OCC had a great article that noted that one can get more information about our players by looking at their performance over time. His methodology was simple; he divided the season into two eight-game groups. He then compared players’ PFF scores for those first eight games to the second game. As he said
The Cowboys record of the final eight weeks suggests that the Cowboys played better football in the second half of the 2012 season. It also stands to reason that some players were better in the second half of the season than they were in the first half.
His thesis was that the season cumulative grades can hide the improvement of the players over the season. Often, we see only the final team record and player scores and declare them hits of misses. This is good for determining who gets to go the playoffs, but not so good for evaluating where our needs for the future lay.
I commend OCC for this article to show trends. Trend analysis is an important tool. Not only can we see the trends over time, but can look at specific inflection points, where the trend line changes, to help us find true causes for further analysis.
Here we can use Bernadeau as an example. Bernadeau’s season can be described in four parts – not just two. Let’s look at Bernie’s season.
His overall score is misleading. One, his ranking at guard is based on the cumulative scores AT GUARD, so missing games cut his scores down. He had only 14 rated games at guard but is compared to players who had 16 games rated at guard. Technically missing one or more games provides an opportunity to raise or lower his score. Yet, given how he was playing at the time, he probably would have raised his score.
Two, he was terrible at center. He had learning pains in his first two games at center in the NFL. Just like the Tyronasaurus had two terrible games in his first games in the NFL at LT, this may not be the final answer to how well Bernie could actually play at center in the future. Yet I think his natural position, along with our depth at center, means that Bernie is a guard.
Then, just like reading a book and getting interrupted, you don’t go back to pick up exactly where you left off. He was bad at center and it took a while to return to form at guard when he moved back. Yet every game at RG after his return was better than the game before and he ended with a last game of 2.0, which is much better than average.
Part one – open the season after recovering from off-season surgery – BAD and memorable for everyone just like Costa.
Part two – healthy at RG – major improvements and what the scouts expected
Week 4….-1.4……2.9…not yet fully healthy but major improvement
Week 9…..1.3…..-1.8 but still noteworthy good
Part three - move to center and the return to guard – bad
Week 11…..-3.5…………not comparable as in a new position
Week 13…..-2.8…………again not comparable in a new position back to guard
Week 14…..-1.2…..1.6…huge gain in performance though still bad
Part four – return to guard – start slow and improve every game to end at a high level again
Week 15…..-0.7…..0.5…below average but still a major gain in performance
Week 16…..-0.3…..0.4…below average but still a major gain in performance
Week 17…...2.0…..2.3…again at a high level
The key is that I divided the season into groups AFTER the season was over. The change is performance was a key to ask what happened to cause the change. In retrospect, it was obvious that he was not really healthy at the start of the season. As he got healthier, so did his performance.
Then, his performance was dramatically changed when he shifted positions. It is important to note that his scores at center and guard are so different. Yet one should not compare PFF scores across different positions, as they are normed by their positions.
Dividing the season into halves, thirds, or quarters or any other number has the advantage of providing a way to measure change over the season. Yet, the division is rather arbitrary.
One criterion is to have a large enough sample size so that the statistics are stable and relatively immune from extreme results or outliers. Bill Parcells, no stat whiz, said four games or so stabilized the results.
Another is ease. Dividing the season into halves or quarters is relatively simple 16/2 and 16/4 get 8 and 4 games respectively. Dividing seasons into thirds is more difficult, but we can use per game statistics to smooth out the differences of 16/3 or 5+ games.
Yet another important factor one can us is divide the season based on important events. One such event is major injury. Every player has his own cycle, but we also can note important events that affect major units.
Here we have a good way to divide the last season based on who was the center. This obviously affected Bernie, but also the rest of the OL a whole.
We used Cook as the starting center. This is important as Cook was the 3rd string center obtained in a last minute trade. He did not know the playbook or the terminology.
We got Costa back for a game and a half, and then we used Cook for another two games until he was injured. We used Bernadeau as the 4th string center. The hallmark of the middle of the season was a lack of continuity. No three games in a row with the same OL.
We had the game with Smith also out during this spell and used Parnell. The interior of the OL was Livings, Bernie and Dockery. Livings, who was normally the rock but was between a project backup LT in Parnell and a 4th string center who had never played at the position in the NFL. Dockery is just bad.
Parnell seemed to do well. Free did even better than Parnell. Yet the middle of the OL was so bad that the D Coordinators feasted on them and the tackle scores should be evaluated accordingly.
We got Cook back. We can note the following for the OL. Note these are the OL stats; there were other sacks, hits etc. that can be attributed to other players. I used the PFF stats for attribution.
Sacks/game…..01.8…….02.3……..01.8…..middle of the season inflates the total
We get more information than just the final scores bu using trend analysis. We can track their performance over time and see increases or decreases. Further, we by looking at specific points where the performance makes a major change we can use that as an indicator for more investigation to determine why the change occurred,
We should note that the OL, as a unit, was much better by the end of the season, compared to early in the season. This should bode well for our off-season evaluation.
The key is to get the starters healthy and practicing together. They did not play as a unit until the last week before the opener. That lasted for 3 snaps. The OL took some time to adjust, but they did. They were playing well until Cook went down with another injury. Yet again they adjusted over time.
Next season we should have basically the same OL for the first time in years. We can work on continuity more than evaluation. Further we have some additional young players for solid depth in Arkin and Leary before we add anyone else in FA or the draft.