Wednesday, April 13, 2016

Mocking Mock Drafts

These mock drafts are hilarious, really. They're fun and they get eyeballs on the screen or perusing the newspaper, I understand that part. But they are seriously flawed from two important perspectives. First - the mock drafters put their draft together purely on the basis of the need of the team, or at least their perceived need of the team. The teams on the other hand, at least pay lip service to the philosophy of drafting the BPA, best player available. On the surface, these approaches directly conflict with each other. Occasionally at the top end of the draft, the team's needs coincide with the BPA and everyone is happy. On the other hand, I think the teams always tint their player evaluations with team need, or with some other subjective measure. For example, the best QB in the draft will be more important and will have a higher grade than the best LG in the draft. But player evaluations are more complex than that, because it is not only a question of who is the best player, or what is the ranking of all the players at a particular position. It is also a question of whether the player is NFL ready. In one year it could be a very weak draft for WR, for example. Obviously, the best WR in that draft may not warrant a high first round pick. Each team puts together their draft board based on quality of player, importance of position, overall grade compared to the ideal player at the position and they must sprinkle in at least some part of team need. In the early rounds of the draft, the contribution of team need plays less into the decision making, because the team really is looking for a star, regardless (almost) of position. In the later rounds, when the players are more bunched together in talent and the team is less likely to find a star, team need will factor more into the selection. If you believe the teams, they use these factors to calculate a score for each player and list them in order from highest to lowest. They take the best player left on their board whenever their turn comes up. Last year, Giants needed a S and they loved Collins, who was high on their board. It was a moment when need coincided with the draft board, so they traded up to get him in the 2nd round.

Two great stories about drafting BPA vs. team need.

In 1980, Giants were coming off an awful season and had the second overall pick in the draft, behind New Orleans Saints. The Giants were just an awful team, no offensive talent to speak of - they did have Phil Simms at QB, though he was going through his early career injury period, but their OL and skill position talent was really poor. On defense they had a player or two on the DL and in the secondary, George Martin and Mark Haynes to be specific, but that's it. What they did have was three outstanding LBs in Harry Carson, Brian Kelly and Brad Van Pelt, which was the only strong unit on the team. With that second overall pick, the fans, the TV/radio media and the print media were campaigning for the Giants to pick some offensive talent or perhaps some defensive players on the DL and in the DB-field, but not a LB. The first pick was by Bum Phillips, newly appointed coach / GM of the Saints and he took George Rogers, Heisman Trophy winner, RB out of South Carolina. George Young picked Lawrence Taylor and the rest is history. George Young said that the fastest he has ever run was to get the pick of Taylor up to the podium.  LT was clearly the BPA and Young picked him despite having the only real talent on the team at his LB position. Phillips, on the other hand, made a need pick with Rogers - he believed that the RB was the most important position on the team, having just coached Earl Campbell in Houston. How'd that work out for those teams?

Second story is about the Dallas Cowboys and their legendary HC, Tom Landry. He was famous for drafting good players and for being faithful to the best player available philosophy. He said that he deviated from this strategy just once - in the 1979 draft - and it cost him greatly. He had an aging Roger Staubach at QB on the roster, who would retire after the 1979 season. He also had the QB in waiting, the QB of the future on the roster, a young good QB named Danny White. Consequently, he deviated from his BPA philosophy just this once and drafted some nondescript players in first few rounds. The BPA, the player he passed up because he had his QB of the future already on the roster was Joe Montana. 'Nuff said.

Monday, April 4, 2016

Statistics and Injuries

Those of you who know me know that I am interested in mathematics and statistics and will occasionally look at them in order to prove a point or guide a discussion. In the e-world, many Giants fans have been pointing to the strength and conditioning program and how the many injuries that the Giants have had are entirely the fault of the S&C program not keeping up with modern training techniques. The flip side is the Coughlin perspective who said over and over again that the piling up of injuries is just bad luck. The truth probably lies somewhere in between - you certainly cannot blame every injury that occurred on a flawed S&C program. Football is a violent sport, with heavy collisions that put a medical and orthopedic strain on the body. No doubt injuries will occur no matter how well conditioned the athletes are. Having said that, there are certainly training programs that can limit or contain some of the injuries, especially those that are not direct result of collisions and are more muscular and soft tissue injuries. Last year, after Beatty tore his pectoral muscle while working in the weight room, Coughlin instituted some changes in the program. There were "spa days" or recovery days, they used GPS devices to track the number of hits and the force of the collisions that each player experienced during practice, but still the injuries continued. (Note: later Beatty tore the rotator cuff in his shoulder while rehabbing from his pectoral injury.) The changes that Coughlin instituted after the Beatty injury, I believe were effective in reducing injuries. But they didn't address the core problem - a strength and conditioning program that was mostly focused on the "strength" part of the program and not enough on the "conditioning" part. Core strength, balance, flexibility, long lean muscles, efficiency of movement and pure conditioning are more important than how much weight the player can lift. Strength is important, but not at the expense of conditioning. Better conditioning contributes not only to reduced injuries but also to more effective play on the field.

Now to my statistical analysis of why this injury bug just can't be luck. Football Outsiders tracks a statistic called adjusted games lost (AGL) for each team. The reason it is adjusted games lost and not just total games lost is that they adjust the number on basis of which player was lost. They value loss of a starter or regular rotation player more than a bench player, rendering the statistic a little more meaningful. The Giants finished last of all 32 teams in the league in this AGL statistic for the last 3 years, something that's very hard to do. But it's too easy to simply say, it's unlikely and that this could happen by chance without putting some hard measurements around it. Here goes.

I looked at the AGL statistics for the last two years, 2015 and 2014. I made the fairly safe assumption that the  AGL sample for each year was normally distributed. I then calculated the average and standard deviation for each year. The standard deviation is a measure for how widely spread the measurements are around the mean. For example, if a student's average score on two tests that he took is 80, he may have gotten 75 and 85 on his two tests or may have gotten scores of 100 and 60. In both cases the average is 80, but in the second case the standard deviation is greater because the scores are more widely spread around the mean. For a "normal distribution" 99.7% of all the possible measurements are within 3 standard deviations of the mean. In 2015 the Giants' AGL was more than 3 standard deviations away from the mean and in 2014 it was slightly less than that, about 2.59 standard deviations away. Using these statistics and the normal distribution, you can calculate the probability that a team will get a sample score less than (or greater than) a particular number.

OK ---- for those of you whose eyes are glazed over with these boring statistics, let me turn this into English. The probability that by pure chance a number would be less than or equal to the Giants AGL for year 2015 is 99.88%. The same probability for year 2014 is 99.5%.
Looking at it another way, the probability that a random score would be greater than the Giants for 2015 and 2014 are respectively: .119% and .482%

Putting these together - the probability that any team could do as badly as the Giants in both years together is .000574%

Understanding this better - that is 5.7 chances out of a million.
(With apologies to D&D, quoting Lloyd Christmas... " so you're saying there's a chance")

Looking at the second worst team in the two year period, the Washington Redskins were .38%

In other words, Giants were roughly 1,000 times worse than the second worst team in the league over the 2 year period.

Here's hoping the new S&C coach does a little better than the last one.