Author: Victor

The debate of who is the pound for pound best fighter in mixed martial arts is often brought up after some of the world’s greatest fighters compete. However, how does one rank fighters who have never competed against each other and most likely never will? Win/loss records are often used to compare fighters but more often than not, these types of debates are quite subjective. David Coelho, a software architect, used social network analysis in an attempt to answer the question of who is pound for pound, the best fighter in the world.

In Coelho’s network, fighters were nodes and directed edges represented their fights. Nodes with edges pointing to them meant that they won their match against the other nodes. The figure below is an example of how the network looked. 

After creating the network, Coelho used Page Rank to determine who the best pound for pound fighter was for each gender. In CSCC46, the class was taught to think of Page Rank as a circulating fluid between nodes. In Coelho’s social network analysis, he thought of Page Rank as the relevancy of fighters that would get passed from one fighter to the next, depending on whether the fighter lost or won their fight. Fighters that won would receive the losing fighter’s relevancy and thus increasing their own. Coelho adjusted the Page Rank algorithm to account for fighters losing relevance when they lost fights. The figure below displays how the network looks when including the Page Rank algorithm. As you can see, relevancy pools up on the most successful fighters.

What interested me to read and explore this article was how the author was able to use social network analysis to answer one of the most subjective questions, about one of my favourite sports. Perhaps one day, social network analysis can be seen as the most credible tool to rank the greatest fighters in the world. For those curious fight fans, the tables below display who Coehlo’s social network analysis determined as the best pound for pound male and female fighters in the world.

Source: https://www.linkedin.com/pulse/network-science-analysis-history-mma-david-coelho/

Four weeks into the NFL season and teams have begun to realize exactly how successful they can be this year. Some teams are undefeated, while others have yet to record a single win. How exactly do team executives and coaches decide on signing or re-signing a player they believe can improve their team? One obvious way is to use player statistics from prior seasons. However, what if there was a way that did not involve using any player statistics. A study by Salim & Brandão attempted to use a NFL social network to predict team success. 

In the study by Salim & Brandão, nodes in the network represented quarterbacks and teams. Edges represented the labor relations between quarterbacks and teams. The figure below is the NFL social network the researchers came up with.

The researchers calculated the clustering coefficient of the NFL social network and compared it with that of the Erdös-Renyi network. Since the clustering coefficients of these two networks were very different, the researchers were able to determine that the NFL social network was not a random network. In class, it was explained that the Erdös-Renyi Random Graph can be used to determine if networks are random or not, which is exactly what Salim & Brandão demonstrated in this study.

Using the Pearson coefficient, the researchers determined a positive correlation between node degree from the NFL social network and team success in one of two NFL conferences. Figure 11 displays the positive correlation between node degree and team success in the American Football Conference (AFC). This result is significant considering the researchers compared this model’s findings with another model that used a quarterback performance statistic called passer rating, to predict team success. In that model, there was no pattern of correlation between passer rating and team success (Figure 10). This meant that the passer rating metric did a poorer job in predicting team success, compared to the model that used the property taught in class called node degree. 

In my opinion, the results from this study can have major implications in the sports world. What intrigued me to read and write about this study was how the researchers used concepts taught in our class, in order to make meaningful observations about one of my favourite sports. In the competitive world of professional sports, teams are always looking for an edge over their competition. Instead of using player statistics to make roster decisions like every other team in the league, team executives could start utilizing social networks to determine if they should sign or re-sign certain players. It will be fascinating to see if social network analysis can change the way professional sport teams conduct their business in the future.

Source: https://api.semanticscholar.org/CorpusID:14037766