Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Sept 5, 2012 9:26:40 GMT -8
Rocky is trying to out smart himself and so are all of you.
|
|
|
Post by TheSanDiegan on Sept 5, 2012 9:35:10 GMT -8
I've read this thread and compared my comments/info to yours. My conclusion based on your mixups is that you are Googling the concepts. This has reached a level of absurdity that is staggering. Goodbye, scrub. I have a feeling you are accusing me of doing what you have done. Again deflection, because you have lost the issue. You now realize you can not come up with the percentage that you claim can be calculated. That is because there is a lack of relevant data. No, he never "lost" the issue. And while I will happily leave it to SPK to defend his articulation of formal statistical methodologies, I stand behind my own earlier assertion there is more to consider than a fluffy, ill-defined bucket of NCAA-wide macrostats. While there is a limited amount of data pertaining to a specific opponent, there are likely to be enough data points of similar situations to adapt your go/no-go criteria to a given score, game time, opponent, or personnel package... something akin to a more advanced and developed approach to the published article you included a link to yesterday (interesting article, BTW, something I plan to revisit at some point - thanks for the link). I spent nearly a decade mathematically modeling process-driven workflows for IT companies and teaching multivariate analysis to engineers, project, and program managers within said companies. Isolating and quantifying the variability inherent to the situation at hand (i.e., projecting a success rate on a 2-pt. conversion against a given opponent) would appear to be significantly simpler than applying the same methodologies in industry.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Sept 5, 2012 10:11:12 GMT -8
Math nerds, why is it then that both Chip Kelly and Lane Kiffin were going for 2-point conversions against teams that were so INFERIOR that the game was decided the second the ball left the kickers foot at the start of the game? There has to be some 'reason' behind their decisions as it is NOT due to lack of a good kicker on scholarship. So therefore it may have something to do with getting his players prepared for this circumstance 'when' there is a game on the line? Of course that was their thinking. Has nothing to do with Rocky's decisions, of course, since WE were the underdog. However, it does serve as reason to question the article linked by Bill, which I'll now comment about. I see a number of problems with the thing. Here's a couple. 1. He starts out talking about going for to "late in the game." However, the stats he uses are for going for two at any point. So, for example, I have to think there was a little less pressure on Kelly's team to convert than there would be on say a freshman QB trying to bring his team back "late in the game." 2. He says going for a tie is dumb because, among other things, you subject yourself to the vagaries of the OT coin toss. Although that would have considerable validity with regard to how the NFL has done it, it is almost totally or arguably totally irrelevant to college games given the way the NCAA's OT rules work. 3. Particularly consider this, and I quote. "Go for two both times makes no sense logically or mathematically, unless the team's extra point special team is so terrible that its chance for success is less than the chance for making a two-point conversion." Based on the author's data, two-point conversion attempts are successful 43% of the time. If I understand the author, the chances of the average team converting two one-point efforts is the percentage of a one-time success squared. If that's correct, the chances of converting twice is .43 X .43, or about 18.49%. So the chances of converting one of the two two-point tries Rocky had on Saturday was only about 43% and the chance of converting both was about 18.5%. However, even Abel Perez made about 96% of his PATs a year ago. So until convinced to the contrary, this is what I will continue to think. If McMorrow's back prevents him from kicking, redshirt the kid and use Feer to kick PATs and use the soccer player or Joel Alesi to KO. And use Alesi to punt - he was OK on Saturday. With regard to fourth downs - which is a whole separate issue from PATs - keep going for it from midfield to about our opponents' 20 and then make situational decisions thereafter according to how Feer handles PATs. Oh and please don't think I'm a math nerd. I have great respect for those who truly are and I'm truly not.
|
|
|
Post by AztecBill on Sept 5, 2012 10:24:39 GMT -8
Math nerds, why is it then that both Chip Kelly and Lane Kiffin were going for 2-point conversions against teams that were so INFERIOR that the game was decided the second the ball left the kickers foot at the start of the game? There has to be some 'reason' behind their decisions as it is NOT due to lack of a good kicker on scholarship. So therefore it may have something to do with getting his players prepared for this circumstance 'when' there is a game on the line? Of course that was their thinking. Has nothing to do with Rocky's decisions, of course, since WE were the underdog. However, it does serve as reason to question the article linked by Bill, which I'll now comment about. I see a number of problems with the thing. Here's a couple. 1. He starts out talking about going for to "late in the game." However, the stats he uses are for going for two at any point. So, for example, I have to think there was a little less pressure on Kelly's team to convert than there would be on say a freshman QB trying to bring his team back "late in the game." 2. He says going for a tie is dumb because, among other things, you subject yourself to the vagaries of the OT coin toss. Although that would have considerable validity with regard to how the NFL has done it, it is almost totally or arguably totally irrelevant to college games given the way the NCAA's OT rules work. 3. Particularly consider this, and I quote. "Go for two both times makes no sense logically or mathematically, unless the team's extra point special team is so terrible that its chance for success is less than the chance for making a two-point conversion." Based on the author's data, two-point conversion attempts are successful 43% of the time. If I understand the author, the chances of the average team converting two one-point efforts is the percentage of a one-time success squared. If that's correct, the chances of converting twice is .43 X .43, or about 18.49%. So the chances of converting one of the two two-point tries Rocky had on Saturday was only about 43% and the chance of converting both was about 18.5%. However, even Abel Perez made about 96% of his PATs a year ago. So until convinced to the contrary, this is what I will continue to think. If McMorrow's back prevents him from kicking, redshirt the kid and use Feer to kick PATs and use the soccer player or Joel Alesi to KO. And use Alesi to punt - he was OK on Saturday. With regard to fourth downs - which is a whole separate issue from PATs - keep going for it from midfield to about our opponents' 20 and then make situational decisions thereafter according to how Feer handles PATs. Oh and please don't think I'm a math nerd. I have great respect for those who truly are and I'm truly not. Point 3: If you make the first 2 point try, then it would make no sense to try again since a 1 point conversion would give you the lead. So the math works like as follows: 1st try43% make 57% miss 2nd try after make (happens 43% of the time) Make 1 point try: 96% .43 * .96 = 41.28% - 3 pointsMiss 1 point try: .43 * .04 = 1.72% - 2 points2nd try after miss (happens 57% of the time) 43% make 2 point try = .57 * .43 = 24.51% - 2 points57% miss 2 point try = .57 * .57 = 32.49% - 0 pointsThis is after trailing by 14 points. Assuming the score is 14-0 to start, after your 2 TDs of the comeback the scores would be. 12-14 - 32.49% LOSS 14-14 - 26.23% TIE 15-14 - 41.28% WIN Assuming a tie is a 50/50 prop that leaves 54.395% WIN 45.605% LOSS Disclaimer: It is up to the coach to decide how he determines what percentage chance there is to convert after a given TD. That percentage may change based upon all the things that are involved in such an endevor. He may use advanced stats or take an educated guess.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Sept 5, 2012 10:31:43 GMT -8
Bill, that's all fine and dandy as a possible template for being down by two TDs "late in the game." I also think as the other poster pointed out above that in a game where you're a heavy favorite, it's also fine to go for two early on since there is virtually no chance of it impacting the outcome and you are therefore doing so only to gain practice for when making it will matter "late in the game" in subsequent games. However, I see absolutely nothing to suggest that going for two after our first TD on Saturday was smart - UNLESS, as the Montana State professor suggests, our placekicking is so "terrible" that the typical model doesn't apply to us.
|
|
|
Post by AztecBill on Sept 5, 2012 10:34:22 GMT -8
Bill, that's all fine and dandy as a possible template for being down by two TDs "late in the game." I also think as the other poster pointed out above that in a game where you're a heavy favorite, it's also fine to go for two early on since there is virtually no chance of it impacting the outcome and you are therefore doing so only to gain practice for when making it will matter "late in the game" in subsequent games. However, I see absolutely nothing to suggest that going for two after our first TD on Saturday was smart - UNLESS, as the Montana State professor suggests, our placekicking is so "terrible" that the typical model doesn't apply to us. I don't place the "late in the game" qualifier on my contention that going for 2 when down 2 TDs is correct. Since you have to score 2 TDs you "know" you will get 2 chances and can therefore combine chances. Which is what this is about. I played Bridge for years which taught me about the advantages of combining chances. You might want to ask Purdue fans about going for 2 when you are the better team. In a bowl game (Drew Brees was the QB) they were ahead 25-0. How did they get to 25? going for 2, 3 times and making none of them. They lost in OT. In retrospect it is easy to say they weren't the "heavy favorite" that the score convinced them they were. I am not saying they were wrong for going for 2, but I am sure Purdue fans differed on your contention after the loss.
|
|
|
Post by Strawberry Puppy Kisses on Sept 5, 2012 10:43:07 GMT -8
I've read this thread and compared my comments/info to yours. My conclusion based on your mixups is that you are Googling the concepts. This has reached a level of absurdity that is staggering. Goodbye, scrub. I have a feeling you are accusing me of doing what you have done. Again deflection, because you have lost the issue. You now realize you can not come up with the percentage that you claim can be calculated. That is because there is a lack of relevant data. Arrgh, you're aggravating. Okay, let's try to run through this patiently because I still believe you can't be as dumb as you're continuing to be. I've had the pleasure (?) of meeting you before and seemed intelligent... but you just don't know this subject like you think you do! 1. The only thing I've lost is my patience. And congratulations to you for that. 2. You're using population-sized data in your argument. I've repeated to you (as well as another educated (statistics) poster), explained to you, and practically had to insult you, that such data is not a good fit. 3. You've cited unpublished articles, that while may be wholly intelligent, have no relationship to my argument in #2. 4. It is low-mid level complexity data mining & statistical analysis to derive a regression-based model, multivariate, that could identify the critical independent variables that classify "success/fail" - with an associated statistical significance (p), based on a chosen alpha. 5. That model, coupled with probability tables, would be an interesting decision support system. 6. Such a model would improve significantly on your ill-fit probability statements. Further it would be scalable, which your "it's just simple arithmetic" system is not. 7. Lastly, the data to produce such a system is available! It would need to be obtained, consolidated, scrubbed, and otherwise prepared for partitioning into training data. It is no small effort to complete #7! But that's what data mining projects are about. You're mind-bogglingly stubborn in the face of something you know nothing about. Your challenges to "prove it" are ridiculous - do you think I have the spare time to just casually dive into this project? That was my original argument you narrow-minded spazz! I said that if Rocky is working w/ statistical methods & resources, I hope he's doing it properly (see above)... otherwise, if his decision system is what you've been spouting off... IT'S CRAP! Yet you go off on a rail, using google searches, and derail the topic. You don't even realize you're all over the place. Stop being such a stubborn fool.
|
|
|
Post by bitteraztec on Sept 5, 2012 10:58:18 GMT -8
So Bill, where are the correlations between the possible results of the first outcomes and the possible results of the 2nd outcomes?
|
|
|
Post by AztecBill on Sept 5, 2012 10:59:12 GMT -8
I have a feeling you are accusing me of doing what you have done. Again deflection, because you have lost the issue. You now realize you can not come up with the percentage that you claim can be calculated. That is because there is a lack of relevant data. Arrgh, you're aggravating. Okay, let's try to run through this patiently because I still believe you can't be as dumb as you're continuing to be. I've had the pleasure (?) of meeting you before and seemed intelligent... but you just don't know this subject like you think you do! 1. The only thing I've lost is my patience. And congratulations to you for that. 2. You're using population-sized data in your argument. I've repeated to you (as well as another educated (statistics) poster), explained to you, and practically had to insult you, that such data is not a good fit. 3. You've cited unpublished articles, that while may be wholly intelligent, have no relationship to my argument in #2. 4. It is low-mid level complexity data mining & statistical analysis to derive a regression-based model, multivariate, that could identify the critical independent variables that classify "success/fail" - with an associated statistical significance (p), based on a chosen alpha. 5. That model, coupled with probability tables, would be an interesting decision support system. 6. Such a model would improve significantly on your ill-fit probability statements. Further it would be scalable, which your "it's just simple arithmetic" system is not. 7. Lastly, the data to produce such a system is available! It would need to be obtained, consolidated, scrubbed, and otherwise prepared for partitioning into training data. It is no small effort to complete #7! But that's what data mining projects are about. You're mind-bogglingly stubborn in the face of something you know nothing about. Your challenges to "prove it" are ridiculous - do you think I have the spare time to just casually dive into this project? That was my original argument you narrow-minded spazz! I said that if Rocky is working w/ statistical methods & resources, I hope he's doing it properly (see above)... otherwise, if his decision system is what you've been spouting off... IT'S CRAP! Yet you go off on a rail, using google searches, and derail the topic. You don't even realize you're all over the place. Stop being such a stubborn fool. Rocky has no data (that can be used for this) about this year's Aztecs offense. Rocky has no data about this year's Washington defense. Any coorelation between offensive success and style against success in 2 point tries and defensive success and style versus 2 point success may be available for a bowl game (when 12 other games have been played) but are surely unavailable in the first game of the season. I placed a disclaimer on my last post about this. This is in deference to you.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Sept 5, 2012 11:04:53 GMT -8
I don't place the "late in the game" qualifier on my contention that going for 2 when down 2 TDs is correct. Since you have to score 2 TDs you "know" you will get 2 chances and can therefore combine chances. Which is what this is about. I played Bridge for years which taught me about the advantages of combining chances. Throwing out all independent variables, the average team will convert two-point attempts 43% of the time while the average team will convert each of two consecutive one-point attempts about 89% of the time. Statistically, therefore, the average team should be kicking the PAT two consecutive times UNLESS, as the Montana State prof said, they have a "terrible" (his word) or at least markedly inferior PAT unit. So although I haven't said it before now, I think those who claim Rocky's strategy could be harmful to the psyche of his kids could be correct about that.
|
|
|
Post by Strawberry Puppy Kisses on Sept 5, 2012 11:11:52 GMT -8
So, following exactly that logic:
Let's pretend that a sufficient sample was obtained and you derived a statistically significant model. It suggests that, as long as your mean OL weight is greater than X and are playing on the road, you should go for two (I'm making this up, obviously).
You're saying that current season data being unavailable invalidates the sample? That's WHY you (sufficiently) sample!!!
My God man. And you're really arguing with me about this?
Seriously, I know I have said it way too many times in this thread - but I'm done talking to you about this. You don't know this subject as well as you think you do!
If I post in this thread again, I will delete my account. And not in yusef terms.
|
|
|
Post by AztecBill on Sept 5, 2012 11:12:52 GMT -8
So Bill, where are the correlations between the possible results of the first outcomes and the possible results of the 2nd outcomes? That is a good point. The decision is made before the first try and the team is locked into the plan. If you look at all teams that try two 2 point conversions in a game, I suspect there would be less 2 point results and more ZERO point results and 4 point results than expected. That does effect the percentages somewhat but since the results are a 9% spread I doubt it would effect it enough to change the decision.
|
|
|
Post by AztecBill on Sept 5, 2012 11:22:19 GMT -8
So, following exactly that logic: Let's pretend that a sufficient sample was obtained and you derived a statistically significant model. It suggests that, as long as your mean OL weight is greater than X and are playing on the road, you should go for two (I'm making this up, obviously). You're saying that current season data being unavailable invalidates the sample? That's WHY you (sufficiently) sample!!! My God man. And you're really arguing with me about this? Seriously, I know I have said it way too many times in this thread - but I'm done talking to you about this. You don't know this subject as well as you think you do!If I post in this thread again, I will delete my account. And not in yusef terms. If what you are saying is correct, the data we would need would have to come after many games (this season. These are different teams than last year). That is because we need data against a sufficient number of teams to try to gauge how good those teams defenses are. Ditto on our opponents defense.
Below is the disclaimer I will use from now on in deference to your concerns: Disclaimer: It is up to the coach to decide how he determines what percentage chance there is to convert after a given TD. That percentage may change based upon all the things that are involved in such an endevor. He may use advanced stats or take an educated guess.
|
|
|
Post by bitteraztec on Sept 5, 2012 11:36:49 GMT -8
So Bill, where are the correlations between the possible results of the first outcomes and the possible results of the 2nd outcomes? That is a good point. The decision is made before the first try and the team is locked into the plan. If you look at all teams that try two 2 point conversions in a game, I suspect there would be less 2 point results and more ZERO point results and 4 point results than expected. That does effect the percentages somewhat but since the results are a 9% spread I doubt it would effect it enough to change the decision. So would you agree that the expected benefit of the strategy would be greater if you have a highly rated offense or are playing a team that has a poorly rated defense?
|
|
|
Post by AztecBill on Sept 5, 2012 11:45:11 GMT -8
So, following exactly that logic: Let's pretend that a sufficient sample was obtained and you derived a statistically significant model. It suggests that, as long as your mean OL weight is greater than X and are playing on the road, you should go for two (I'm making this up, obviously). You're saying that current season data being unavailable invalidates the sample? That's WHY you (sufficiently) sample!!! My God man. And you're really arguing with me about this? Seriously, I know I have said it way too many times in this thread - but I'm done talking to you about this. You don't know this subject as well as you think you do!If I post in this thread again, I will delete my account. And not in yusef terms. There is nothing to compare your model against !! The type of things that will make a difference is not like the weight of players which can be measured before the season begins. It is things like: median yards per carry on short down short field situations normalized against opponents defenses. Completion percentage on short down short field situations normalized against opponents defenses. That needs a lot of games for the teams in question.
|
|
|
Post by AztecBill on Sept 5, 2012 11:56:23 GMT -8
That is a good point. The decision is made before the first try and the team is locked into the plan. If you look at all teams that try two 2 point conversions in a game, I suspect there would be less 2 point results and more ZERO point results and 4 point results than expected. That does effect the percentages somewhat but since the results are a 9% spread I doubt it would effect it enough to change the decision. So would you agree that the expected benefit of the strategy would be greater if you have a highly rated offense or are playing a team that has a poorly rated defense? Yes. All of that effects the expected conversion rate. That rate is key. If it is low enough than kicking is the right plan. If it is better than about 5% below average, then going for 2 in the situations described, should be the plan.
|
|
|
Post by bitteraztec on Sept 5, 2012 12:06:45 GMT -8
Ok, now let's forget about which teams are playing. What about the effect of which week it is?
Typically, at the beginning of the season the offense is less productive compared to later in the season. Meanwhile there is not as big of a difference on defense performance over the course of the season. Therefore, I would expect that the conversion rate is less in week 1 vs week 12.
Is this a reasonable assumption?
|
|
|
Post by AztecBill on Sept 5, 2012 12:09:36 GMT -8
Ok, now let's forget about which teams are playing. What about the effect of which week it is? Typically, at the beginning of the season the offense is less productive compared to later in the season. Meanwhile there is not as big of a difference on defense performance over the course of the season. Therefore, I would expect that the conversion rate is less in week 1 vs week 12. Is this a reasonable assumption? I have no idea. Are offenses less effective relative to defenses in early season games? I didn't know. I wonder how much the Aztecs practice two point conversations. I would think we should practice short yardage, short field situations more now.
|
|
|
Post by bitteraztec on Sept 5, 2012 12:14:11 GMT -8
What about Home vs. Road games?
Is it fair to assume that the conversion rate is lower for road games?
|
|
|
Post by darksidereturns on Sept 5, 2012 13:17:39 GMT -8
Some interesting quotes this week goaztecs.cstv.com/sports/m-footbl/spec-rel/090412aah.html" Q. What's the mentality this week against Army? COACH LONG: Our game plan has not been decided yet. Every game is going to be dealt with individually and differently. We will put our statistics over the last four years, and we'll put Army's statistics in there for the last three years because this is the fourth year Coach Ellerson is there. The last three years we'll put their stats in there and our chart will be completely different than it was that week. Q. What about if the coach's intuition is different than the chart? COACH LONG: Yeah, we have a chart that says: Green, go for it. It says red, don't go for it. And a gray area that allows for coaching and gut feelings. Good question, though, by the way. The chart's not infallible. "
|
|