestions here, on Twitter, or on Discord.

In summary, we leave the final word to you for your comments and insights!

How We Plan to Create a Perfect Model

Our ultimate goal is to earn trust within the community and among team management, much like Transfermarkt, a website that handles football transfers. Transfermarkt relies on experts and volunteers rather than models, and while it may have some margin of error, it has earned trust from the community and teams alike.

To make esports and CS more professional, we believe that Valve needs to establish Financial Fair Play Policies and Transfer Rules to prevent the game from being dominated by just a few big clubs. However, to introduce a transfer platform to the audience, we can't simply provide arbitrary numbers. The pricing format we've presented so far is a foundational step that will need refinement.

Our vision of a perfect pricing model doesn't necessarily rely solely on math models crafted by our best data analysts. It could also involve input from experts and the community. Our algorithms can play a role in ensuring transparency and accuracy in the future by considering even the smallest details.

Esports Transfers Plans

Currently, we are focused on expanding the number of players we track. By the end of the year, we aim to have data on more than 2,000 players, including rising stars from FACEIT and CS2 matchmaking.

Our team is relatively small, and we don't have any partnerships at the moment. As a result, we're mainly refining our player lists and website content.

While we await the CS2 release, we've begun laying the groundwork. For example, we plan to collect and analyze HLTV demos to determine how much time pro players need to check specific positions and angles on the map. By finding the average time among all professionals and then analyzing FACEIT/CS2 demos, we hope to identify promising lesser-known players based on this parameter.

Same could be done with movement, aim, reaction time and many more.

In the meantime, we've already discovered interesting insights, such as the average pro player’s (top-30 HLTV) knife switch rate. Based on 30 demos, players switch their knife 3.73 times per round after the freeze time. 58% of those switches were done to increase the speed, while the remaining 42% had no practical purpose.

Looking ahead, we plan to expand our platform to include other games, but CS stands out as the one with the necessary infrastructure to automate various aspects effectively.

P.S Thank you for taking the time to read this article. We hope you found it interesting and thought-provoking. Your feedback, ideas, and comments are always welcome and valuable to us.

Don't hesitate to explore our Esports Transfers website, follow us on Twitter https://twitter.com/etransfersgg, and engage in discussions about this topic with us.

If you have any questions or inquiries, please feel free to reach out to me on Discord at @solutiontv or send me a direct message here on Reddit. For potential collaboration opportunities, you can also contact us via partners@esportstransfers.com.

Your participation and support mean a lot to us as we continue to work on improving the world of esports transfers. <3