PsicoTSI Project

This application uses information from the online game service Hattrick.org. This use has been approved by the developers and copyright owners of Hattrick.org, Extralives AB

Description	PsicoTSI 1.1.1.7	Instalation Help				Links	Versions History

	PsicoTSI Tool is a FireFox 3.0.* Extension. Click here to install PsicoTSI Tool.		PsicoTSI Tool is freeware. A big effort is needed to develop this tool. If you want to support this project, you can do it by making a donation using PayPal.

Introduction

“Psycohistory” is a science developed by Hari Seldon, during the last years of the Galactic Imperium. In the beginning only the great scientist was working on his idea, that consisted in making a simulation of the Galaxy (or Humanity) behaviour, making use of a simplified system. Together with a team of mathematicians and psychologists, he realized that he could develop the psychohistory science using the behaviour of a single planet, Trantor, the Imperuim capital... and it worked. By means of complex equations, it was possible to calculate the probability of a crisis to happen, predict which actions should be applied to solve such crisis, and finally, check the effects through a new simulation.

I hope this science existed (in hands of kind people).

Motivation

Hattrick is not only a soccer game, it is also a strategy game. Some people use HAM, supporter statistics, hottrick, experience, advises... Well, in addition to those tools, we might use as well "neural networks".

My problem is that, like many other people, I like doing things right, I mean, I like winning. For this issue, playing hattrick, a strong economy is essential to buy players, employees, etc. The way to achieve this purpose is "steptrading", this is adquire players close to skill up, with high skill sublevels, train them to the next level, and sell them. This is the future (or present). So it is. With the last measures adopted by HTs, "daytrading" has ended, or at least is less effective (€ earned/time consumtion relationship). That was the way to reach the highest leagues some seasons before. Just take a look at the transfer history of many great teams.

The PsicoTSI Project consists of predicting TSI using the player's skills; and back, predicting skills using TSI. How?

Neural Networks

Neural Networks (NN) belong to the field of the Artificial Intelligence (AI), which among other things are used to approach functions.

Variables:
• Skills
• Form
Approached function:
• TSI

Basically, the neural networks are functions that optimize their parameters by means of different methods (genetic algorithms, gradient, etc). It is a sort oflinear regression, but much more complex.

Matlab includes a full-functional NN Toolbox. Since I have used it several times for my work at University to solve different problems, I considered to keep on learning, in this case, playing hattrick.

The main problem was the ammount of information required to optimize the NN perforamnce, so that the NN does not adapt herself to the information, but it fits or comprehends the general behavior of the studied phenomenon. An approximate minimal ammount of necessary information is 10 times the number of parameters of the network, which for this problem can be in a range of 100 and 150 parameters.

That's why I am grateful for the collaboration of bendito, sujito and akmanar. Thank to them, a very extensive database (DB) of players (not goalkeepers) under 26 years old, and not injured was completed with... 2000 entries!

With that DB the NN could be properly trained ;)

Results

I don't promise any succes. I don't promise that we can solve completely the problem. A reasonable approximation can be enough. The next graph shows the fitting of the PsicoTSI Tool with the 2K players of the DB.

Look at them, there they are, players with a very high TSI, compared with the rest of the players... By means of the PsicoTSI Tool, we will be able to know which players should we buy and/or train...

Limitations

In spite of the advantages of the PsicoTSI Tool, it has strong limitations with a very difficult solution:

• Injured players or players older than 26 years cannot be analyzed with the PsicoTSI Tool, because their effects have not been included in the Neural Network training.
• The data included in the DB and the player information are very inaccurate. For instance, a passable skill (passable = 6) could be from 6.01 to 6.99.
• If the player to be analyzed has two main skills at the same level, the error of the sublevel prediction can be very high.

Use Instructions

The use of the PSicoTSI Tool is very easy. You must follow this steps:
1.- Install the FireFox extension.
2.- Close and open the web browser.
3.- Open the player details page of any of your players or a pleyer in the transfer list. In the popup menu (right-click) ther will be a new element: PsicoTSI.

Acknowledgements

This tool has been possible thank to bendito, sujito y akmanar. The code is inspired in FoxTrick and Roachfiend. If you want to add new players to the database, have any suggestion or comments, or detect any error/mistake, please, contact me at [email protected].

	Introduction “Psycohistory” is a science developed by Hari Seldon, during the last years of the Galactic Imperium. In the beginning only the great scientist was working on his idea, that consisted in making a simulation of the Galaxy (or Humanity) behaviour, making use of a simplified system. Together with a team of mathematicians and psychologists, he realized that he could develop the psychohistory science using the behaviour of a single planet, Trantor, the Imperuim capital... and it worked. By means of complex equations, it was possible to calculate the probability of a crisis to happen, predict which actions should be applied to solve such crisis, and finally, check the effects through a new simulation. I hope this science existed (in hands of kind people).
	Motivation Hattrick is not only a soccer game, it is also a strategy game. Some people use HAM, supporter statistics, hottrick, experience, advises... Well, in addition to those tools, we might use as well "neural networks". My problem is that, like many other people, I like doing things right, I mean, I like winning. For this issue, playing hattrick, a strong economy is essential to buy players, employees, etc. The way to achieve this purpose is "steptrading", this is adquire players close to skill up, with high skill sublevels, train them to the next level, and sell them. This is the future (or present). So it is. With the last measures adopted by HTs, "daytrading" has ended, or at least is less effective (€ earned/time consumtion relationship). That was the way to reach the highest leagues some seasons before. Just take a look at the transfer history of many great teams. The PsicoTSI Project consists of predicting TSI using the player's skills; and back, predicting skills using TSI. How?
	Neural Networks Neural Networks (NN) belong to the field of the Artificial Intelligence (AI), which among other things are used to approach functions. Variables: • Skills • Form Approached function: • TSI Basically, the neural networks are functions that optimize their parameters by means of different methods (genetic algorithms, gradient, etc). It is a sort oflinear regression, but much more complex. Matlab includes a full-functional NN Toolbox. Since I have used it several times for my work at University to solve different problems, I considered to keep on learning, in this case, playing hattrick. The main problem was the ammount of information required to optimize the NN perforamnce, so that the NN does not adapt herself to the information, but it fits or comprehends the general behavior of the studied phenomenon. An approximate minimal ammount of necessary information is 10 times the number of parameters of the network, which for this problem can be in a range of 100 and 150 parameters. That's why I am grateful for the collaboration of bendito, sujito and akmanar. Thank to them, a very extensive database (DB) of players (not goalkeepers) under 26 years old, and not injured was completed with... 2000 entries! With that DB the NN could be properly trained ;)
	Results I don't promise any succes. I don't promise that we can solve completely the problem. A reasonable approximation can be enough. The next graph shows the fitting of the PsicoTSI Tool with the 2K players of the DB.

	Look at them, there they are, players with a very high TSI, compared with the rest of the players... By means of the PsicoTSI Tool, we will be able to know which players should we buy and/or train...
	Limitations In spite of the advantages of the PsicoTSI Tool, it has strong limitations with a very difficult solution: • Injured players or players older than 26 years cannot be analyzed with the PsicoTSI Tool, because their effects have not been included in the Neural Network training. • The data included in the DB and the player information are very inaccurate. For instance, a passable skill (passable = 6) could be from 6.01 to 6.99. • If the player to be analyzed has two main skills at the same level, the error of the sublevel prediction can be very high.
	Use Instructions The use of the PSicoTSI Tool is very easy. You must follow this steps: 1.- Install the FireFox extension. 2.- Close and open the web browser. 3.- Open the player details page of any of your players or a pleyer in the transfer list. In the popup menu (right-click) ther will be a new element: PsicoTSI.

	Acknowledgements This tool has been possible thank to bendito, sujito y akmanar. The code is inspired in FoxTrick and Roachfiend. If you want to add new players to the database, have any suggestion or comments, or detect any error/mistake, please, contact me at [email protected].