Ultimate22 takes on the Dallas review

The ‘All 22′ review of games, also known as Ultimate22, was given an introduction yesterday on The NY Times Fifth Down Football Blog.  The limitations of making a grade within a single play and the macro statistical advantage of evaluating large numbers of plays were discussed in the comments section at greater length. 

Here is the Reader’s Digest summary of these considerations:

(1) Get over the idea that any single play is always going to get a perfect evaluation.  The only way that is possible is if we ourselves are the coaches and know precisely what was drawn up and intended for execution.  And if you want to have another laugh, just know that Lawrence Taylor went rogue on Rod Rust and had Steve DeOssie make a pre-1991 defensive call out of frustration with the ‘read and react,’ so nothing in this world is 100%, not even the coaches.

2) Appreciate the value of large numbers to get a respectable report on the contributions of individual players overall.  Would you rather get a snapshot of 5 plays from an Offensive Lineman or 55?  When you have 55 plays to observe contributions, the body of work is much more representative of the evening’s game and that player.  We never said that we’d be able to get all 55 perfect.  But we do believe that our conclusions overall will be solid and meaningful.

Here is a question for the entire viewership to think about.  What percentage of the time does a player fail to execute his assignment, ie a lineman is told to block Player A and blocks the wrong guy?  How often does that happen?  Well, let’s think about the issue from a different perspective…  In the 11 person choreographed ballet, which is also known as a single football play, we know that everyone is counted on to do their job or else the play may necessarily fail. So how often do all 11 players on any play ALL do their job?  Is it 10% of the time? Is it 50% of the time?  Is it 90% of the time? 

Below is a table which has the % chance that a single player will do his job on a given play, and the % chance of all 11 players do their job, assuming each player has the same ability.

% chance of a single player doing his job on a single play % chance of all 11 players doing their job on a single play
99% 90%
95% 57%
90% 31%
80% 9%
70% 2%
60% 0%
50% 0%

What percent of the time are all 11 players doing their job on a single play?  The reason why this table was put together was to show you that we would be watching horrible ugly football if only 30% of the time all 11 players were doing their job.  It would be the Keystone Cops.  Using this analysis, we would believe that all 11 players are doing their job on any given play roughly 50% to 90% of the time.  And that implies that any single player is doing his job about 95% to 99% of the time.  This makes sense.  So if there are 60 snaps for a player, this means that we would expect him to be executing his assignment properly on about 57 to 59 of those 60 snaps. (Note: we are not saying they will do their jobs well on 58 snaps, just that they will do what they are supposed to be doing, ie blocking the right guy.)  As professionals, this makes sense again.  We are humans.  We make some mistakes, but hopefully not too often.  We try to limit them by practicing them again and again so that we reduce their occurrence. 

Summary- Assuming players execute their proper assignment ~95% to 99% of the time, that is good enough for us to make informed evaluations over large groups of observations.

This weekend and early next week we will be reporting on how the players did individually and as a group vs Dallas.  The All 22 video feeds only became available late last evening, so it will take a couple of days for the staff to go through all the plays. 

We are expecting some surprises. It is impossible for the human eye to watch more than a few things on a single play, let alone the details of 11 independent things.  Let’s see what the results are.

    

Quantcast