IrishTitansFan Posted January 14 Report Share Posted January 14 8 minutes ago, Pragidealist said: No- Its just one factor to consider. Then put those in context. Stats and analytic based approaches are the norm now. This isn't really worth debating. I promise you no teams are focusing on completion\TD/INT % or anything like that. There are some useful stats like P2S%, under pressure vs clean etc but not the generalisations from boxscore stats Justafan 1 Link to post Share on other sites More sharing options...
Mythos27 Posted January 14 Report Share Posted January 14 3 minutes ago, Justafan said: Kenny Picket in his senior year threw for 4319 yards, 42 TDs & INTs. 66% completion and 8.5 yards per attempt. According to Prag's chat GPT argument, the Titans would have been smart to take him #1 overall! Now I'm interested to see what chatgpt says about the previous like 10 drafts. Pragidealist 1 Link to post Share on other sites More sharing options...
Pragidealist Posted January 14 Report Share Posted January 14 Just now, Mythos27 said: Now I'm interested to see what chatgpt says about the previous like 10 drafts. Ok... I have a bit more time today than normal- not sure I have THAT much time. lol. Mythos27, and IsntLifeFunny 2 Link to post Share on other sites More sharing options...
Pragidealist Posted January 14 Report Share Posted January 14 1 minute ago, IrishTitansFan said: I promise you no teams are focusing on completion\TD/INT % or anything like that. There are some useful stats like P2S%, under pressure vs clean etc but not the generalisations from boxscore stats They are- but are likely using very complicated and specific algorithms to weight things. Link to post Share on other sites More sharing options...
IrishTitansFan Posted January 14 Report Share Posted January 14 1 minute ago, Mythos27 said: Now I'm interested to see what chatgpt says about the previous like 10 drafts. Max Duggan QB1, most likely I don’t think you could accurately check what it would’ve said about previous drafts after the fact Justafan, and Mythos27 2 Link to post Share on other sites More sharing options...
Mythos27 Posted January 14 Report Share Posted January 14 Just now, IrishTitansFan said: Max Duggan QB1, most likely I don’t think you could accurately check what it would’ve said about previous drafts after the fact Right. I forgot it scours the current internet. Link to post Share on other sites More sharing options...
IrishTitansFan Posted January 14 Report Share Posted January 14 Just now, Pragidealist said: They are- but are likely using very complicated and specific algorithms to weight things. They will be charting their own statistics or using agencies to get stuff like on target %, they will not be using box score statistics Mythos27, and Justafan 2 Link to post Share on other sites More sharing options...
Pragidealist Posted January 14 Report Share Posted January 14 Just now, IrishTitansFan said: Max Duggan QB1, most likely I don’t think you could accurately check what it would’ve said about previous drafts after the fact The problem is the data. Its mostly scanning the internet. If I went and put together a data set for it to use- it could. Mythos27 1 Link to post Share on other sites More sharing options...
Pragidealist Posted January 14 Report Share Posted January 14 Just now, IrishTitansFan said: They will be charting their own statistics or using agencies to get stuff like on target %, they will not be using box score statistics I imagine they use it all in a predictive algorithm with stats like completion percentage being weighed lower in importance. Link to post Share on other sites More sharing options...
Popular Post oldschool Posted January 14 Popular Post Report Share Posted January 14 14 minutes ago, Mythos27 said: @Pragidealist The Chatgpt stuff is interesting. Obviously we wouldn't make decisions based on it but it's definitely interesting to look at as an aggregate. You can't trust chatgpt though. it fucks up simple two step equations all the time. Asking it to aggregate stats without giving it the source to generate the stats from i.e. profootballreference.com... even then you don't know if its actually calculating the data correctly. Its a Garbage in Garbage out scenario. Pragidealist, rns90, Atlas, and 4 others 6 1 Link to post Share on other sites More sharing options...
titanruss Posted January 14 Report Share Posted January 14 7 minutes ago, Pragidealist said: Analyzing quarterback performance against top-tier competition provides valuable insights into their potential at the professional level. Let's compare Cam Ward's 2024 season statistics against ranked opponents with those of Caleb Williams, Drake Maye, and J.J. McCarthy from their respective seasons. Key Metrics for Evaluation Completion Percentage (Comp%): Measures passing accuracy. Yards per Attempt (YPA): Indicates efficiency per pass. Touchdown-to-Interception Ratio (TD:INT): Reflects decision-making. Sack Rate: Percentage of dropbacks resulting in sacks; lower rates suggest better pocket presence and processing speed. Cam Ward's Performance Against Ranked Teams Cam Ward has faced ranked opponents multiple times in his career. Here's a summary of his performance: Games Played: 10 Passing Yards: 2,852 Touchdowns: 21 Interceptions: 6 Passer Rating: 149.4 Note: Detailed statistics such as completion percentage, yards per attempt, and sack rate specifically against ranked teams are not readily available. Caleb Williams' Performance Against Top 25 Defenses Caleb Williams has had notable performances against top 25 defenses: Completion Percentage: 51.4% Touchdown-to-Interception Ratio: 1:1 Touchdowns: 6 Interceptions: 6 Note: These statistics highlight challenges faced by Williams against top-tier defenses. Drake Maye and J.J. McCarthy Specific game-by-game statistics for Drake Maye and J.J. McCarthy against ranked opponents are not readily available in the provided sources. Conclusion While Cam Ward has demonstrated solid performance metrics against ranked opponents, including a passer rating of 149.4, the lack of detailed statistics such as completion percentage, yards per attempt, and sack rate limits a comprehensive comparison. In contrast, Caleb Williams has faced challenges against top 25 defenses, with a completion percentage of 51.4% and a 1:1 touchdown-to-interception ratio. Due to the unavailability of specific data for Drake Maye and J.J. McCarthy, a complete comparative analysis is constrained. Sources: Cam Ward vs. Ranked Teams Caleb Williams' Numbers Against Top 25 Defenses Note: The availability of detailed statistics varies, and some data may not be publicly accessible. Sources 4o *shrugs* The stats may still hold.. but you do realize theres a difference between "ranked teams" and "top 25 defenses" right? Link to post Share on other sites More sharing options...
Pragidealist Posted January 14 Report Share Posted January 14 1 minute ago, oldschool said: You can't trust chatgpt though. it fucks up simple two step equations all the time. Asking it to aggregate stats without giving it the source to generate the stats from i.e. profootballreference.com... even then you don't know if its actually calculating the data correctly. Its a Garbage in Garbage out scenario. Absolutely... but its fun to play around with. Link to post Share on other sites More sharing options...
Pragidealist Posted January 14 Report Share Posted January 14 NFL analytics teams use a combination of predictive algorithms, statistical models, and machine learning techniques to evaluate prospects. These approaches aim to quantify a player's potential for success in the NFL by analyzing performance, physical traits, and situational factors. Here are some of the primary methods and tools used: 1. Regression Models Purpose: To predict NFL success (e.g., passing yards, Pro Bowl selections) based on college performance. Key Variables: Completion percentage, yards per attempt, and touchdown-to-interception ratio for quarterbacks. Speed, agility, and strength metrics for skill positions. Types of Models: Linear Regression: To assess how a single variable (e.g., YPA) correlates with NFL performance. Logistic Regression: To predict binary outcomes (e.g., "Will this player make a roster?"). 2. Machine Learning Purpose: To uncover complex patterns and interactions in large datasets. Examples: Random Forests and Gradient Boosting: To rank the importance of various metrics (e.g., arm strength, pocket presence) in predicting success. Neural Networks: Used for more nuanced predictions, such as evaluating the likelihood of injury based on biomechanics. 3. Performance Comparisons Similarity Algorithms: K-Nearest Neighbors (KNN): Identifies past players with similar profiles (e.g., size, college stats) to compare trajectories. Example: Comparing Cam Ward to NFL players with similar completion percentages and mobility metrics. Player Archetype Matching: Matches prospects to existing NFL archetypes (e.g., "Drew Brees-like accuracy" or "Josh Allen-like arm strength"). 4. Success Threshold Analysis Purpose: To define minimum thresholds for key metrics associated with NFL success. Examples: QBs with a college completion percentage below 60% historically have lower NFL success rates. 40-yard dash times under 4.5 seconds for wide receivers strongly correlate with higher draft stock. 5. Contextual Data Models Purpose: To adjust for external factors influencing performance. Examples: Adjusting passing stats for offensive line quality or strength of schedule. Incorporating weather, altitude, and field conditions into performance metrics. 6. Bayesian Inference Purpose: To update the probability of a player succeeding based on new data. Example: Starting with a prior assumption (e.g., QBs from spread offenses struggle in the NFL) and updating it based on how a specific player performed against pro-style defenses. 7. Injury Risk Models Purpose: To predict injury likelihood based on historical and biomechanical data. Methods: Motion-capture analysis to assess joint stress and alignment. Statistical correlations between college workload and NFL durability. 8. Decision Trees Purpose: To simulate draft outcomes and optimize selections. Example: If a top QB is off the board, which position should the team prioritize based on value and need? 9. Situational Metrics Purpose: To evaluate performance in specific contexts. Examples: Performance on 3rd-and-long or in the red zone. QB performance against blitz-heavy defenses or top-ranked teams. 10. Combine and Pro Day Data Integration Purpose: To incorporate physical testing results into evaluations. Examples: Combining 40-yard dash, shuttle times, and vertical jump to predict explosiveness for wide receivers. Using hand size, release time, and velocity for QB evaluations. Real-World Applications The S2 Cognition Test: Measures cognitive processing speed and decision-making in quarterbacks. Used as a predictive tool for evaluating mental processing in high-pressure situations. PFF Data: Teams integrate Pro Football Focus metrics, such as pressure-to-sack ratios and CPOE (Completion Percentage Over Expected). Game Simulation Models: Simulating how a player’s skills translate to an NFL system based on scheme and coaching. Challenges and Limitations Small Sample Sizes: College data is often limited, especially for players in non-Power 5 conferences. Scheme Dependence: Performance in college systems doesn’t always translate to NFL schemes. Unquantifiable Traits: Leadership, work ethic, and adaptability are harder to measure but critical for success. Its a complicated area... kinda fun to explore but I am in no way trying to do what they do. I'm messing around during a slower day... doing the type of stuff I used to spend way too much time doing in my 20's doing on here. . Just doing it a lot faster. Just for interesting conversation Link to post Share on other sites More sharing options...
Pragidealist Posted January 14 Report Share Posted January 14 5 minutes ago, titanruss said: The stats may still hold.. but you do realize theres a difference between "ranked teams" and "top 25 defenses" right? I do but chat didn't. I just didn't bother to correct it. Its that stuff that can be a time drain when using or playing with it. Link to post Share on other sites More sharing options...
Justafan Posted January 14 Report Share Posted January 14 14 minutes ago, Mythos27 said: Now I'm interested to see what chatgpt says about the previous like 10 drafts. If you give it the right prompt, you can get it to say virtually anything you want. Mythos27, and titanruss 2 Link to post Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.