One of the fascinating things in sports (and in business) has been the growth of the data science area over the past 10 or so years. I can say from my own professional career, I've been involved in a number of big data projects in a couple of industries and it's a massively growing field (fun fact for those going to college or coming out in the work force, go on a focus for data analysis, data science or statistics and you will get a job).
As we know, Major League Baseball has been at the forefront for analytics, mainly because it's a statisticians' dream. 162 games, tons of data to sort through, etc. It used to be that the front offices in baseball was made up of former players or scouts, and maybe had a college kid as the analytics guys. Now the analytics guys run the teams. And it's now also happening on the field as pitching coaches are no longer necessarily a former pitcher, but one of these data nerds (I say that lovingly), who will break down film, check spin rates, arm angles, etc. Look no further than the Yankees passing on David Cone or Andy Pettitte for Matt Blake, who is one of these new wave pitching gurus that use data and video.
Anyway, the NFL has been trying to play catchup. They held a Data Bowl last year to provide essentially a data dump of collected information and asked prospective data scientists to come up with their best analytical models in return for Super Bowl tickets. Honestly, that's a brilliant strategy to get identify some smart minds at a pretty low cost all things considered. And anyone who sticks out would get hired by the NFL data group or individual teams. The NFL has been using data for years (the Eagles among the early adopters), but it still seems very much Wild West in the application.
Anyway, the NFL has been posting information about their data/stats for a wider audience in the link below.
So what does all this mean? Well, it's interesting (at least to me) to see who can use the data the best. I can say from personal experience at work, that the key is sorting through the data to find the proper statistical model that allows you to make a decision. I think the NFL and teams are trying to still figure this out (not nearly as mature as MLB or the NBA), and they aren't there yet but it's coming. And I know that we rolled our eyes at Gettleman's comment about analytics, but I have to believe that the Giants organization is actually utilizing this information in their decision making, even if Gettleman downplays it. But I'm not sure the coaches really know how to parse it. On the surface at least, Shurmur inconsistently uses analytics to inform his decisions (he cites percentages when going for 2, etc). Part of it might be that the analytics are not mature enough to work for coaches just yet. Or that Shurmur just is using it wrong (or could be both).
Anyway, as the Giants look in the mirror at themselves, I really hope that Mara and Tisch are looking hard at this new data group the NFL is building, and as they are in NYC where the NFL offices are located, you would think they would have an advantage to pluck the best and brightest right out of there for a job in the front office to help this team. We know that Ernie Adams has been doing this forever as Belichick's sidekick. We know the Eagles are all over this (is it a coincidence they've been the best team in the NFC East for years and years). For the Giants to turn this mess around, they need to get with the times. There is so much to fix in this organization. George Young helped create an organizational structure which led to the 1980s championships. It's been 40+ years, it's time to adapt a new organizational structure, and it should be one based on competitive data and building a team around that to win.
https://operations.nfl.com/stats-central/stats-articles/ - (
New Window )
That doesn't sound very fun.
75% of that data is probably meant to sell you sh-t you don't need and make money, or steal it.
Stood out to me since many believe (perhaps rightly) that DG is anti-analytics.
That and knowing that when you don't have a very talented team, you need to adopt your game plan on a weekly basis to exploit your opponents weaknesses. If you are going against a stingy D, that means going for it more aggressively on the plus side of the field to maximize points. You don't necessarily have to do this against bad defenses, but I don't think Shurmur's in-game strategy is that nuanced.
All marketing buzzwords dumb people use to try to sound smart.
2 - 8
2 - 8
2 - 8
2 - 8
2 - 8
When Machine Learning eventually starts to help break down data there will be some big movement with the massive amounts of data we have collected.
I can't imagine Shurmur uses data to inform his decisions, because his mistakes are inconsistent. His fourth down decisions and end of half time management would be unacceptable from a Madden player.
But seriously, Matt in SGS for GM.
I can't imagine Shurmur uses data to inform his decisions, because his mistakes are inconsistent. His fourth down decisions and end of half time management would be unacceptable from a Madden player.
But seriously, Matt in SGS for GM.
Ha, God help us. As BigBlueDowntheShore says, Machine Learning is where all this is heading to try to get out all the junk and focus on the real information and start to build out trend analysis.
I hate the piss out of them, but this was a great article about the Houston Astros and how they have built themselves into the best team in baseball over the past few years. It's no accident when you get smart guys in the room who can parse it and build a foundation. The Yankees are doing it now across their entire pitching system from the minors on up to the Bronx.
There is great information there for NFL teams, now go out and find the nerds to help make sense of it and translate to the "football people". It's a hand in hand thing for it to actually work.
Or maybe Coach Red is right, and it's all a bunch of bullshit that makes dummies sound smart.
https://www.theringer.com/mlb/2019/6/3/18644512/mvp-machine-how-houston-astros-became-great-scouting - ( New Window )
So the "Giants Way" nepotism needs to end. Completely. Until then, nothing is going to change.
You could not have data nerds with decision making power and continue the way things are now.
I certainly am in favor of improving upon and facilitating its application and definitely if it can produce or help a Giants lineman blowsomeone off the line of scrimmage.
Maybe I'm being fickle but that doesn't ring true. Data has ALWAYS been produced. Not all data has been captured and not all data has been analyzed. Sorry to be douchy, I work in a 'big data' company and we spout that produced line from time to time and it bugs me.
There is no way yet to read your quarterback's mind-state with 1 minute to go down 7 in the 4th quarter...or your kicker's brainwaves down 3 with 0:05 on the clock. Or your team chemistry. Or a winning locker room mentality.
I certainly am in favor of improving upon and facilitating its application and definitely if it can produce or help a Giants lineman blowsomeone off the line of scrimmage.
It has always been done in some form or another. But the point is that now there is more data available than ever before as technology has improved. You always see some variation of it over time. One example was Coughlin timing lineman after 10 yards in the 40 yard dash. His theory (and it's a correct one), most offensive lineman aren't really going to do much beyond 10 yards to initiate contact for a block. So who give a shit how fast he is running 40 yards, he's not going to go 40 yards unless there is a big play by the offense and he's chasing it down, or there is a big return coming the other way on a fumble or pick and if you are relying on your OL to catch a guy 40 yards downfield, you are pretty much screwed. That's a data point in evaluating offensive lineman. Tom Coughlin, data analyst.
Hell, Bill Parcells was interviewed about the 1986 NFC Championship game in the wind and he quoted some stats that showed the amount of field position saved because McConkey fielded all the punts and didn't let it hit the ground compared to the Redskins returner who let it hit the ground.
So it's not like it was never used, but now with the right data, they can identify some specifics about speed in certain areas (ie- the gunner stat shown in the link) and you can use that to find the right player to fit a scheme. There is information about players going sideline to sideline with real data behind it, so when you want to get a "sideline to sideline" linebacker, you have real information to evaluate.
And of course, there are the charts and models of when to go for 2, when to go for it on 4th down, the right time to take time outs. Clearly, Shurmur either doesn't have any of that information or doesn't understand it if he does. But all that is driven by data analysis. As someone said, I'd love if they had some coaching nerd sitting there with that chart to provide that information to the head coach and crunch the probabilities in real time. You see it in baseball all the time.
Someone famous said that ;)
I remember the time I was at a program review and when presented with a machine solution to a problem by the data scientists, asked why it worked. The answer I got was basically "why does that matter". That was an eye-opener, at least for a dinosaur like me.
And of course, there are the charts and models of when to go for 2, when to go for it on 4th down, the right time to take time outs. Clearly, Shurmur either doesn't have any of that information or doesn't understand it if he does.
For the life of me, I don't understand why someone hasn't come up with a "chart" for calling TOs at the end of the half. And make it situational - based on what you have left, scoreboard, and time.
There is no way yet to read your quarterback's mind-state with 1 minute to go down 7 in the 4th quarter...or your kicker's brainwaves down 3 with 0:05 on the clock. Or your team chemistry. Or a winning locker room mentality.
But that's the point of this. You don't remove the human element or even the "football people", but you augment it with the data group. It's a hand in hand thing to work. It will never be (or doubtful) just a stat guy as the head coach. Baseball is much more conducive to utilizing stats because there is a much larger data set to work from. And in football, so much can change that impact the data (bad weather, injuries across your OL so you don't block anyone, etc). However, we have seen so many bad decisions made by coaches, and Shurmur has not shown any feel for any in game decision. But he had the right information on the right type of play call to run against a certain defense based on the way they previously played a situation (in baseball it's telling pitchers what type of pitch to throw and what not to throw in a situation...basically how a catcher used to call a game but now driven by data). In addition to the obvious (when to go for 2, when to go for it, when to punt, when to call time outs). Essentially if done right, it's a cyborg setup. Computer data providing the recommendation and the humans to decide how to best implement it or ignore it (because they know their kicker has the shits or the QB is playing like shit, etc).
For the life of me, I don't understand why someone hasn't come up with a "chart" for calling TOs at the end of the half. And make it situational - based on what you have left, scoreboard, and time.
BillKo - I am not much into making coaching evaluations based on game management decisions as it's a minor part of an HC's job (Reid is terrible at it yet is a very successful coach). BUT I will make an exception of it for Shurmur. The fact that he doesn't simply have a planned, programmed and intelligent approach to various end of half situations developed over his many years of HC and OC is experience is a damning indictment of his level of preparedness.
I remember the time I was at a program review and when presented with a machine solution to a problem by the data scientists, asked why it worked. The answer I got was basically "why does that matter". That was an eye-opener, at least for a dinosaur like me.
There is no way yet to read your quarterback's mind-state with 1 minute to go down 7 in the 4th quarter...or your kicker's brainwaves down 3 with 0:05 on the clock. Or your team chemistry. Or a winning locker room mentality.
These types of things are incorporated into computer models to some extent. We can assess different groups to determine their level of chemistry and add that to the computer model. (For example, getting data about group of people who have been together for years vs. a group that is just thrown together). We can also add in-situ observations or our confidence level in certain parts of the problem. I've seen several hybrid models with that combine analytic background information with "in the field" observation. For example, if I'm building a computer model to use for football decisions, I can have slide a bar left or right to indicate my OL performance is crappy or great.
I'm not sure it would take another 15 years for a computer/human hybrid model to outperform the average NFL coach acting alone.
Quote:
I remember the time I was at a program review and when presented with a machine solution to a problem by the data scientists, asked why it worked. The answer I got was basically "why does that matter". That was an eye-opener, at least for a dinosaur like me.
. Off-topic, Ray, but I think your question was legitimate and the data scientists need to be able to answer it, not just wave it off.
In this case, we would have to have given ~$1.5M over 3 years to a university professor to tell us what the physics were behind the machine solution. It seemed better to just accept the better performance by the machine method then spend that money. (It wasn't actually my decision).
There is no way yet to read your quarterback's mind-state with 1 minute to go down 7 in the 4th quarter...or your kicker's brainwaves down 3 with 0:05 on the clock. Or your team chemistry. Or a winning locker room mentality.
True. But there is plenty of opportunity for game situation - time management, tendencies, 4th down, goal line, etc.
I'm very interested in trying to convert college profile data to pro ability. From college stats to personality to family life to conference, etc, I think there is something there to create advantages to having better hit rates in the draft...
It's funny to read this now and see that it's more or less the same discussion. In the 1960s, computers were fairly new to use in this way. Now we are in a world were you have tons of data, but how to use it right.
And I found this amusing in the article that answers why you always need a person to evaluate what's going on
As Qureishi pointed out, computers are not creative thinkers and cannot react to situations like this. No matter how sophisticated the machines get, the game will always be dependent upon human inspiration and human error.
https://www.si.com/vault/1968/01/29/670175/make-no-mistakes-about-it - ( New Window )
Quote:
For the life of me, I don't understand why someone hasn't come up with a "chart" for calling TOs at the end of the half. And make it situational - based on what you have left, scoreboard, and time.
BillKo - I am not much into making coaching evaluations based on game management decisions as it's a minor part of an HC's job (Reid is terrible at it yet is a very successful coach). BUT I will make an exception of it for Shurmur. The fact that he doesn't simply have a planned, programmed and intelligent approach to various end of half situations developed over his many years of HC and OC is experience is a damning indictment of his level of preparedness.
He's all over the map, I hear ya. He's going on feel, and while "feel" can be applied to other areas, this is not one IMO.
To me, in-game management is huge. Games are won and lost based on that.
Time out management probably doesn't cost games per say, but it can easily put you in better positions.
That's why I am amazed there isn't a chart - similar to the 2 point conversion chart - that can be referenced by all NFL coaches.
Quote:
thanks for the post. I think Data analysis has been around in football for a very long time as well as other professions. The good organizations have always done it. I understood it to be preparation.
I certainly am in favor of improving upon and facilitating its application and definitely if it can produce or help a Giants lineman blowsomeone off the line of scrimmage.
It has always been done in some form or another. But the point is that now there is more data available than ever before as technology has improved. You always see some variation of it over time. One example was Coughlin timing lineman after 10 yards in the 40 yard dash. His theory (and it's a correct one), most offensive lineman aren't really going to do much beyond 10 yards to initiate contact for a block. So who give a shit how fast he is running 40 yards, he's not going to go 40 yards unless there is a big play by the offense and he's chasing it down, or there is a big return coming the other way on a fumble or pick and if you are relying on your OL to catch a guy 40 yards downfield, you are pretty much screwed. That's a data point in evaluating offensive lineman. Tom Coughlin, data analyst.
Hell, Bill Parcells was interviewed about the 1986 NFC Championship game in the wind and he quoted some stats that showed the amount of field position saved because McConkey fielded all the punts and didn't let it hit the ground compared to the Redskins returner who let it hit the ground.
So it's not like it was never used, but now with the right data, they can identify some specifics about speed in certain areas (ie- the gunner stat shown in the link) and you can use that to find the right player to fit a scheme. There is information about players going sideline to sideline with real data behind it, so when you want to get a "sideline to sideline" linebacker, you have real information to evaluate.
And of course, there are the charts and models of when to go for 2, when to go for it on 4th down, the right time to take time outs. Clearly, Shurmur either doesn't have any of that information or doesn't understand it if he does. But all that is driven by data analysis. As someone said, I'd love if they had some coaching nerd sitting there with that chart to provide that information to the head coach and crunch the probabilities in real time. You see it in baseball all the time.
Matt, I am agreement with applying and improving the process. I think a big benefit is it can better utilize time so coaches can focus on other things.
Baseball makes sense but I never played the game. They seem to going to playing percentages and responding accordingly.
Didn't they bring a analytics gut to Cleveland? I believe he utilized trading down and having more picks. Even with that system you have flaws. Say you have 20-24 picks over two years and half turn out very good. What do you do when you have to pay all 12 with a salary cap? Is there enough experience on the roster? Sure it could work over time but I am not sure it will be a more viable strategy. We need more Data Analysis on this!
Mmmm. Not a big baseball guy but what then is causing the Yankees to get big hits in the playoffs against good pitching? Has he not gotten to solving that part?
Quote:
Forget the difference in sports, I'm pretty sure you could name him the Giants GM and he'd have this thing turned around in less than 3 years. Not because he relies on analytics, but because he knows how to use them to make data driven decisions, and he knows how to find people to help him accomplish that goal.
Mmmm. Not a big baseball guy but what then is causing the Yankees to get big hits in the playoffs against good pitching? Has he not gotten to solving that part?
You mean the fact that they *only* made it to the ALCS?
Quote:
In comment 14677640 jcn56 said:
Quote:
Forget the difference in sports, I'm pretty sure you could name him the Giants GM and he'd have this thing turned around in less than 3 years. Not because he relies on analytics, but because he knows how to use them to make data driven decisions, and he knows how to find people to help him accomplish that goal.
Mmmm. Not a big baseball guy but what then is causing the Yankees to get big hits in the playoffs against good pitching? Has he not gotten to solving that part?
You mean the fact that they *only* made it to the ALCS?
And lost to a team that has since been proven to cheat ;)
It isn't.
Quote:
In comment 14677640 jcn56 said:
Quote:
Forget the difference in sports, I'm pretty sure you could name him the Giants GM and he'd have this thing turned around in less than 3 years. Not because he relies on analytics, but because he knows how to use them to make data driven decisions, and he knows how to find people to help him accomplish that goal.
Mmmm. Not a big baseball guy but what then is causing the Yankees to get big hits in the playoffs against good pitching? Has he not gotten to solving that part?
You mean the fact that they *only* made it to the ALCS?
Baseball is not like the NFL. All teams spend close to the salary cap in the NFL and free agency is nothing like the NFL nor is the injury factor. The Yankees should be in the playoffs most years imo.
From a data science perspective baseball and football are definitely very different - on the football side, the analytics are a lot more complex because there are a lot more moving parts.
The focus though would be on efficiency and maximization - something that the Giants are pathetic at. Wasted resources left and right, from signing players and trading them right after to retaining Eli while drafting a QB with the plan of sitting him a year then starting him 2 games later. A halfway decent data science department could have prevented the Giants from making some of these blunders, though then again. common sense could have taken them at least halfway there.
From a data science perspective baseball and football are definitely very different - on the football side, the analytics are a lot more complex because there are a lot more moving parts.
The focus though would be on efficiency and maximization - something that the Giants are pathetic at. Wasted resources left and right, from signing players and trading them right after to retaining Eli while drafting a QB with the plan of sitting him a year then starting him 2 games later. A halfway decent data science department could have prevented the Giants from making some of these blunders, though then again. common sense could have taken them at least halfway there.
I am not a big baseball guy and never played the game. I was unaware about the cap as I thought teams could spend as much as they want but taxed after the max. I thought teams could also still have like 50 million dollar payrolls and did not have a minimum they had to spend that was close to the max.
I will say one area I would love analytics to help with is predicting a player's longevity. Obviously some injuries can't be predicted but maybe there is something they would provide meaningful outcomes.
Too many Giants over the years have faded at the 30 year old mark. Osi, Tuck, Snee, McKenzie, etc. They you have Suggs still playing who came in the same year as Osi.
For example:
Eyes say that if a DB is in position for a PD, and yet tries for the full INT instead, the WR often ends up with the ball. Due to its an adjustment in body position at high speed for the DB.
BUT, is this a point of emphasis right now on this team? That statistically you go for the much higher probability of a PD you will win more often as opposed to thinking "INT all the way" and the WR gets the ball ? Probably it's not discussed.
Take that a step farther.
If. IF ..if a DB is in position for the INT and yet is also measuring for the run and 6pts.. often he fails to gather in the INT! Hello. Highlights don't = wins .
We have seen it a million times.
BUT.. are these types of points of emphasis covered on this team at this time? It doesn't seem so.