The Draft
Every year, the National Football League holds a draft in which eligible college players who declare for the draft are selected by 1 of the 32 teams in the league. The draft consists of 7 rounds, and the order in which teams are granted picks in each round is determined by the records of the teams in the previous season, with the worst record selecting first and the best record selecting last. Teams may either use their picks to select a player, or they may trade their right to draft a certain pick to another team in return for players, other picks, or a combination of both. The following was the order for the first 3 rounds of the recent 2019 Draft:
What Determines Twitter Sentiment?
NFL Teams have access to every possible data source available surrounding potential draft picks, understand what positions the team needs to fill, and most often meet with the players they draft before they draft them. Thus, it can be assumed that each NFL team is making the best available selection given the information they have. Yet, fans and analysts seem to have polarizing opinions about most picks, and take to Twitter to express them. What factors surrounding a given draft pick, if any, affect the twitter reaction to the pick?
To analyze which factors affect Twitter reactions, I will be focusing on the first round, as it is the most watched & reacted to.
The following are the first round picks of the 2019 NFL draft, and the average sentiment (derived using the syuzhet package in R) of all original tweets mentioning the player's name in the 24 hours following the first round of the draft (scraped using the twitteR package) .
To analyze which factors affect Twitter reactions, I will be focusing on the first round, as it is the most watched & reacted to.
The following are the first round picks of the 2019 NFL draft, and the average sentiment (derived using the syuzhet package in R) of all original tweets mentioning the player's name in the 24 hours following the first round of the draft (scraped using the twitteR package) .
Does the Picking Team Determine Sentiment?
One factor to pick reaction is the team making the pick. Do players picked by teams who have a history of picking high quality players in the first round receive a more positive sentiment in their Twitter reaction? To determine the quality of previous picks, I used Pro Football Reference's database. Pro Football Reference has created an algorithm for evaluating the "Approximate Value" of any player in a given season. The methodology for the value can be found here. Using the Pro Football Reference database (assembled nicely in a database by Kaggle user Ron Graf), we can calculate the average amount of "Approximate Value" that each NFL team's first round picks have generated over the first 4 years after they were drafted to determine team picking ability.
From the above chart it is clear that some franchises have had more success in picking players in the first round than others have. Does a teams ability to pick quality players influence how their picks are reacted to?
The above graph reveals that there does appear to be positive relationship between historic pick success per team and Twitter reaction to their pick. This could mean that analysts and fans are more confident in picks from teams who have a history of making good selections However, the relationship is extremely slight. Another factor to consider is how a team's presence on Twitter affects the reaction to their picks. Does a larger fanbase on Twitter affect average reaction to teams picks (perhaps fans are more positive, or maybe more critical)?
The above graph shows a positive relationship between the amount of Twitter followers a team has and the sentiment of the reactions to their picks, indicating that larger Twitter fanbases might influence a more positive Twitter reaction. Again, the relationship is extremely slight. A final thing to consider about the team making the pick, is how the team performed last year.
From the above graph, it appears that there is a negative relationship between 2018 win percentage and Twitter sentiment reaction. This could indicate that fans of teams who did poorly the year prior are more optimistic about their picks, or that analysts may be more optimistic about a players ability to be impactful on a worse team. This relationship is also slight, and so it is worth explore factors outside of the picking team to see what other factors are determining pick sentiment.
Does the Player Himself Determine Sentiment?
According to the NCAA, there were 15,000+ college football players eligible for the draft in 2019. Of these players, only 32 were selected in the first round. These players represent the best of the best, yet why are some received more positively than others? One factor to consider may be the quality of the college program they are coming from. A way to measure the quality of the program is to used ESPN's measure of Football Power Index (FPI), which is its measure of team strength for college teams. We can use the 2018 season end FPI to determine program quality coming into the draft.
The above chart illustrates the disparity between program quality that many first round draft picks are coming from. While it makes sense that good players would be coming from good teams, drafting players is about choosing the best available individual players regardless of program. However, does coming from a better program yield a more positive outlook, and thus more positive sentiment for a player?
Here, we see a positive relationship between school FPI and sentiment. This relationship also appears to be much more positive than the relationships we found from drafting team statistics. This may indicate that players coming from better programs are perceived as being higher quality picks and thus receive a more positive reaction on Twitter.
Another way to factor in individual player values is to look at the player's rank himself going into the draft. The best way to determine this rank is through mock drafts (when analysts predict when every player will be chosen in the draft). Sports analyst Ben Robinson has conducted a popular study in which he compiles all reputable mock drafts and creates a mock draft based on the average of all the ones collected. This gives each player an "ADP" value (Average Draft Pick) going into the draft, and we of course also have the actual pick that the player was taken at. This gives us a difference between the predicted value of ADP and the actual pick , and tells us whether the player was taken too early or too late relative to the popular opinion of sports analysts.
Another way to factor in individual player values is to look at the player's rank himself going into the draft. The best way to determine this rank is through mock drafts (when analysts predict when every player will be chosen in the draft). Sports analyst Ben Robinson has conducted a popular study in which he compiles all reputable mock drafts and creates a mock draft based on the average of all the ones collected. This gives each player an "ADP" value (Average Draft Pick) going into the draft, and we of course also have the actual pick that the player was taken at. This gives us a difference between the predicted value of ADP and the actual pick , and tells us whether the player was taken too early or too late relative to the popular opinion of sports analysts.
The above chart gives us the actual, ADP, and difference for each player in the first round. A negative difference means the player was drafted before he was predicted to go (too early), and a positive difference means the player was drafted after he was predicted to go (too late). Does this measure of difference to predicted pick affect sentiment?
The above chart shows that there is a positive relationship between difference from prediction and sentiment. It can be inferred that this relationship is because fans and analysts might be reacting to picks based on when they were predicted to go. They may consider "early" picks to be of bad value to the drafting team, and "late" picks to be of good value to the drafting team. This relationship is also much more positive than the drafting team metrics. What happens if we create a variable of Team FPI + Difference (both similar in scale) from Prediction?
This combination of variables yields us our most positive relationship. There appears to be a clear trend between a player's draft pick relative to his prediction along with his school strength and the Twitter reaction to his pick.
Conlcusion
When it comes to sentiment, it appears that while the team picking the player may have some influence over the ensuing reaction, what is more important is the quality fo the program the player is coming from and where in the draft he is selected relative to where he was predicted to go. This could have various implications. From a marketing perspective, it may make sense to endorse players who are perceived to have been a good deal in the draft, particularly those from good programs to capitalize on positive sentiment. From a sports data analytics & research perspective this shows the importance of mock drafts and their ability to determine the perceived value of players going into the draft. From a perspective of college football programs, better programs might use this info to try to recruit young talent, with promises of a more positive social reception.