Front Office Football Central - View Single Post

MalcPow · 07-14-2008, 09:00 PM

I think what people are getting at in a lot of these posts is the concept of what you're controlling. For all Jim's very astute observations in the initial post about the complexities of football and the inadequacies of its statistics (I strongly agree on both points), FOF is still very much a statistical simulation and its approach to "gameplanning" is heavily skewed by that perspective. Most football games that are played between human beings are about matchups. Teams seek to create favorable matchups that they can exploit. Rookie cornerbacks that they can expose, undersized DE's they can run at, inexperienced QBs that can be deceived and intimidated... these are the data points that matter when teams analyze upcoming opponents and gameplans are created with an eye to taking advantage of these weaknesses or finding ways to create favorable mismatches. Most teams have their general packages and schemes in place (they may work to tweak their formation or personnel usage to disguise things from week to week), but they're basically attempting to leverage their advantages and exploit an opponent's weaknesses. This sort of thing is intuitive to the vast majority of FOF's user base (relatively informed football fans) because they understand that football is far more about getting Randy Moss one on one with an inferior player than it is about throwing "long" 68% of the time on 2nd and 5. I don't want to throw long per se, but I want to throw long (or at least in Moss's direction) whenever I get that kind of favorable matchup.

Along those same lines, it's clear that FOF's percentage-driven approach to playcalling and gameplanning makes for an effective statistical simulation, but it will also never lend an intuitive feeling of control for just that reason. It doesn't model, or pretend to, the key points of information that most people associate with football decision making. I know that a QB's "Timing" rating somehow approximates his effectiveness (or simply his willingness) in taking what the defense gives him and throwing to a better option as the play develops, but of course there's no real tangible depiction of that or sense of the criterion involved so I don't really know what's going on. When in reality, these sorts of decisions are at the core of every game of football. Many schemes are geared entirely around reads at the line of scrimmage or decisions made based on how a safety reacts in coverage, the concept of calling a "long pass" can be pretty foreign in those types of schemes.

So for me, before I ramble on completely, it's not how much, but what kind of control that matters. I want to be able to say let's get Westbrook 25 touches, and let's try to take some shots downfield with Moss, or let's target this defensive back, with a slider for how strictly I want it followed (i.e. keep feeding him the damn ball no matter what/if it's not working early, let it go). I'd also like some general settings that allow me to set a slider for something like conservative/aggressive playcalling (not the right words probably), but something that accounts for the fact that I've got Tom Brady and I'm willing to air it out and take some chances. And the flip side, I've got Joe Rookie out there and we're gonna hand it off on third and long inside our 20. (This is probably as simple as a general modifier setting for each quarterback that activates if they end up in the game.) I realize that the pushback here comes in the difficulty of coding an AI that can handle these kinds of variables, but in reality, i think the current percentage system creates just as many issues. If you shift the thinking toward this kind of matchup-based system, while accounting for some general guidelines for pass and rush frequency, I think you'll find that people can connect more to the control they have over their team.

07-14-2008, 09:00 PM	#13
MalcPow College Benchwarmer Join Date: Jun 2005 Location: San Diego	I think what people are getting at in a lot of these posts is the concept of what you're controlling. For all Jim's very astute observations in the initial post about the complexities of football and the inadequacies of its statistics (I strongly agree on both points), FOF is still very much a statistical simulation and its approach to "gameplanning" is heavily skewed by that perspective. Most football games that are played between human beings are about matchups. Teams seek to create favorable matchups that they can exploit. Rookie cornerbacks that they can expose, undersized DE's they can run at, inexperienced QBs that can be deceived and intimidated... these are the data points that matter when teams analyze upcoming opponents and gameplans are created with an eye to taking advantage of these weaknesses or finding ways to create favorable mismatches. Most teams have their general packages and schemes in place (they may work to tweak their formation or personnel usage to disguise things from week to week), but they're basically attempting to leverage their advantages and exploit an opponent's weaknesses. This sort of thing is intuitive to the vast majority of FOF's user base (relatively informed football fans) because they understand that football is far more about getting Randy Moss one on one with an inferior player than it is about throwing "long" 68% of the time on 2nd and 5. I don't want to throw long per se, but I want to throw long (or at least in Moss's direction) whenever I get that kind of favorable matchup. Along those same lines, it's clear that FOF's percentage-driven approach to playcalling and gameplanning makes for an effective statistical simulation, but it will also never lend an intuitive feeling of control for just that reason. It doesn't model, or pretend to, the key points of information that most people associate with football decision making. I know that a QB's "Timing" rating somehow approximates his effectiveness (or simply his willingness) in taking what the defense gives him and throwing to a better option as the play develops, but of course there's no real tangible depiction of that or sense of the criterion involved so I don't really know what's going on. When in reality, these sorts of decisions are at the core of every game of football. Many schemes are geared entirely around reads at the line of scrimmage or decisions made based on how a safety reacts in coverage, the concept of calling a "long pass" can be pretty foreign in those types of schemes. So for me, before I ramble on completely, it's not how much, but what kind of control that matters. I want to be able to say let's get Westbrook 25 touches, and let's try to take some shots downfield with Moss, or let's target this defensive back, with a slider for how strictly I want it followed (i.e. keep feeding him the damn ball no matter what/if it's not working early, let it go). I'd also like some general settings that allow me to set a slider for something like conservative/aggressive playcalling (not the right words probably), but something that accounts for the fact that I've got Tom Brady and I'm willing to air it out and take some chances. And the flip side, I've got Joe Rookie out there and we're gonna hand it off on third and long inside our 20. (This is probably as simple as a general modifier setting for each quarterback that activates if they end up in the game.) I realize that the pushback here comes in the difficulty of coding an AI that can handle these kinds of variables, but in reality, i think the current percentage system creates just as many issues. If you shift the thinking toward this kind of matchup-based system, while accounting for some general guidelines for pass and rush frequency, I think you'll find that people can connect more to the control they have over their team.