The Bayes Theorem offers a useful tool for measuring the usefulness of different tools of analysis when considering draft prospects. The theorem essentially measures the "degree of belief" with which one supports a certain position. The main goal of the theorem is to measure whether a position is "likely true" or "likely false." When used by scouts, the theorem can allow scouts to avoid overvaluing certain features of the pre-draft process. Therefore, it is a useful tool for all who are interested in the NFL Draft. I will begin with a statement of the Bayes Theorem:
P(A|B) = P (B|A) P(A)
P(B)
Where A and B both represent given events and B is not equal to zero.
The terms are definable as follows:
P(A) and P(B) are the probabilities of observing A and B without regards to each other. I.E. they are the probability of whether A or B will happen on its own.
P(A|B) is a conditional probability. It tells us the probability of observing event A given that B is true.
P(B|A) is the probability of observing event B given that A is true.
How does this apply to the NFL Draft? Let me show you. Let us consider the imaginary draft prospect Dick Donlin. Dick Donlin has just completed a strong collegiate football career at Hamline University in Minnesota. Will he be a first round pick this year? Let's plug some variables into the equation. We'll set B as the probability that Dick Donlin will be a first round pick based on game tape and A as the probability that Donlin will gave a strong Combine performance. These variables are independent to an extent; Donlin can have a horrible Combine and still be a first round pick given his collegiate performance.
Let's assume that Donlin is a can't miss prospect. We'll call him the best defensive end in college football, which almost guarantees that he will be a first round pick. Therefore, we'll set P(B) as a high number. Call it a 95% chance that Donlin will be a first round pick based on game tape. That starts us out as follows:
P(A|B) = P (.95|A) P(A)
P(.95)
Given Donlin's strong college performance, he should have a decent Combine performance. However, there have been some questions about Donlin's conditioning, so the probability of a great Combine performance is lower. Still, he's a decent athlete. Let's give him an 85% chance to have a strong Combine performance. That gives us the following:
P(A|B) = P (.95|.85) P(.85)
P(.95)
Then:
P(A|B) = P (0.8075) P(.85)
P(.95)
Then:
P(A|B) = 0.6864
.95
This gives us the probability that Donlin will have a strong Combine performance given the fact that he has performed like a first round draft pick of 72%. These numbers are just given for the purpose of demonstration, but something stands out here. When one approaches the sum of Donlin's college film and his Combine performances on equal ground, a poor Combine can kill Donlin's draft stock. A poor Combine lessens Donlin's chance to be a first round pick, even if the film shows him to be a surefire first round option, if one places the Combine, or a Pro Day, on equal footing to his college performance on the field.
There's a lesson to be learned here. The value a scout places on data from pre-draft evaluations should be weighted based on the importance of that data. Is a Combine or Pro Day workout performance really important enough to make or break a prospect's draft stock? The experiences of JaMarcus Russell and Teddy Bridgewater suggest that, at least to some scouts, they are. Teddy Bridgewater dropped in the draft following a subpar workout, but he has turned out to be a decent NFL quarterback. JaMarcus Russell rocketed to the first overall pick of the draft following a stellar workout as the Raiders overlooked the obvious risks that came up throughout his college career (weight, inaccuracy, etc). Teddy Bridgewater was a worthwhile first round pick. JaMarcus Russell is rightly labeled as one of the biggest busts in the history of the NFL Draft.
How should we value data? I suggest valuing data proportional to the prospect's career. Let's consider a prospect who has started every game of his career and is entering the draft as a senior. That's 4 seasons of 12 games, assuming no Bowl games or time missed due to injury. Just for an ultra simple example, we'll consider a 50-50 split between offensive and defensive game time, and also assume that the player played on special teams. So out of 48 one hour games, our player has put 24 hours of game tape on film. That's a sizeable amount of data.
Now let's consider a combine or workout. The prospect likely puts an hour or two of usable data on film, at most. In reality, the data is significantly less. The significance of that data should therefore be much less than the significance of the available game film. There lies the major problem with our Dick Donlin example. We valued Combine data and game tape data equally, and a poor workout killed Dick's draft stock. Of course, if we reverse the numbers, a strong workout can inflate a player's draft stock heavily even given questionable game film. Is this the proper way to evaluate players? Of course not.
There are multiple problems with putting too much stock in workout and Combine data. A player can train hard for the events in the workout to make up for subpar on-the-field performance, and inflate their own draft stock. Of course, a player who's workout is significantly better than their on-the-field performance may artificially inflate their own draft stock. These players will be over-drafted in the NFL. Another key problem with the workout. A workout is a one-day event. If the player has a bad day, then the workout could artificially deflate his value. Granted, players can work out at both the Combine and at their Pro Day, but still, two days compared to a record of 48 games. I'll take the 48 games over four seasons. Workouts are generally not in full pads, and never show the prospect in a real game-day situation. Therefore, the value of a workout should be seen as significantly less than the value of a player's film record.
Game tape is clearly the most important assessment tool available for demonstrating a player's future NFL potential given the amount of data. The next most important piece of data varies depending on the player. If you're looking at a small school player, Senior Bowl or Shrine Game practice and game performances might show you that the player in question can perform well against top tier talent. If a player has character issues, then interviews are very important. If a player has had medical problems in the past, and your evaluation of him as a draft prospect is dependent on his medical status, then medical examinations are key.
Workouts and Combine data should be at the bottom of the barrel as far as value as an evaluation tool. While these events can measure a player's athletic ability, they measure the player's ability to train for the events in the workout more than the player's ability to play on the football field. When a fast 40 time makes a player rocket up draft boards, don't buy in if the new standard value on the player is significantly higher than it was based on their game tape. The reverse is also true; don't let a poor workout kill your value on a player if their game tape says otherwise. Game tape is always the truest measure of a player's potential as a football player. Combine workouts are useful when comparing similarly ranked prospects, or when looking to draft a player from a different system. For example, if you plan on drafting a DT who played in a 4-3 in college to play nose tackle in a 3-4, you might look for a strong bench press performance to show that he has the upper body strength to handle multiple blockers. However, don't let bench press numbers dictate your ultimate value on the player. If the film looks good, and his bench press looks good, go get him. If the film looks iffy but he bench pressed well, maybe he will be a great fit in your 3-4, but then again, maybe he will struggle like he did in a 4-3. Watch the film and see how he performs against blockers before making your final decision.
My final point here is not to throw out Combine and workout data entirely. Rather, this data should be valued accordingly. It is foolish to over-emphasize a player's workout data at the expense of their game tape. Doing this leads to drafting JaMarcus Russell first overall, and to Teddy Bridgewater nearly falling out of the first round of the draft. Workout data is useful, but it is not nearly as important as a player's game tape. As our quick foray into the Bayes Theorem shows, if one over-values workout and Combine data, their final evaluation of a player can be fundamentally flawed. Don't let that happen to you. Watch the tape, watch more tape, look at workout and Combine data, and then watch the tape again. Don't let a workout warrior trick you into drafting you higher than you should based on film; instead, focus most intently on the film, and then value your other pieces of data based on their ability to shed light on the player's long-term potential in the NFL. --Mike B.
P(A|B) = P (B|A) P(A)
P(B)
Where A and B both represent given events and B is not equal to zero.
The terms are definable as follows:
P(A) and P(B) are the probabilities of observing A and B without regards to each other. I.E. they are the probability of whether A or B will happen on its own.
P(A|B) is a conditional probability. It tells us the probability of observing event A given that B is true.
P(B|A) is the probability of observing event B given that A is true.
How does this apply to the NFL Draft? Let me show you. Let us consider the imaginary draft prospect Dick Donlin. Dick Donlin has just completed a strong collegiate football career at Hamline University in Minnesota. Will he be a first round pick this year? Let's plug some variables into the equation. We'll set B as the probability that Dick Donlin will be a first round pick based on game tape and A as the probability that Donlin will gave a strong Combine performance. These variables are independent to an extent; Donlin can have a horrible Combine and still be a first round pick given his collegiate performance.
Let's assume that Donlin is a can't miss prospect. We'll call him the best defensive end in college football, which almost guarantees that he will be a first round pick. Therefore, we'll set P(B) as a high number. Call it a 95% chance that Donlin will be a first round pick based on game tape. That starts us out as follows:
P(A|B) = P (.95|A) P(A)
P(.95)
Given Donlin's strong college performance, he should have a decent Combine performance. However, there have been some questions about Donlin's conditioning, so the probability of a great Combine performance is lower. Still, he's a decent athlete. Let's give him an 85% chance to have a strong Combine performance. That gives us the following:
P(A|B) = P (.95|.85) P(.85)
P(.95)
Then:
P(A|B) = P (0.8075) P(.85)
P(.95)
Then:
P(A|B) = 0.6864
.95
This gives us the probability that Donlin will have a strong Combine performance given the fact that he has performed like a first round draft pick of 72%. These numbers are just given for the purpose of demonstration, but something stands out here. When one approaches the sum of Donlin's college film and his Combine performances on equal ground, a poor Combine can kill Donlin's draft stock. A poor Combine lessens Donlin's chance to be a first round pick, even if the film shows him to be a surefire first round option, if one places the Combine, or a Pro Day, on equal footing to his college performance on the field.
There's a lesson to be learned here. The value a scout places on data from pre-draft evaluations should be weighted based on the importance of that data. Is a Combine or Pro Day workout performance really important enough to make or break a prospect's draft stock? The experiences of JaMarcus Russell and Teddy Bridgewater suggest that, at least to some scouts, they are. Teddy Bridgewater dropped in the draft following a subpar workout, but he has turned out to be a decent NFL quarterback. JaMarcus Russell rocketed to the first overall pick of the draft following a stellar workout as the Raiders overlooked the obvious risks that came up throughout his college career (weight, inaccuracy, etc). Teddy Bridgewater was a worthwhile first round pick. JaMarcus Russell is rightly labeled as one of the biggest busts in the history of the NFL Draft.
How should we value data? I suggest valuing data proportional to the prospect's career. Let's consider a prospect who has started every game of his career and is entering the draft as a senior. That's 4 seasons of 12 games, assuming no Bowl games or time missed due to injury. Just for an ultra simple example, we'll consider a 50-50 split between offensive and defensive game time, and also assume that the player played on special teams. So out of 48 one hour games, our player has put 24 hours of game tape on film. That's a sizeable amount of data.
Now let's consider a combine or workout. The prospect likely puts an hour or two of usable data on film, at most. In reality, the data is significantly less. The significance of that data should therefore be much less than the significance of the available game film. There lies the major problem with our Dick Donlin example. We valued Combine data and game tape data equally, and a poor workout killed Dick's draft stock. Of course, if we reverse the numbers, a strong workout can inflate a player's draft stock heavily even given questionable game film. Is this the proper way to evaluate players? Of course not.
There are multiple problems with putting too much stock in workout and Combine data. A player can train hard for the events in the workout to make up for subpar on-the-field performance, and inflate their own draft stock. Of course, a player who's workout is significantly better than their on-the-field performance may artificially inflate their own draft stock. These players will be over-drafted in the NFL. Another key problem with the workout. A workout is a one-day event. If the player has a bad day, then the workout could artificially deflate his value. Granted, players can work out at both the Combine and at their Pro Day, but still, two days compared to a record of 48 games. I'll take the 48 games over four seasons. Workouts are generally not in full pads, and never show the prospect in a real game-day situation. Therefore, the value of a workout should be seen as significantly less than the value of a player's film record.
Game tape is clearly the most important assessment tool available for demonstrating a player's future NFL potential given the amount of data. The next most important piece of data varies depending on the player. If you're looking at a small school player, Senior Bowl or Shrine Game practice and game performances might show you that the player in question can perform well against top tier talent. If a player has character issues, then interviews are very important. If a player has had medical problems in the past, and your evaluation of him as a draft prospect is dependent on his medical status, then medical examinations are key.
Workouts and Combine data should be at the bottom of the barrel as far as value as an evaluation tool. While these events can measure a player's athletic ability, they measure the player's ability to train for the events in the workout more than the player's ability to play on the football field. When a fast 40 time makes a player rocket up draft boards, don't buy in if the new standard value on the player is significantly higher than it was based on their game tape. The reverse is also true; don't let a poor workout kill your value on a player if their game tape says otherwise. Game tape is always the truest measure of a player's potential as a football player. Combine workouts are useful when comparing similarly ranked prospects, or when looking to draft a player from a different system. For example, if you plan on drafting a DT who played in a 4-3 in college to play nose tackle in a 3-4, you might look for a strong bench press performance to show that he has the upper body strength to handle multiple blockers. However, don't let bench press numbers dictate your ultimate value on the player. If the film looks good, and his bench press looks good, go get him. If the film looks iffy but he bench pressed well, maybe he will be a great fit in your 3-4, but then again, maybe he will struggle like he did in a 4-3. Watch the film and see how he performs against blockers before making your final decision.
My final point here is not to throw out Combine and workout data entirely. Rather, this data should be valued accordingly. It is foolish to over-emphasize a player's workout data at the expense of their game tape. Doing this leads to drafting JaMarcus Russell first overall, and to Teddy Bridgewater nearly falling out of the first round of the draft. Workout data is useful, but it is not nearly as important as a player's game tape. As our quick foray into the Bayes Theorem shows, if one over-values workout and Combine data, their final evaluation of a player can be fundamentally flawed. Don't let that happen to you. Watch the tape, watch more tape, look at workout and Combine data, and then watch the tape again. Don't let a workout warrior trick you into drafting you higher than you should based on film; instead, focus most intently on the film, and then value your other pieces of data based on their ability to shed light on the player's long-term potential in the NFL. --Mike B.
RSS Feed
