Home
Articles
Photos
Interviews
Forums
New posts
Search forums
Georgia Tech Recruiting
Dashboard
What's new
New posts
New profile posts
Latest activity
Chat
Log in
Register
What's new
Search
Search
Search titles only
By:
New posts
Search forums
Menu
Log in
Register
Install the app
Install
Home
Forums
Georgia Tech Athletics
Georgia Tech Football
Your best guess....What say you about Miami game?
JavaScript is disabled. For a better experience, please enable JavaScript in your browser before proceeding.
You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an
alternative browser
.
Reply to thread
Message
<blockquote data-quote="IEEEWreck" data-source="post: 351666" data-attributes="member: 617"><p>I sympathize with the sky is falling kind of folks, sort of. I experience plenty of heart-wrenching terror that sticks around until I know the circuits I designed have passed burnout overvoltage tests, but I guess I just use it all up at work.</p><p></p><p>That's human nature, right? You come up with the reasons things might be terrible, and you focus on them to mitigate risk. Sometimes it works well. I'm not sure why I'd examine @miami scores since 2009, but then anxiety is not rational, right? </p><p></p><p>I humbly suggest that football might be more fun if we treat it like a brain teaser, and engage ourselves with a sort of hobbyist Bayesian approach. You might find yourself driven to closer observation that makes the exercise more fun, and you might find the whole thing less anxious in general. Here, I'll use my thoughts on Miami as an example.</p><p></p><p>Using the time honored tradition of pulling things out of my *** (known in academic papers as heuristic methods for generating prior probability distributions) let's lay some ideas down:</p><p></p><p>Offense:</p><p>We've looked pretty good. Blocking is excellent, and we've gotten consistent 10 yard gains from basic penetration (i.e. the line and linebackers were blocked, but either a farther back defender got in or a jumble at or near the line prevented a footrace. Not, that I've seen, has there been a lot of 'well, the play was executed correctly, but there wasn't a lot of space/a key block on the second level was missing and we consistently have to smash some mouth for 3 yards. If I were serious about this, I'd go work with LongestDay's stuff and try to actually quantify the probability of yardage gains given all blocks made, X blocks made but Y blocks missed, etc, as well as the frequency of those blocking situations. But I'm just some smug EE on the internet, so I'm gonna take wild guesses instead, as so:</p><p></p><p>45% of plays have good primary and secondary blocking, or are completed midrange passes with +-2 sigma yardage in the 6-14 yard area.</p><p>35% of plays have good primary but broken secondary blocks, with +-2s in the 3-5 yard area.</p><p>12% of plays are missed reads or other (successfully predicted counter plays like the rocket toss or called dive) that result in +-2s in the -2 to 3 yards area.</p><p>8% of plays either have perfect blocking and become footraces or long passes for 30+ yards</p><p></p><p>Now, a real statistician would stop here and update the probabilities based on observation. But what kind of fun is that? Let's prognosticate and tweak!</p><p></p><p>Miami's line looks meh from the few clips I've seen. Midlines, Iso's, and called dives look like they could be purely dominant. FSU hammered them up the middle until they mysteriously decided to stop. Let's say a 35% chance of pure run it up the middle scoring all game long. </p><p></p><p>Other thing to consider is that Miami has historically gone pretty far in on trying to jam the mesh and time the snap. Results have been mixed. I'm actually less worried because successful execution of no-huddle has shown we seem to have the nerves to mess with our timing, if not directly our counts. I really, really hope we see fake a-back motion to catch the the defense looking. We'll tweak generously for Miami, and say 40% good, 35% meh, 17% bad and a range tweak to -5 to 2 yards, 8% to the house.</p><p></p><p>Defense:</p><p></p><p>Here's a nearly totally uninformed prior distribution for Miami:</p><p></p><p>5% 30+ yds</p><p>55% 10-30 yds</p><p>30% -1 to 10 yds</p><p>10% -8 to -1 yds</p><p></p><p>GT has shown two different things on D so far to my inexperienced eyes:</p><p>1. We're getting effective pressure</p><p>2. We're making tackles consistently and well </p><p>and a same thing</p><p>3. We have smart eyes and make well timed picks</p><p></p><p>I think the recaps of FSU I've watched show Miami's QB is really, really good. He's not going to blink in the face of pressure. At least, he sure didn't against a FSU defense that's darn good at bringing it. Where FSU ate it was a mixture of pretty, pretty, highly difficult throws and catches that made the highlight reel, but that were set up by fairly basic execution failures 2-3 plays in a row on denying yards with solid tackling. I think FSU made a conscious decision to go ahead and have a battle of talent on the secondary in order to avoid giving Miami a rhythm to get going in the middle ranges. I think GT takes the opposite strategy, thinking we're more likely to make a good pick over the middle than win consistently one on one on the big throws. I predict swarm fans will be miffed. </p><p></p><p>Turnovers and special teams:</p><p></p><p>I'm not optimistic about our special teams. I'm looking for reasons to update my distribution. I think Miami makes comfortable field goals, but isn't amazing from marginal ranges. </p><p></p><p>I think (or do I hope) we've got a fire in our hearts about turnovers, and hold it to 1. </p><p>I think we get 1 pick against Miami, because that QB has some skills, but we have time and film study. </p><p></p><p>Just for fun, let's say we and Miami both have a 30% chance of scoring or so deep redzoning a turnover that it's effectively more a score than a stop. </p><p></p><p>Ok, great. Now all we need to do is integrate over the various Bayesian outcomes (i.e. P(yards gained on drive > yards to TD on drive given yards to TD on drive is {bins of yards to go}). Luckily, we've set up bins and assigned uniform probabilities over the bins, so we can substitute (Riemann) multiplication for integration. </p><p></p><p>I kid, I kid. Instead I'll just look at what I thought about and use the 'whatever comes to mind' Bayesian replacement heuristic and say:</p><p></p><p>It's almost certain we knock Miami's teeth in on offense, with substantial ability to control the clock.</p><p></p><p>It's almost certain Miami moves the ball consistently on us, but less than 20% probability they can move so successfully that they overcome time of possession and all the efforts of our defense.</p><p></p><p>Mean score before turnovers, GT+8</p><p></p><p>I give a 20% chance no turnovers, 50% chance 1 turnover, 20% chance 2 turnovers, 10% 3+. We'll say each turnover is +9.1 for Miami (7 for stopping a drive, .3*7 =2.1 for pick six)</p><p></p><p>thus:</p><p>.2 gt by 10.1</p><p>.5 gt by 8</p><p>.2 miami by 1.1</p><p>.1 miami by 10.2</p><p></p><p>gives me GT by 4.78.</p><p></p><p>Was that all bs? Of course. But ask any entrepreneur or executive and they'll tell you that the plan is almost always worthless, but the act of planning indispensable.</p></blockquote><p></p>
[QUOTE="IEEEWreck, post: 351666, member: 617"] I sympathize with the sky is falling kind of folks, sort of. I experience plenty of heart-wrenching terror that sticks around until I know the circuits I designed have passed burnout overvoltage tests, but I guess I just use it all up at work. That's human nature, right? You come up with the reasons things might be terrible, and you focus on them to mitigate risk. Sometimes it works well. I'm not sure why I'd examine @miami scores since 2009, but then anxiety is not rational, right? I humbly suggest that football might be more fun if we treat it like a brain teaser, and engage ourselves with a sort of hobbyist Bayesian approach. You might find yourself driven to closer observation that makes the exercise more fun, and you might find the whole thing less anxious in general. Here, I'll use my thoughts on Miami as an example. Using the time honored tradition of pulling things out of my *** (known in academic papers as heuristic methods for generating prior probability distributions) let's lay some ideas down: Offense: We've looked pretty good. Blocking is excellent, and we've gotten consistent 10 yard gains from basic penetration (i.e. the line and linebackers were blocked, but either a farther back defender got in or a jumble at or near the line prevented a footrace. Not, that I've seen, has there been a lot of 'well, the play was executed correctly, but there wasn't a lot of space/a key block on the second level was missing and we consistently have to smash some mouth for 3 yards. If I were serious about this, I'd go work with LongestDay's stuff and try to actually quantify the probability of yardage gains given all blocks made, X blocks made but Y blocks missed, etc, as well as the frequency of those blocking situations. But I'm just some smug EE on the internet, so I'm gonna take wild guesses instead, as so: 45% of plays have good primary and secondary blocking, or are completed midrange passes with +-2 sigma yardage in the 6-14 yard area. 35% of plays have good primary but broken secondary blocks, with +-2s in the 3-5 yard area. 12% of plays are missed reads or other (successfully predicted counter plays like the rocket toss or called dive) that result in +-2s in the -2 to 3 yards area. 8% of plays either have perfect blocking and become footraces or long passes for 30+ yards Now, a real statistician would stop here and update the probabilities based on observation. But what kind of fun is that? Let's prognosticate and tweak! Miami's line looks meh from the few clips I've seen. Midlines, Iso's, and called dives look like they could be purely dominant. FSU hammered them up the middle until they mysteriously decided to stop. Let's say a 35% chance of pure run it up the middle scoring all game long. Other thing to consider is that Miami has historically gone pretty far in on trying to jam the mesh and time the snap. Results have been mixed. I'm actually less worried because successful execution of no-huddle has shown we seem to have the nerves to mess with our timing, if not directly our counts. I really, really hope we see fake a-back motion to catch the the defense looking. We'll tweak generously for Miami, and say 40% good, 35% meh, 17% bad and a range tweak to -5 to 2 yards, 8% to the house. Defense: Here's a nearly totally uninformed prior distribution for Miami: 5% 30+ yds 55% 10-30 yds 30% -1 to 10 yds 10% -8 to -1 yds GT has shown two different things on D so far to my inexperienced eyes: 1. We're getting effective pressure 2. We're making tackles consistently and well and a same thing 3. We have smart eyes and make well timed picks I think the recaps of FSU I've watched show Miami's QB is really, really good. He's not going to blink in the face of pressure. At least, he sure didn't against a FSU defense that's darn good at bringing it. Where FSU ate it was a mixture of pretty, pretty, highly difficult throws and catches that made the highlight reel, but that were set up by fairly basic execution failures 2-3 plays in a row on denying yards with solid tackling. I think FSU made a conscious decision to go ahead and have a battle of talent on the secondary in order to avoid giving Miami a rhythm to get going in the middle ranges. I think GT takes the opposite strategy, thinking we're more likely to make a good pick over the middle than win consistently one on one on the big throws. I predict swarm fans will be miffed. Turnovers and special teams: I'm not optimistic about our special teams. I'm looking for reasons to update my distribution. I think Miami makes comfortable field goals, but isn't amazing from marginal ranges. I think (or do I hope) we've got a fire in our hearts about turnovers, and hold it to 1. I think we get 1 pick against Miami, because that QB has some skills, but we have time and film study. Just for fun, let's say we and Miami both have a 30% chance of scoring or so deep redzoning a turnover that it's effectively more a score than a stop. Ok, great. Now all we need to do is integrate over the various Bayesian outcomes (i.e. P(yards gained on drive > yards to TD on drive given yards to TD on drive is {bins of yards to go}). Luckily, we've set up bins and assigned uniform probabilities over the bins, so we can substitute (Riemann) multiplication for integration. I kid, I kid. Instead I'll just look at what I thought about and use the 'whatever comes to mind' Bayesian replacement heuristic and say: It's almost certain we knock Miami's teeth in on offense, with substantial ability to control the clock. It's almost certain Miami moves the ball consistently on us, but less than 20% probability they can move so successfully that they overcome time of possession and all the efforts of our defense. Mean score before turnovers, GT+8 I give a 20% chance no turnovers, 50% chance 1 turnover, 20% chance 2 turnovers, 10% 3+. We'll say each turnover is +9.1 for Miami (7 for stopping a drive, .3*7 =2.1 for pick six) thus: .2 gt by 10.1 .5 gt by 8 .2 miami by 1.1 .1 miami by 10.2 gives me GT by 4.78. Was that all bs? Of course. But ask any entrepreneur or executive and they'll tell you that the plan is almost always worthless, but the act of planning indispensable. [/QUOTE]
Insert quotes…
Verification
What's the good word?
Post reply
Home
Forums
Georgia Tech Athletics
Georgia Tech Football
Your best guess....What say you about Miami game?
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
Accept
Learn more…
Top