Big Turk Posted November 27, 2019 Posted November 27, 2019 (edited) Analytics are all the rage these days and are being used more than ever. In regards to certain things like 4th downs, however, it is not always a smart bet to use them. The reason comes down to sample size. In a game you might face 4th down 6 or 7 times, on average. Now, analytics will tell you that your chance to win will be higher if you go for it because of the expected conversion rate. However, that is over a huge sample size. When you are looking at a small sample size, there will be a LOT of random variation in there. You might not convert any of those 4th downs in a game and then the next game you might convert all of them for instance. Even in a season, when looking at maybe 100 chances, that is STILL a very small sample size in terms of random variability. An easier way to think about it is in terms of flipping a coin. Over a huge sample size, the odds are 50% of getting heads versus getting tails. However, if you took a coin and flipped it 10 times, or even 50 or 100 times, you will find that you might get 60-70% of either heads or tails in there. Why? Because it's not linear. You are not necessarily going to get 50% heads and 50% tails in these smaller sample sizes(but you still might), only when you get to larger sample sizes do these averages come into play. Which brings you to coaches. If they get stuck with bad luck due to random variation, they will probably get fired. This makes it very unattractive for them to risk their jobs and the jobs of many of their coaches to random sample variation. And this is what the analytics people never seem to mention. That yes, while it DOES make sense to go for it over a very large sample size due to the expected conversion rates, it many times does NOT make sense to do it within the sample size they are working with. Edited November 27, 2019 by matter2003
I'm Spartacus Posted November 27, 2019 Posted November 27, 2019 Mike Schopp from WGR550 would love to argue this point with you! Analytics is all he likes to talk about. 1
2012spiller Posted November 27, 2019 Posted November 27, 2019 You’re taking a very simplistic approach to analytics. In your coin flip example, yes it take a lot tosses to get to 50%, but what if you drill down and say coins flipped by a left handed person results in this %, layer on left handed person flipping a quarter vs a dime, and or the surface you’re flipping on, etc. Analytics/modeling is only as good as the variables you factor in. So yes using basic statistics isn’t helpful, but using statistics that are based on scenarios that are close or identical to your situation is very helpful.
Albany,n.y. Posted November 27, 2019 Posted November 27, 2019 Here's an example where not all analytics are created equal. Big difference if it's 4th & inches in enemy territory between having Allen or Barkley in at QB. Allen is almost automatic, not so with a smaller QB or someone like Rob Johnson who may get the 1st down & at the same time go down with a season ending injury.
MJS Posted November 27, 2019 Posted November 27, 2019 Sample size isn't the only issue. You also have to make decisions based on the pulse of your team and the opponent you are facing. Weather can be a factor. Injuries can be a factor. Analytics are a tool in the tool belt. You should absolutely use them to inform your decisions, but they aren't the only piece. 3
Boatdrinks Posted November 27, 2019 Posted November 27, 2019 I’m really not a fan of all this analytics/ numbers geek stuff in sports. I just think NFL coaches are an overly conservative lot, and playing to win ( being smartly aggressive ) vs playing not to lose is a better way to coach a football team. All the rules benefit the offense, yet many coaches are timid when they have possession of the ball.
Just Joshin' Posted November 27, 2019 Posted November 27, 2019 7 minutes ago, MJS said: Sample size isn't the only issue. You also have to make decisions based on the pulse of your team and the opponent you are facing. Weather can be a factor. Injuries can be a factor. Analytics are a tool in the tool belt. You should absolutely use them to inform your decisions, but they aren't the only piece. Thanks - saved me from typing. They are one part of the decision making process, not the end all. I think it is like poker - the analytics may say fold the but situation may lead you to raise. Analytics being situation based is great, pure analytics not so much.
BringBackFlutie Posted November 27, 2019 Posted November 27, 2019 (edited) 16 minutes ago, 2012spiller said: You’re taking a very simplistic approach to analytics. In your coin flip example, yes it take a lot tosses to get to 50%, but what if you drill down and say coins flipped by a left handed person results in this %, layer on left handed person flipping a quarter vs a dime, and or the surface you’re flipping on, etc. Analytics/modeling is only as good as the variables you factor in. So yes using basic statistics isn’t helpful, but using statistics that are based on scenarios that are close or identical to your situation is very helpful. I agree. If you have the numbers that factor in ranks of O and D of both teams, yard line, type of play, and players getting the ball, etc. that will probably refine things to the point that you know whether to go for it or not, regardless of immediately previous results. The odds of winning could STILL be better than kicking, and you'd want to continue to go for it anyway. Because had you not, and you kicked instead, there are chances that that didn't work out on the positive side either. Edited November 27, 2019 by BringBackFlutie
Westside Madness Posted November 27, 2019 Posted November 27, 2019 (edited) 14 minutes ago, Boatdrinks said: I’m really not a fan of all this analytics/ numbers geek stuff in sports. I just think NFL coaches are an overly conservative lot, and playing to win ( being smartly aggressive ) vs playing not to lose is a better way to coach a football team. All the rules benefit the offense, yet many coaches are timid when they have possession of the ball. You just did a mild/loose form of analytics ('geek stuff' <-- this really made me lol) right there. Your term of 'playing to win' vs. 'playing not to lose'. How can you say its a better way to coach a football team without some form of high level analytics? You are making that statement based on your perception that 'playing to win' vs. 'playing not to lose' (by your definitions) results in a greater chance at winning based on the sample size of football you have watched, listened to, etc. #geekstuff Edited November 27, 2019 by Westside Madness
row_33 Posted November 27, 2019 Posted November 27, 2019 victory has a thousand fathers defeat is an orphan 1
LSHMEAB Posted November 27, 2019 Posted November 27, 2019 (edited) 47 minutes ago, matter2003 said: Analytics are all the rage these days and are being used more than ever. In regards to certain things like 4th downs, however, it is not always a smart bet to use them. The reason comes down to sample size. In a game you might face 4th down 6 or 7 times, on average. Now, analytics will tell you that your chance to win will be higher if you go for it because of the expected conversion rate. However, that is over a huge sample size. When you are looking at a small sample size, there will be a LOT of random variation in there. You basically just made the case in favor of deferring to analytics in those situations. If you plan on CONSISTENTLY making the correct decision, you want to look at the large sample size. Edited November 27, 2019 by LSHMEAB
BillsShredder83 Posted November 27, 2019 Posted November 27, 2019 The whole post idea is pretty lazy. Was anybody advocating going for it on 4th down without taking into consideration other variables? I just dont see where this was going
Boatdrinks Posted November 27, 2019 Posted November 27, 2019 1 minute ago, Westside Madness said: You just did a mild/loose form of analytics ('geek stuff' <-- this really made me lol) right there. Your term of 'playing to win' vs. 'playing not to lose'. How can you say its a better way to coach a football team with some form of high level analytics? You are making that statement based on your perception that 'playing to win' vs. 'playing not to lose' (by your definitions) results in a great chance at winning based on the sample size of football you have watched, listened to, etc. #geekstuff If you’re paying attention, the analytics / numbers “ geeks” have infiltrated sports discussion. I don’t find that all that interesting to discuss. Anyone who watches football a lot probably thinks coaches are too conservative. Again, the rules overwhelming favor offense. I’m not saying never punt, or anything like that ( although Schopp on WGR subscribes to that theory) just that coaches get too timid and try to gain a few yards of field position instead of trying to keep the ball and score. That’s my impression. Your mileage may vary.
Westside Madness Posted November 27, 2019 Posted November 27, 2019 2 minutes ago, Boatdrinks said: If you’re paying attention, the analytics / numbers “ geeks” have infiltrated sports discussion. I don’t find that all that interesting to discuss. Anyone who watches football a lot probably thinks coaches are too conservative. Again, the rules overwhelming favor offense. I’m not saying never punt, or anything like that ( although Schopp on WGR subscribes to that theory) just that coaches get too timid and try to gain a few yards of field position instead of trying to keep the ball and score. That’s my impression. Your mileage may vary. Understood. Just like whiskey and cheesecake, all things are better in moderation... 1
Big Turk Posted November 27, 2019 Author Posted November 27, 2019 11 minutes ago, LSHMEAB said: You basically just made the case in favor of deferring to analytics in those situations. If you plan on CONSISTENTLY making the correct decision, you want to look at the large sample size. But you might not be around to see how it plays out over that larger sample size 11 minutes ago, BillsShredder83 said: The whole post idea is pretty lazy. Was anybody advocating going for it on 4th down without taking into consideration other variables? I just dont see where this was going Because many analytics basically say never punt and to always go for it on 4th down no matter what.
row_33 Posted November 27, 2019 Posted November 27, 2019 analytics: memorizing all the odds on your poker hands, and thinking you are a "hawk" after reading a world champion's book reality: saints protect you....
Albany,n.y. Posted November 27, 2019 Posted November 27, 2019 51 minutes ago, I'm Spartacus said: Mike Schopp from WGR550 would love to argue this point with you! Analytics is all he likes to talk about. Another reason Schopp doesn't understand how valuable Josh Allen is to the team. His play on the field doesn't translate well to statistics. The best way to properly evaluate him is to watch him play & never look at his stats. That's one reason why the Bills though so highly of him over the other Josh.
Rock'em Sock'em Posted November 27, 2019 Posted November 27, 2019 7 minutes ago, matter2003 said: But you might not be around to see how it plays out over that larger sample size Generally, not always, you'd have a better chance of staying around by making good decisions over less good ones.
row_33 Posted November 27, 2019 Posted November 27, 2019 5 minutes ago, Albany,n.y. said: Another reason Schopp doesn't understand how valuable Josh Allen is to the team. His play on the field doesn't translate well to statistics. The best way to properly evaluate him is to watch him play & never look at his stats. That's one reason why the Bills though so highly of him over the other Josh. happy medium somewhere not sure Jim Kelly had the best stats, sure played like a tough sonnuvagun out there and was a true leader
Recommended Posts