 Good morning to everybody Good morning. First of all, thank you for coming for this presentation. My name is Andrea Alcon. I'm a maternity specialist in fellowship and as she has said, I will talk about artificial intelligence in professional football Artificial intelligence has been applied successfully in a lot of areas of our society For example in medicine industry advertising in economics, and it has obtained very good results Till the point of gaining an important spot in most of these areas In fact, it is also really interesting to notice that artificial intelligence influence has increased a lot In almost all business systems. For example in the professional sports industry This industry has been overwhelmed by the capacity of accessing a lot of great amount of data Which has helped to gain an understanding in this sport Also, it's very interesting to notice that Inside the sport, inside the sport clubs Almost all the medical departments also collect big amounts of data Which help them to collect as much information as they can of their players For example in one football match Millions of events are recorded that so finally they show the performance of the overall players in the game So We are in a scenario where we have great amount of data and we want to extract valuable information from them Here is where artificial intelligence plays a big role because using machine learning algorithms We are able to convert that data into knowledge What's Olozip? Olozip is the pioneering company in the research development and application of artificial intelligence in professional soccer We develop products with scientific rigor and that provide technological innovation Our products and services are crucial for the management of soccer clubs and their related companies And we have great databases which allow us to create top quality predictive models So as in recent decades the amount of data collected in in a sport has raised a lot It has helped to gain an understanding in the sport But also came along with an extremely high potential of being able to extract benefit from that data Also for companies for clubs for both of them So it is important to gain this benefit It is important to be able to extract valuable information from the data and develop models that are capable of obtaining this information Here is where Olozip plays a big role because we are able to create top quality predictive models that are able to transform real temporal data into a transparent statistical models So I want to show you how one of our database looks like we have a set of instances of The data we are working with We later choose the variables that are more important to fulfill to fulfill the objective we are looking for And we use intelligent algorithms to develop an automatic knowledge that is able to predict with high accuracy The objective we are looking for Now I want to present you the leaders of our of our group of our company The company is led by Esteban Granero a professional football player from a Spanish football club He with the engineer guys Kazami Sente both of them from the company in 2016 and the technological partners of Of the company are Pedro Aranyaga and Concha Bielsa Both of them are professors in artificial intelligence and have thousands of publications and papers that are very Notorious and have win a lot of prices Nationally and internationally Both of them are working in the UPM So with this introduction, I want to show you that we have Olozip has deep knowledge in three main aspects Of the market we are working on we have deep knowledge in artificial intelligence deep knowledge in technology and deep knowledge in a sport in this garden diagram, I want to show you how the Data analysis market is a structure We can see four principal dimensions the descriptive dimension the causal dimension the predictive dimension and the prescriptive dimension Almost all the companies working in this field Exploit the descriptive dimension they analyze past data that has been recorded from the past So this is very interesting but it's far from the full potential that we can obtain from this data So here is where Olozip also plays a big role because we work on the four dimensions we describe We diagnose we predict and we suggest we start from the origin what has happened till the Couch why has it happened? We make a prediction about what will happen in the future and we also make suggestions in order to be able to Reach what we want to happen So I want to remark also that our objective is not to replace the human but to try to Give the human the ability of reaching his full potential So I want to see a one of our models. This is how one of our most how looks like how it looks like This is a Bayesian network It each of the nodes represents a variable of our model and as you can see They are connected between them. There are a lot of connections which show the relations between the variables So as you can see our models are transparent. This is a one big important fact of our enterprise because we are able to show which are the weak points of the of the team We are different to the artificial intelligence companies that develop models that are black boxes Which means that you are not able to see what is happening in your model because if your model gives an concrete output With black boxes Artificial intelligence models you are not able to know why the algorithm is given this output in our models as they are transparent You could look into the model and you could see between the connections that the variables Are in you could see the weak points of your team You could see strong points you could change the output of the algorithm You could see okay I'm getting this this output because this variable is related with this one and both of them are related with this Another one so this is very flexible and gives a lot of information for the clubs So when we develop our product what does the clap see which is the interface they are working with So we develop a big tool that is called tct clap Which introduces artificial intelligence in the key departments of the clap and is composed by three big products That are tct coach with real-time suggestions tct doc for injury prevention tct scout and tct scout for scouting and performance of players So now I want to show you the three main products The first one is tct coach. This is a smart complex and intelligent tool that is able to give Advice help the management team of our clap of a clap For example, he helps the trainers. He helps the analyst to take decisions during the game. We have a Our data is collected in real-time. We have data points that are collected in real-time from the so the match that is being played And we have a database for more than 1000 football games so Which are the strong points of our product we can provide a information in real-time We can anticipate weaknesses of our team or strong points as our team We also have the probability of goal of both teams in real-time So if I analyze this example for example We can see In this a part These are the events that are happening in the game And if we look down, this is the probability of goal that our model is calculating So for example, it is notorious to notice that our model is Is learning because as we can see the probability of goal in this case is a match between hetafe and a bar so The probability of goal varies during the game analyzing the real data that is being extracted from the game And as we can see Get off as probability of goal start to increase in the minute 23 And as it goes up suddenly a goal is scored by yet So for example, this shows that our model is learning well If we continue analyzing this example, we could see that a vast probability of goal At this point start to increase start to increase start to increase and reaches the point a very high point in which In real-time they score a goal So as you can see we are able to predict with high accuracy the events that are happening in the game So if we go to the prescription that our model gives the trainer is able to set an objective in this case it could be for example scoring a goal and With this objective the model will be able to recommend which is the best strategy that the teams will Have to increase that chance of scoring a goal for example our model could Give an instruction that is your actual pressing lines are very low if you increase these pressing lines To a high point you will increase your chances of getting the objective in an 8% And it is also very important to notice that We have we are very flexible with objectives of the of the of our tool for example If the trainer sets a goal that is we want to score a goal in we're in the match We are tight and we want to score a goal when the team succeeded in scoring a goal Maybe his his plan is to try the rival not to score a goal against us So you you could easily change this objective and the model will give you another strategy that is able to Maximize your the chance your chance of not being scored so in this way We are very flexible and we provide the trainer a lot of ways to be able to change his mind to work with Instructions to make changes of their players will give him a lot of a lot of chance Our second tool is tct doc tct doc is an intelligent tool developed to avoid professional players injuries In this case we have a very large data database of a lot of variables that is are collected from the team We have genetic variables. We have anthropometric variables. We have workload variables We have the injury historical of the players and with all that information. We built our model So what is our model capable of? In this case our interface will look like this This is the part where we saw all the descriptive data from the player The blood analytical the workloads Anthropometric values all of this is descriptive So our model is capable of making predictions based on the data. We are giving him as I was saying Based on the descriptive data that we have collected our model is capable of making predictions which type which type of predictions We will have a prediction of the maximum workload that a player could have and also the optimal workload that that players will be having in on training some games and Also, we will have a real-time prediction of the probability of getting injured of that player after the training before the training we will have an exact probability of The players probability of being getting injured in each part of his body For example, he will have a different probability of getting injured his name his wrist His uncles everything will have a concrete probability of being injured. So this is very useful for trainers to be able to prepare Personalize Trainings to be able to avoid his player getting injured We also have a global rating of election ability that will show which is the overall status of our players And it also helps it also helps to give instruction to be able to first of all avoid injuries second prepare players that have been injured to return in good conditions to the team and also make make overall Overall informants or give overall informants of all the players of our roster So we will be able to have information of four of our players their probability of injured The workload they should have the time they should rest it will provide a lot of information And finally our third product is TCT scout TCT scout is a tool that helps scouting teams To get the best players for the team we have inside this big tool We have four tools that are progression which shows the progression of our players how these players are going to play the next season Scouting which shows how a sign in a player will affect not just his performance, but also the performance of the whole team Similar players with sauce imagine you are a club. You want a concrete player, but you can't You can't sign in because it's too expensive So you will be able to put this player in your team make a prediction and makes on the predict Based on this prediction, you will be able to search for similar players And we also will have the impact tool That is very useful because it allows us to set the time or that a player should be playing our team imagine Do you sign a young player and you want to see how he's going to play if he plays each game 30 minutes? you could put for him 30 minutes and You will have a prediction if you want to change that you will you could increase his minutes you could give more minutes to another striker and you will see how the The performance of the whole team and his performance his personal performance is going to change So that is very useful for you to set the optimum number of minutes that that players will be playing so now I want to show you some examples of a Example that we have made in this case. We are making a prediction for Cristiano Ronaldo's performance in Juventus We want to see how he's going to play this year in Juventus And we want to compare that with his previous statistics at Real Madrid So we will put we this we are using now a scouting tool the scouting tool will put Cristiano Ronaldo in Juventus And we will have a prediction we will have prediction in three main aspects the offensive Performance the overall offensive performance of the player the overall defensive performance of the player and the overall build-up performance of the player So in this case, we will have in red the performance of Cristiano Ronaldo in Real Madrid And we will have in blue the performance the overall performance that Cristiano Ronaldo is going to have in Juventus this year so it is very interesting to notice that With our prediction Cristiano Ronaldo will rise in the number of assists that he's going to have he's going to make But he will decrease in the number of goals that he was scoring in Real Madrid Here you will see the exact percentage of all the variables and you could compare them with the new ones and you could see the same imagine if you are signing up a Defensive player for you are more interested in the defensive variables So you will have you will analyze the defensive variables if I can tell you some of the variables that we are using for this For these models in the offensive model. We are using assist crosses, goals, pass verticality, shots Take his goal All the offensive variables that we are that we have in our model. In the defensive area, we have tackles radio, total interceptions Tackles won, tackles lost and in the build-up Area we have back passes switches of play, pest radio Sots verticality losses of ball losses. So As you can see We are able to make predictions of how our play how a new player is going to play in in our new team and We also have the prediction of how Juventus is going to play with Cristiano Ronaldo In this case if we look on the offensive variables of the team, we will see that Despite Cristiano is going to score less goals I think Cristiano Ronaldo to this team helps to increase the probability Goals that the team is going to to have so we can see that for example offensively in this case as Goals are going to be more and assist also are going to increase We will see that Cristiano Ronaldo is fitting well in the new Juventus team And we will make the same thing with the defensive variables that are not going to change a lot because Cristiano Ronaldo as we know he's offensive player and the build-up attributes also so As you can see we can put each Player in the new team and predict players performance and teams performance And this is the impact tool We can see that in this case We have switched all the minutes that Gonzalo was playing that he was playing 33% of all the minutes of the of the matches And we have assigned all these minutes to Cristiano Ronaldo if we want to change this if we want to see how Cristiano Ronaldo is going to play if he plays all the minutes of the Of the league we could take some minutes from the ball up or manjukic or Bernat deski We could decrease Cristiano Ronaldo's minutes We could increase the ball as minutes and we will have completely different results because we will see how this change of Minutes is going to affect to our team and to our players And now I want to show you our similarities tool in this case. We are Inter Milan club and we want to see Similar players to look at my bridge as you know this summer Inter Milan wanted to say to sign a look at my bridge, but finally he stayed in Real Madrid and We want to see how which will be a great alternative to to look at my bridge So I want to remark that we make predictions based on how look at my bridge is going to play in inter No, how he was playing in Real Madrid. We make a prediction of look at my reaching inter Milan And we see similar players that will play in the same way of look at my reaching inter Milan so First of all, we will define the set of variables that we want to analyze In this case, we are looking similar players to look at my bridge in assists in build a play in Interceptions in past verticality and in tackles based on the attributes that you that we set The result will be completely different because we will looking for other profile players So our algorithm will be will give an answer in this case The most similar player to look at my bridge with those attributes is Julian Traxler with a 90% of similarity But the next players are on grand say with a 99% continue Erickson David Silva You get a big list of similar players with his level of similarity So this is also a very useful tool for weak teams because they can scroll down in in our results In our results and they can look for the player that fits their condition if they don't have as much money as Real Madrid can have for example they could scroll down till the point they find a player that suits them by the Probability of similarity the similar a similar player and will fit their budget So for this type of teams, this is also very interesting to have this tool And now I want to remark that all of our products all of our services and are fully validated. We Make a lot of calibration methods. We use a lot of a validation tools k-fold course validation a lot of validation techniques that allow us to have a product that is fully validated and To show that our results are are real We are not changing any result and as you can see the real probability The real probabilities very similar to the predicted probability. So we are very very accurate making predictions And I want to sum up Which are our what are we giving the claps? We have a products with scientific rigor We have we generate exclusive knowledge. We predict results. We make patient of results We reduce the uncertainty. We can prevent injuries before they happen We are transparent. We give claps transparent models That you can analyze every clubs are very interested in transparent models because they want to switch variables they want to see how changing available will affect the outcome of the game or How will affect the performance of a play of a player? We also give a feedback to the key departments of the claps And we have very flexible model because you can switch your strategy you can switch the goals of your club and We generate in direct economic compensation because for example if we avoid injuries of players This is money that the team is wasting because his players not able to play Also, if we help teams to win matches that also is very beneficial for them and also we can avoid a But singings of players we can recommend the best player that is going to fit in the team And you avoid some some of the signings that they are always some of them are Are not very good for the team. They lose a lot of money is sending top players that later They don't play really well in there in their new teams And I want to show you that we have have an international impact because our produce have been used internationally We have appeared in a lot of media. We make a prediction of look a mother's the one that I have showed you and Some of the big newspapers policy it for example last police that canal is is one of them Similar players to look a mother it's with the variable that I saw I saw before we also appeared in Jean-Lucas images Twitter because We make the prediction that I have showed you about Cristiano Ronaldo and he was very interesting We make a prediction of how Neymar was going to play if he comes back to Barcelona We compared a bidon Paulinho We also appear in a lot of news in Italy related with our Cristiano Ronaldo's prediction Which are the next step of all of it. We are working where we working on we are working also with running companies We are developing products for basketball. We are also developing tools for tennis We are also entering in the e-sports world and Our products have the capacity of entering a lot of business we can enter in bank We have very top quality models that will fit very well in bank in In low in a lot of in health. We also have very important products for health So there are a lot of ways of implanting our tools in all these areas And now it's time for questions if I can thank you very much for your attention And if I can answer any question I will be pleased Hello, very interesting Can you tell us please some examples of the clients you have? I can't tell the names, but we have clients of the first division of Spain. We have clients also. We have a lot of Clients in in the representative in the representation agencies and For example, I in the city dog I think I we have so one of our clients in the DCD product that is from the first division of Spain and we also have clients that are playing in England in Italy in in the top divisions We have clients in the DC. I have a question which is somewhat related to that So I think for some of the real-time predictions that you mentioned in the first part in the code product You probably need real-time data to feed For this for the model and I think that this kind of data is only currently available in the top leagues What's so I'm wondering are you only interested in clients in those leagues? Or do you expect that in the near future that kind of real-time data will also be available for in other leagues? We are now we are developing tools that will be able to We are going to install this type of technology in also clubs of the second division of Spain For example, we are trying also to reach to that type of clubs that don't have the Nowadays the technology, but we are developing a software that will be able to get this real-time data Hi Can you explain us a little bit more? How you get the Data from the players in real-time and how you are loading the data to the system to make the predictions in real-time Yes, we have cameras that and we have a software that Is the flow between the cameras on our and our software? We love all the information that the cameras are recording and the track system of the players And we process it in our model and the model is able to make prediction based on the data We are collecting from from the cameras and the GPS in real-time Cameras from the public TV system and so on we have cameras in the stadiums We have put our cameras in the stadium and we collect the data They have a GPS system that we are we have treats with them We have if we have Treats with them and we are able to extract this information and to load it in our system And be able to process it later when processing real-time and make the predictions Yeah, but we have treats with the clubs before this time the match started We have treats with them that if they are if they can provide us the data and if they have said we can make that prediction for that game Yes, yes, of course. Yes So going back to the cameras thing so you only have cameras in the clubs that you're working with or Because if I wasn't using your company You will 100% not have cameras because I wouldn't want you helping my rival, right? No, we said cameras we have before a match If our team is playing we have them the they allow us to put our cameras in the stadium And if we have that permission, we are able to make predictions. Yes, we need the teams permission to be able to Establish our product. So if your team is playing Away, then you might not be able to give them we might not be able but in this point We have had that chance of putting our cameras Hey, could you speak a bit more about what kind of data you're using? So we saw on the screen things like assists or maybe how much they were running But do you also for example use use geospatial data? Yes, yes dependent on the product for example in coach we use The positioning of each player needs moment of the game. We have variables For example in the DCT talk Product we have a genetic variables all the variables related with the health of the player in coach we have variables of positioning of Total passes total all of the data that is recorded from the game and Do you also account for the interactions between players, of course? Yes We have tackles one tackles lost that is radio if a player wants an aerial radio all the interactions between them They are collected Sorry, we don't have time for more questions. We have to end here. Thank you very much for your talk. It was very interesting Thank you