A look at the results of the NFL draft from 1990 to 2009. The Power of Statistics I need some questions answered. All season long we’ve been trying to make sense of Mike Glennon’s statistics and play. Some have argued there great some not. We’ve thrown out common phrases like "The best way to learn is to play" and "all rookies make mistakes". We’ve thrown out statistical comparisons on Mike Glennon and debated how they compare to Andrew Luck or Russell Wilson; we’ve talked about draft position and the power of perception. We’ve talked about statistical manipulation and small sample sizes. I still believe game film diagnosis is the best way to evaluate, but I also recognize the power of a good, meaningful statistical study. Through all this I’ve still had questions because I wanted to know what the statistically relevant information says and I honestly didn’t know the answer because I have not found an empirical study that is on point. So I decided to build one, by asking myself the questions first, developing a statistical hypothesis and then trying to prove or disprove that point. My Questions: Do rookie statistics have any correlation to how a QB will perform over their NFL career? If so is there an average measure of lifetime improvement in terms of completion percentage, TD-INT ratio, or Yards per attempt? Does "sitting" for one or more seasons indicate that a QB is more likely to be successful? My Methodology I needed to produce a large stratified survey which requires three things: 1. A large relevant sample size 2. Objective standards 3. A minimum confidence level of .95 and falsifiable null hypothesis For a large relevant sample size: I’ve gone with Career passing numbers as they apply to all Quarterbacks drafted between 1990 and 2009. For relevant numbers I then sub-selected all QBs who made at least 5 starts in one season, within 5 years of entering the league. Objective Standards: While I love my tier system and its been handy for making quick comparisons, for this I wanted something statistically based (it is a statistical comparison after all). Range A-G (7 Groups) Group A (The Burger Flippers)– Fails to meet any below criteria Group B(The Insurance Salesmen) – At least 4 Seasons on an NFL roster Group C(The Future Coaches)- At least 4 seasons on an NFL roster with career yardage greater than 3,000 Group D(Starters)- At least 4 seasons on an NFL roster, with 60 career starts Group E(FranChosen)- At least 4 seasons on an NFL roster, with 60 career starts and 4 seasons where the player started no fewer 10 games while throwing for more TD’s than INT’s Group F(Franchish)-At least 4 seasons on an NFL roster, with 60 career starts and 4 seasons where the player started no fewer than 10 games while throwing for more TD’s than INT’s and a completion percentage above 60% Group G (The Titans)- At least 4 seasons on an NFL roster, with 60 career starts and 4 seasons where the player started no fewer than 10 games while throwing for more TD’s than INT’s and a completion percentage above 60% AND being selected league MVP or 2 time 1st Team All-Conference or 4 Pro Bowl selections(non-injury fill in) The Findings To keep this article as neutral as possible, I’m merely giving the results, the results from a statistical method cannot be disputed (unless you challenge the methodology) , however as with all statistics the what they mean and how they apply is open to interpretation and I invite everyone to propose their own "what they mean". Overview: 224 QBs were selected between pick numbers 1-260 (The Modern Draft) 36 Started Five or More Games as a rookie 67 of them started five or more games(in the same season) within 5 years of entering the league 46 of them were first round picks 103 of the 224 had enough starts in their career to produce significant correlations Only 9% (Less than 1 in 10) managed to belong to the top 2 groups that every franchise wanted Group Number % of 224 Burger Flippers 70 31% Insurance Salesmen 59 26% Future Coaches 45 20% Starters 18 8% FranChosen 12 5% True Franchise 13 6% The Titans 7 3% Round of Pick Breakdown 1 46 2 21 3 23 4 32 5 25 6 31 7 36 All QBs (Throughout their career) Career Completion Percentage did not change in a statistically meaningful way, only 9 players showed a baseline change outside of double the standard deviation (statistically significant) in career completion percentage, 8 improved 1 regressed – (More on this later the ones who did improve were important) Almost all studied experienced a change in their TD – INT ratio. While the median change was an improvement of just .00641; of the 103 who had enough starts to be studied, 51 showed significant improvement and 25 showed significant regression. Yards per attempt deviated almost as much as TD-INT ratio, 71 deviations were recorded (69%), with 48 showing meaningful improvement and 23 regressing. Rookies vs. Those who sat and waited Rookie QBs were more likely than others to show improvement over the course of their career in all three categories than QBs who waited to start. Rookies showed slightly statistically significant improvement in Yards Per Attempt and Completion Percentage, and dramatically likely to improve their TD-INT ratio. QBs who waited showed, as a group, flatline statistics. QBs who started as rookies were more likely to reach 60 NFL career starts but slightly (though still significantly) less likely to reach the Elite groups represented by G&F above (20% vs. 17%). Players whose first year as a starter came after their rookie season, did not as a whole improve over the course of their career, they did however play much better as a group than rookies in their first year. Inaugural Season Career Rookie Waited Rookie Waited COMP % 53.8% 56.8% 57.1% 57.2% TD-INT 0.9 to 1 1.3 to 1 1.1 to 1 1.2 to 1 YPA 6.2 6.6 6.5 6.6 Failures (Groups A,B,C) versus Okay(D&E) versus Best(G&F) The "Best" in Groups G&F showed two significant differences: They had dramatically higher YPA scores than their peers (most from their first season) and they showed abnormally high improvement levels from their first year to their career numbers. Of the 8 players referenced above for extraordinary improvement in career completion %, 4 of them were in the Elite groups(Both Mannings, McNair & McNabb). Half of those drafted in the Best Group were selected at or before the 11th overall pick (10 of 20). The lowest drafted player to make it into one of those groups was former Tampa Quarterback Brad Johnson (pick 227 , then Round 9). Of the 46 QBs taken, 4 failed to start 5 games in any one season within 5 years of entering the league (included Tommy Maddox) , 15 fell into groups ABC, 15 in D&E, and 12 in groups G&F. The most common attribute of those who started 60 games but failed to excel (groups D&E) was a lack of overall improvement. In fact yards per attempt, as a whole, slightly decreased. They also showed a lower YPA average than QBs in groups F&G but higher than ABC. The most common attribute from members in groups AB&C was regression in their ability to protect the football. They were the only group that had, as a whole, a decline in YPA and TD-INT ratio. The only group members who showed significant statistical improvement and have failed to get to 60 starts were Craig Erickson (Inconsistency), Quincy Carter(Drugs), and Kevin Kolb(Injuries). Groups First Year Career Comp% YPA TD-INT Comp% YPA TD-INT AB&C 54.5% 6.22 1.03 to 1 55.0% 6.21 .87 to 1 D&E 55.4% 6.55 1.12 to 1 57.5% 6.64 1.23 to 1 F&G 59.8% 6.96 1.68 to 1 62.6% 7.35 1.81 to 1 Conclusions Some of this I found quite illuminating. I knew Football Outsiders had called YPA the best enduring statistic for QB success. I certainly have to agree given the difference over time. That said I found the improvement of several players quite interesting, equally interesting is the long list of those who failed to improve over time. Those who showed promise but never quite fulfilled it. Obviously the best chance to find a franchise QB is still in round 1, particularly in the first 11 picks of round 1. That doesn’t mean other quality franchise guys will not come from later in the draft but that development and long term improvement, in addition to YPA are crucial for success. At this time I close my usually opinionated self and invite you , dear reader, to provide your own conclusions based on the analysis presented. Source: Continue reading...
