| “It’s about winning and I’ve seen someone told one of the reporters I’m in the teardown,’’ Gettleman said recently. “We’re not spending $62.5 million on Nate Solder, spending the money on [28-year-old guard] Patrick Omameh, we’re not trading for Alec Ogletree. If it’s a teardown, we’re not doing that. We evaluated the roster, we’ve developed a plan moving forward. It’s about winning now. Who wants to lose? I don’t.’’ |
I saw this on twitter today - link to article below. A lot of people say this wasn’t a “win now” year or philosophy. I keep seeing the point of “60 percent roster turnover”. This quote says a lot & they are staring down 5-11.
Link - (
New Window )
It's one thing to not like their plan - but to wonder if they have one at all seems a bit silly. I'm fairly certain they're not just proceeding at random here without actually having discussed any course of action.
It's one thing to not like their plan - but to wonder if they have one at all seems a bit silly. I'm fairly certain they're not just proceeding at random here without actually having discussed any course of action.
OK, so let's rephrase - a coherent plan.
People point out a number of possibilities - like the fact that the team had a lot of holes, that Eli was nearing the end but they might have been forced to keep him because of the cap hit. Fine - at that point, trading guys like JPP and Snacks make perfect sense. Overpaying for Solder and a journeyman like Omameh, not to mention Stewart because BBI has resolved him to being a rounding error - don't exactly scream coherent plan.
but it is not incompetent to move fast and avoid uncertainty with little time on the clock ...and a very uncertain scouting effort still in gear so a sure hand fast mattered.
The point is that moving fast was a defensible choice at the time. other teams were looking for GMs which meant a 6 to 8 week process would compromise both Fa and the draft.
The next GM choice matters a ton for it's a person we might live with awhile
To me the Maras have earned current doubt. But not a kangaroo court before this plays out 2 to 3 years
It's one thing to not like their plan - but to wonder if they have one at all seems a bit silly. I'm fairly certain they're not just proceeding at random here without actually having discussed any course of action.
In general, they certainly have a global plan. Make the roster better, manage the cap to allow for that, draft needs, etc.
But I am leaning more towards the point of view that they are improvising a plan for QB - short and long term. And that is unsettling.
Gettleman came out with a lot of bluster, spent a lot of cash and made some clear mistakes. I think his plan was based on the Giants being closer than he realized.
There were a couple of clear points where his actions showed he got how far away the team was. 1) Week before game 1 when we was still mix-matching, knowing guys would have ZERO reps going into a real game 2) moving Snacks and Apple for arguably low return.
I don't like Gettleman, but I respect his actions in-season.
Again, this is all about the QB. It's a third-rail, daring to sniff post-Manning got his predecessor fired.
It's no excuse, and the next 2 years will judge his actions. But the fan base, ownership, Manning Inc. all have to be considered, not just what's right on the field.
Quote:
Do people really believe the Giants have no plan at all? Like they're just flailing wildly with no calculated rhyme or reason behind it?
It's one thing to not like their plan - but to wonder if they have one at all seems a bit silly. I'm fairly certain they're not just proceeding at random here without actually having discussed any course of action.
In general, they certainly have a global plan. Make the roster better, manage the cap to allow for that, draft needs, etc.
But I am leaning more towards the point of view that they are improvising a plan for QB - short and long term. And that is unsettling.
I think it's just a unique and difficult situation with the QB. This isn't "okay, we need to move on from Blake Bortles - he's not getting it done.."
This is us trying to move on from a 2x Super Bowl MVP and Giants legend who keeps insisting he wants to play and wants to play here. It's hard to shove him out the door. After the uproar last year during the entire Geno Smith fiasco and remembering how it ended with Simms, I think Mara is petrified of creating a rift between the org and Eli or this becoming a major ordeal that makes the Giants look bad.
I think the procession plan and everything else will fall into place easier when we have closure regarding #10. That entire situation is kind of clouding everything else right now and making it hard to move on.
In other words, it takes time to tear down and rebuild a roster. Especially one that was top heavy in terms of the salary cap, and not well cobbled together on the whole to begin with by the prior regime. It takes time to get rid of bad contracts, and sometimes negative personalities, and time to restock with quality that fits your master plan.
I hate watching my team lose. Being a team that is struggling and at times feeling like there is no plan or evolution in progress is bothersome as hell. How could all of these seasoned professionals seem so clueless, etc. Ultimately, teams take 2-3 years to build properly. We seeing these teams turn it around in one year, but most of them hard their key pieces in place, and saw some benefit from a weaker schedule, having no film on tape to game plan against, no $20M QB who won titles in the past to say goodbye to.
This process will require patience. BBI is the place to vent, I get it, but this will take some time. Can't tell others how to act and comport themselves, but fuming over 2018 is an exercise in futility. Even with a precise blueprint in hand, there's 31 other teams to compete with for talent, on and off the field.
Hope to see a bump up in 2019 and keep building up the talent base, get those systems in place, lose the players that won't be part of the next run. Get lean under the cap, be wise with pricey acquisitions. I'm not sold on DG or PS, and perhaps they aren't here two years from now if this thing burns to the ground again.
It will take at least through next year to clear the deadwood completely and even with the right pieces I would be surprised with 8-8 in 2019.
Let's see what the plan is before condemning DG. I'm sure pieces will be let go come late January and early February so you will get a sense of their direction or plan.
I'm not saying any are inevitable -- but moving on from the face of the franchise is scary. It would be easier if he was an asshole like Favre.
I think it's just a unique and difficult situation with the QB. This isn't "okay, we need to move on from Blake Bortles - he's not getting it done.."
This is us trying to move on from a 2x Super Bowl MVP and Giants legend who keeps insisting he wants to play and wants to play here. It's hard to shove him out the door. After the uproar last year during the entire Geno Smith fiasco and remembering how it ended with Simms, I think Mara is petrified of creating a rift between the org and Eli or this becoming a major ordeal that makes the Giants look bad.
I think the procession plan and everything else will fall into place easier when we have closure regarding #10. That entire situation is kind of clouding everything else right now and making it hard to move on.
I get it.
But we're heading down a path, and I know you get it too, where there is going to be a lot more barking about what we should have done instead; and how many years may have been wasted playing the situation too delicately.
Interesting how HC/GM/Pied Piper Bill Belichick was reportedly ready to execute his transition plan by keeping JimG and sending Brady to the exits. By ole Bob Kraft stepped in, drew a line in the sand, and said that ain't happening...
Different circumstances because Brady continues to be productive and win, but we see there was a plan in place, cold blooded and all...
This is us trying to move on from a 2x Super Bowl MVP and Giants legend who keeps insisting he wants to play and wants to play here. It's hard to shove him out the door. After the uproar last year during the entire Geno Smith fiasco and remembering how it ended with Simms, I think Mara is petrified of creating a rift between the org and Eli or this becoming a major ordeal that makes the Giants look bad.
I think the procession plan and everything else will fall into place easier when we have closure regarding #10. That entire situation is kind of clouding everything else right now and making it hard to move on.
[quote] You try to win every season. In the age of parity, with a couple of breaks, things can happen. It didn't go our way this year, but a break here or there and we could be fighting for the playoffs, too. [/quote Yup If the refs do not screw us in the Carolina and second Philly game we are still alive and have a great shot to make the playoffs .So frustrating
did you really see us as a playoff team playing the Titans last week, I sure didn't, it was humiliating as hell..
Or the fact that the Giants themselves felt it necessary to add two other candidates to the list, which is what really made it look ridiculous. One was Marc Ross - who stood exactly 0.00% chance of being hired (we'll call this the courtesy interview for a long standing employee).agree on this. Then there was Louis Riddick - someone who wasn't working in a FO capacity, and wasn't very senior before he left, either.I think everyone was a little surprised by this interview. I wouldn't be surprised if this was a courtesy done for his agent/at his request to get some experience, which is the kind of thing neither party would publicize. Most likely (and we're all guessing here) the Giants had a good idea that they wanted DG all along and the interview was little more than a formality. That obviously bothers those who feel that a massive change was needed more than those who liked DG and felt he could make necessary changes while not throwing out the baby with the bathwater. My take anyway.
Dan - we'll agree to disagree on some of the differences you've mentioned. If you're hiring rank and file, sure, there's no noncompete clause to deal with, and a larger pool of talent. Finance has plenty of niche roles that are without a huge pool of talent, where banks fall over each other for the right to overpay for one of the few available candidates. Agreed, and in those cases might they not feel a need to act quickly? Given that the talent is specialized enough? And plenty of senior management changes firms carrying over a NC that prevents them from poaching people they're comfortable with from their prior role. Again, agreed, and in the cases where they require a specific talent who is not under contract (poachable) where the pickings are very slim, is it not likely we see a similar approach? Is it wrong if they make an offer to a senior leader with unique skills after a very brief interview process? In my opinion, I think it would be normal/acceptable.
Responses inline in bold
Quote:
Or, it's a comment that cant be substantiated but you'll pretend you're smarter than the room instead of poorly supporting it with a bunch of subjective takes.
One of the first things that happened when DG came in was to start cleaning house. Getting rid of guys or not resigning guys who didn't seemingly care. Hart was first. Eventually DRC, Jerry, Flowers and Apple were gone. Pugh, Richburg and Fluker weren't resigned.
You labeling players as 'guys who seemingly didnt care' is more of that stuff that sounds great but ultimately isnt supported by anything. Your best argument here is Hart and Pugh, guys who actually had stories about them follow. Everything else comes off as 'dissatisfied fan wants to see someone come down hard on players because of a bad season. Is there something that would support that Jerry, Fluker, Richburg, and DRC just didnt care about their jobs?
There's something about management laying down the law on the labor that somehow thrills fans. 'Roster churn'? so 17 players are left from last year, and the year was a bomb anyway. But bonus points and a clear win for the grumpy tough guy because they didnt quit on the field, it must be because last year they just werent scared enough of losing their jobs. Maybe Gettleman should be more afraid of losing his job so he can cut down on the dead money and mistake contracts.
JonC will support that these whispers existed.
Outside of Fluker, the rest of those players are gone primarily because of attitude or a combination of attitude and performance.
Are you really going to act like we didn't hear anything about those players? And I'm just going on what's at the surface, and not the underlying information I know.
Jesus.
You labeling players as 'guys who seemingly didnt care' is more of that stuff that sounds great but ultimately isnt supported by anything. Your best argument here is Hart and Pugh, guys who actually had stories about them follow. Everything else comes off as 'dissatisfied fan wants to see someone come down hard on players because of a bad season. Is there something that would support that Jerry, Fluker, Richburg, and DRC just didnt care about their jobs?
There's something about management laying down the law on the labor that somehow thrills fans. 'Roster churn'? so 17 players are left from last year, and the year was a bomb anyway. But bonus points and a clear win for the grumpy tough guy because they didnt quit on the field, it must be because last year they just werent scared enough of losing their jobs. Maybe Gettleman should be more afraid of losing his job so he can cut down on the dead money and mistake contracts.
I agree with you - there are whispers and anonymous reports about attitudes but I haven't heard anything official. It's a little too easy to simply label all those guys as not caring.
I liked Bill2's contribution yesterday on this thread where discussing analytics he described its use in evaluating player value. I think it much more likely that if the Giants are using some kind of present/future value model and assigning a cap value to players which leads to some otherwise head-scratching moves.
Fans and media using only their own evaluation techniques may not understand and the result is what we're hearing now about players no longer here.
Bottom line - I don't think anyone outside of the org knows for sure why specific players were let go.
Responses inline in bold
Dan - I think we agree to disagree here. IMO, the Giants had absolutely nothing to lose by interviewing another viable candidate, even if they had pretty much decided that Gettleman was the guy (for the reasons outlined above). They stood no chance of losing him, so the opportunity cost of having additional discussions was only time - and that could have been measured in days, not weeks.
The worst part is that Gettleman is older and after the hiring was revealed to have a serious health issue, which means the Giants could be in the GM boat the next time around. This hiring process could have identified some up and comers who weren't ready to be hired yet, but warranted monitoring, no different from trying out players who you don't sign at first but keep an eye on. Instead, the search ended early, with the only real benefit being time. I don't see how a few more days of GM search would have cost the Giants anything.
Dan - I think we agree to disagree here. IMO, the Giants had absolutely nothing to lose by interviewing another viable candidate, even if they had pretty much decided that Gettleman was the guy (for the reasons outlined above). They stood no chance of losing him, so the opportunity cost of having additional discussions was only time - and that could have been measured in days, not weeks.
The worst part is that Gettleman is older and after the hiring was revealed to have a serious health issue, which means the Giants could be in the GM boat the next time around. This hiring process could have identified some up and comers who weren't ready to be hired yet, but warranted monitoring, no different from trying out players who you don't sign at first but keep an eye on. Instead, the search ended early, with the only real benefit being time. I don't see how a few more days of GM search would have cost the Giants anything.
Fair enough.
I think you should know that I hold you and your opinions in high regard. IMO you are truly a valued BBI contributor.
Cheers and hope your holiday season is/has been a great one.
Dan - I think we agree to disagree here. IMO, the Giants had absolutely nothing to lose by interviewing another viable candidate, even if they had pretty much decided that Gettleman was the guy (for the reasons outlined above). They stood no chance of losing him, so the opportunity cost of having additional discussions was only time - and that could have been measured in days, not weeks.
The worst part is that Gettleman is older and after the hiring was revealed to have a serious health issue, which means the Giants could be in the GM boat the next time around. This hiring process could have identified some up and comers who weren't ready to be hired yet, but warranted monitoring, no different from trying out players who you don't sign at first but keep an eye on. Instead, the search ended early, with the only real benefit being time. I don't see how a few more days of GM search would have cost the Giants anything.
This is very well said. I’ve said similar remarks just not as well...
it may reveal who might have been spoken to and how wide the hidden net went. Not for the GM job (the one where someone in charge without a time of drift during the end of year and pre-fa and draft evaluation process) but for the coaching job.
Take a definitive statement like “we had a good draft” people want to say this as if it’s a fact, some of our favorite luddites on this board are brandishing this with their I <3 Gettleman T Shirts. Yet I would say the jury is very much still out on this. And I know this may be difficult for some to understand but this statement has nothing to do with who Barkley is as a player who I love more than any Giants in a long time. I think from a game theory analytical perspective it is healthy to look at Barkley’s closest comps in lieu of complex model results that I don’t have. Players like Peterson and Sanders two of the very best RBs in history. The Vikings were 80-79 with Peterson on the team and the Lions were 78-82 with Sanders. If you look at the records after their rookie deals 44-51 and 48-48 it brings up some important questions. Questions that clearly concerned the Steelers with Le’Veon Bell. Now they hired this Carnegie Mellon professor in 2015 and their team seemed to think that it could win with Bell at $15 Million but not at $18 Million this is the kind of thing that absolutely points to analytics. Here is where Gettleman comes in and fateful clip. My issue is not with the pick of Barkley. My issue his response to this question. He had the wrong answer. It did not reflect an understanding behind probability theory necessary to minimize the mistakes in his data driven job. The bottom line is we don’t know if Barkley is not so good at a time that many of us all agree that the team isn’t ready to compete that he requires resources that from a team construction standpoint will not allow us to compete when we are ready. Say Barkley like Bell and because of the wear and tear on his body in 3 years says I want $25M a year or I sit? Don’t you dare franchise me. Do we know that he won’t do that? Do we know that he won’t break down? Can a team win with an RB making $25M a year? $20M? What if he does get injured? The point is, there are still worlds out there where this is a good or bad move and “getting into the analytics of it all” is an important part of a process in ensuring you are making these choices for all the right reasons. The right answer is we weighed a lot of factors and believe we made the right choice or something in that vein. What makes this clip even more troubling is that he uses Jonathan Stewart as an example to defend his point. There is no other way to slice this, but he looks very stupid for saying that. Not because he fell off the cliff, there was only some probability that that would happen this year. But the bravado which he states he hasn’t fallen off yet. He is framing the problem in the wrong way mathematically. The fact that Stewart hasn’t dropped off and his age in fact raises the probability that he will imminently, it is not a basis to think a player is immune to wear and tear. Now it would be nice if it looked as though when they were evaluating signing Stewart that someone was in his ear reminding him of this. Perhaps our GM wouldn’t say things that he honestly should regret saying, he didn’t need to go out on a limb like this, but it shows the kind of overconfidence in an outdated process that should send up red flags to any technologist. (I’d be interested in your take on this as a technologist Bill2, I share this clip so often because all my friends that are football fans and data people have not stopped mocking me about it)
This leads me to my next point. You could say back that the contracts Gettleman handed out to these people were mistakes anyone would make, I say back, well, are all teams using the same search parameters now? Could we be more backed into a corner in ways of fixing our team because of our process? Could things that worked for Gettleman 5 years ago not have the same efficacy now given the advancements of other teams? This is where the article shared today on another thread comes in on the current state of analytics. First, they bring up the Giants and the Gettleman video as an example of a team that is behind. Do any of us really want the Giants to be a team that stands out for this? Shouldn’t our ownership be ashamed of itself given its access to Wall St. and being a hub of business that we are a team on the wrong side of this article? Is it not embarrassing that given our record against the Eagles that they have had an advanced analytics program for 20 years and a guy with Ty Siam’s background is our rising star? This broad analytics article shared in another thread very much supports a lot of the points I have been making all season.
“It’s about translating that data ASAP and being very, very in tune with the numbers. You can’t be a year behind, you can’t be a month behind.” —Thomas Dimitroff, Atlanta Falcons general manager
“It’s about translating that data ASAP and being very, very in tune with the numbers. You can’t be a year behind, you can’t be a month behind,” said Falcons general manager Thomas Dimitroff.
Teams are examining details they’ve never studied before to get an edge. One scouting department graded a defensive back prospect as an undrafted free agent due in part to his slow 40-yard dash. When that department was able to measure his game speed using tracking data, it determined that it should have listed him as a midround pick. Other teams in the market for linebackers have homed in on what kind of closing speed elite tacklers need. For instance, Zebra Technologies, a company whose MotionWorks service collects game-day data, found that the Cowboys’ Leighton Vander Esch reaches 16 to 17 miles per hour on his best plays.
“Teams can go deep on rosters to leverage the tracking data to scout players for the future, maybe in free agency,” said John Pollard, vice president of Zebra Technologies.
I’ve been saying for a bit now that this is the reality. And I’ve been criticized for bringing up age in relation to this, but this article does as well, and it is generally true that younger people are more interested in incorporating new ideas. I’ve brought up Zebra several times because I do think that is one of the keys, but it is honestly just a small part of what it is I think we need to do. This arms race and articles like this are also important to note. If you have an original football model or ideas what are the chances you shoot an email to a guy like Gettleman vs. Ernie Adams or Howie Roseman? This is a problem that if we don’t start addressing today, we ARE doomed to fall in a deeper and deeper competition chasm. We need to be known widely as a team that embraces new ideas and technologies not scoffing at it. And the fact remains, and I don’t think this can be emphasized enough, even if we compete the next few years with Gettleman he could be causing damage in competitive advantage that lasts multiples of years longer. This is why I really am trying to emphasize that while the jury is out on Gettleman’s ability to compete today, there is enough evidence that we need to bring a strong analytics mind near the top of our organization ASAP if we want to prevent a serious long-term competitive disadvantage.
Reading between the lines of these threads I’d say that it might be the case that Bill2 either knows the Tisch family or is one of the people the Giants talk to about these issues. So despite the fact that I’ve resisted doing this given the fact that I perceive this to be valuable information that I could be paid for I’m going to go ahead and share it because if they aren’t already implementing a system like this, they should be. And I’ll be the first to say, great if they are but their internal staff does not suggest they have the knowhow to do this and I will explain that even if they are working with an outside firm this is not optimal to gain or maintain any kind of performance advantage.
It’s best to start with how to frame these problems from a high level. People that know me as a big optimist, I think you need to be one in order to make good forecasting systems. Every time I begin to try to predict something new the first take is going to be nothing compared to what I end up with after a few months of work. I’ve predicted NPV of cash flows for every securitized mortgage in the US, huge college and corporate loan packages, stock outcome probabilities and index addition probabilities, trade predictions for a global banks entire fixed income portfolio (22K ISINs paired with 13K clients), healthcare classifications and at one point I even dabbled in Football. I’ve done this for many years, each time I improve my process and the technology platform I use gets better and smarter. This is all to say even if we hired the best person in the world at making football predictions or our staff learned really well on the job there are things you pick up in academia trying to solve many different signal processing issues or in the working world dealing with many different problems that we are still most likely falling short in. This is why I brush back at the idea that Ty Siam or Buffalo IT consultant guy can accomplish something like what is needed to provide value in football. There is nothing to suggest that they can even leverage open source tree-based algorithm code, let alone boosting, bagging or neural net code. They would have to have a deeper understanding of this before they could build bespoke systems that can try to optimize around different hyperparameters and time series modifications within the training of algorithms which requires writing or changing code within these algorithms. Situational ensembling is also quite necessary for these complex non-linear equations IE if I want to predict the outcome of the next play is a single neural net might be best or random forest combined with a few weak learning XGboost algos and a recurrent neural net. And maybe, maybe when trying to predict the simple Buffalo Bills you need just a regression but when predicting the Pats you do need a complex amalgam of algorithms for the most efficacy. It could also be true that different constructions could be better if a team is playing in the rain or snow or a dome. What I’ve just outlined is really the difference between a machine learning system and an effective AI system. And with that, my plan:
1) I stopped working on my fantasy football prediction project because I concluded I needed play by play data which was out of the scope of my core business and too expensive. I first tried to forecast individual player performance but there was too much volatility in the touchdown models (which I separated from yardage models). Next I tried forecasting scores in games individually using team based numbers and then distributing the touchdowns to players individually which yielded better results but I realized I did need better data because when I tested my system live my eye test vs the results told me that I wasn’t accounting for the individual CB v WR and line play matchups well enough. IE if you just take a team’s rankings vs. opposing WRs you are grouping too much because some teams have 1 shutdown corner that can take a top WR out of the game and others have 3 great ones that is better neutralizing a pass attack with an even distribution of WRs. I suspected the problem would be too complex for the data I had, but you always learn things by trying out a solution with whatever you have, and I determined these to be the next step in building something that might have more efficacy.
a. A play type forecaster - this is how you solve your little timeout problem with game theory. Feed in game scores and timestamp every play with amount of time left, quarter, down, distance and then start with something I’d say as simple as timeout, run offensive play, FB, punt, run defensive play, FG block, punt block, punt return then 2-point conversion as well. Each of these will have a win probability associated with choosing that type in a vacuum. This isn’t a very hard system to build and something that either we should have built after Shumur butchered not taking the timeouts in the Saints game with 1:44 left and if not just subscribed to EDJ Sports at the very least, people ask why are you so sure the Giants don’t have good in house analytics? This is Exhibit ABCDEFGHIJ not getting this kind of simple game theory-based prediction correctly in game should drive any technologist crazy. So it’s not really a leap at all to suggest that we don’t have a good technologist in house.
b. An outcome forecaster - First you would need one of those handy dandy autoencoders that the Patriots have. This would convert film into signals that could be processed with timestamps. I will not get into how to make one of these (must hold something back) but there are many different ways to make these learn better IE do you want to put a label on what a play is? Do you want to label if there appears to be an audible? Do you want to label if your receiver is on an option route? The more you teach the machine, the better it understands what it is seeing and the better it will predict). Throw in the Zebra data. As well as, as much individual information on the players as possible attached with player IDs. This can give you the beginnings of an excellent per play forecaster where different types of plays and play outcomes are assigned probabilities and these can be rolled up into yardage and expected points for quarters, games, then seasons.
c. Combine a and b the better you label the better the more you can get out of it. IE run off guard to the right vs. off tackle to the right, what are the different win probabilities deltas given the situation?
d. Improve labeling efforts for better results IE was there a star on this play? Was a player victimized? These are the foundations of matchup exploitation.
This is also just one segment, the approach for a draft model, injury model and even personality model weighing psychological questions as well as social media data would could all be nested into a broader ROI / team construction model. Best practices on this is not to have a fixed algorithm or analysis structure but instead a dynamic one which also considers feedback from the system which is also engineered. IE encoding after the fact for certain situations what various staff members would have done. Or forming this feedback in different ways to send different information into the system like asking who would you have picked in the draft is a very different question than rank your top 10 picks in that spot or project stats for all of the following players or a complex draft simulation methodology where all of your scouts individually pick players given many random simulations of who is available. It looks like we are deficient in various areas needed to build this system on staff as well as an understanding of the kind of way an organization needs to position itself in order to attract the right kind of people.
A final note on reinforcement learning. Google it, understand it. I posted about Chess Centaurs and the biggest part that is needed to understand the value reinforcement learning is that the data created by subject matter experts interacting with predictive data is most probably an even more valuable data set than the predictions themselves. There is nothing better for a reinforcement learning system than various opinions as well feeding in other sources like an EDJ model which has it’s own win probability data which is only additive. All that being said, using an outside firm or outside data is not only delaying the creation of our own data set but it is hurting our competitive advantage overall and I’ll tell you why. This architecture I laid out, different teams, different architects would all end up with different code and different results even if both systems were built by world class talents there are so many ways to “skin the cat” here. Even if we have Shurmur or Gettleman leveraging these systems (which there is no evidence whatsoever of and evidence to the contrary) using outside firms or people we are giving away our institutional knowledge instead of encoding and storing it ourselves where only we can leverage it. If we believe we have the best people, it is quite dumb to engage any outside company to build our system. Even if we are working with a provider exclusively for now, unless it is a lifetime exclusive, we are essentially saying hey, let us help you train your algorithms with our experts, so you can then sell it to other people.
Again, I want to say that I have no ax to grind with anyone here, especially not Gettleman who I want to succeed. I do want him to start hiring the right people though. And also I think it is important as a fanbase that we hold the team accountable to this which is why people like Fatman bother me because even if he convinces one person this isn’t a problem when there is plenty of evidence that it is a legitimate one, from a probability standpoint, there is a lesser chance that a move is made because the Mara’s actually do care about public opinion.
Other teams might have systems like this built out already or the team in place to do so as shown in the broad analytics article I shared. This industry is billions of dollars we should be able to spend the $1-2M a year needed to build out a proper implementation internally even if we want to continue to outsource in the short term. We need to see that Giants page updating with real technical minds soon, there is absolutely no reason to be secretive about it, if we have something going it's time to spill the beans so we can begin to organically attract many more of the kinds of people needed to win the technology arms race already underway.
it may reveal who might have been spoken to and how wide the hidden net went. Not for the GM job (the one where someone in charge without a time of drift during the end of year and pre-fa and draft evaluation process) but for the coaching job.
B2, not to miller the thread but not on here much. Wishing you and yours a very happy holidays.
Take a definitive statement like “we had a good draft” people want to say this as if it’s a fact, some of our favorite luddites on this board are brandishing this with their I <3 Gettleman T Shirts. Yet I would say the jury is very much still out on this. And I know this may be difficult for some to understand but this statement has nothing to do with who Barkley is as a player who I love more than any Giants in a long time. I think from a game theory analytical perspective it is healthy to look at Barkley’s closest comps in lieu of complex model results that I don’t have. Players like Peterson and Sanders two of the very best RBs in history. The Vikings were 80-79 with Peterson on the team and the Lions were 78-82 with Sanders. If you look at the records after their rookie deals 44-51 and 48-48 it brings up some important questions. Questions that clearly concerned the Steelers with Le’Veon Bell. Now they hired this Carnegie Mellon professor in 2015 and their team seemed to think that it could win with Bell at $15 Million but not at $18 Million this is the kind of thing that absolutely points to analytics. Here is where Gettleman comes in and fateful clip. My issue is not with the pick of Barkley. My issue his response to this question. He had the wrong answer. It did not reflect an understanding behind probability theory necessary to minimize the mistakes in his data driven job. The bottom line is we don’t know if Barkley is not so good at a time that many of us all agree that the team isn’t ready to compete that he requires resources that from a team construction standpoint will not allow us to compete when we are ready. Say Barkley like Bell and because of the wear and tear on his body in 3 years says I want $25M a year or I sit? Don’t you dare franchise me. Do we know that he won’t do that? Do we know that he won’t break down? Can a team win with an RB making $25M a year? $20M? What if he does get injured? The point is, there are still worlds out there where this is a good or bad move and “getting into the analytics of it all” is an important part of a process in ensuring you are making these choices for all the right reasons. The right answer is we weighed a lot of factors and believe we made the right choice or something in that vein. What makes this clip even more troubling is that he uses Jonathan Stewart as an example to defend his point. There is no other way to slice this, but he looks very stupid for saying that. Not because he fell off the cliff, there was only some probability that that would happen this year. But the bravado which he states he hasn’t fallen off yet. He is framing the problem in the wrong way mathematically. The fact that Stewart hasn’t dropped off and his age in fact raises the probability that he will imminently, it is not a basis to think a player is immune to wear and tear. Now it would be nice if it looked as though when they were evaluating signing Stewart that someone was in his ear reminding him of this. Perhaps our GM wouldn’t say things that he honestly should regret saying, he didn’t need to go out on a limb like this, but it shows the kind of overconfidence in an outdated process that should send up red flags to any technologist. (I’d be interested in your take on this as a technologist Bill2, I share this clip so often because all my friends that are football fans and data people have not stopped mocking me about it)
This leads me to my next point. You could say back that the contracts Gettleman handed out to these people were mistakes anyone would make, I say back, well, are all teams using the same search parameters now? Could we be more backed into a corner in ways of fixing our team because of our process? Could things that worked for Gettleman 5 years ago not have the same efficacy now given the advancements of other teams? This is where the article shared today on another thread comes in on the current state of analytics. First, they bring up the Giants and the Gettleman video as an example of a team that is behind. Do any of us really want the Giants to be a team that stands out for this? Shouldn’t our ownership be ashamed of itself given its access to Wall St. and being a hub of business that we are a team on the wrong side of this article? Is it not embarrassing that given our record against the Eagles that they have had an advanced analytics program for 20 years and a guy with Ty Siam’s background is our rising star? This broad analytics article shared in another thread very much supports a lot of the points I have been making all season.
Quote:
We’ve reached this high point for a couple of reasons. The rise of smarter, younger GMs and coaches is part of it. A bigger part of it, though, is the spread of the NFL’s player-tracking data, which is being shared leaguewide for the first time this season. Having access to that data allows teams to build models to analyze plays and players differently, and to simply know more about the game. That’s been a boon to a movement that had already been embraced by a handful of the smartest teams. As other teams try to catch up, they’ve created an arms race to get the best numbers. Essentially, the smartest teams are getting significantly smarter, the average teams are trying to get better, and the dumbest teams are going to be very dumb if they don’t act soon.
“It’s about translating that data ASAP and being very, very in tune with the numbers. You can’t be a year behind, you can’t be a month behind.” —Thomas Dimitroff, Atlanta Falcons general manager
“It’s about translating that data ASAP and being very, very in tune with the numbers. You can’t be a year behind, you can’t be a month behind,” said Falcons general manager Thomas Dimitroff.
Teams are examining details they’ve never studied before to get an edge. One scouting department graded a defensive back prospect as an undrafted free agent due in part to his slow 40-yard dash. When that department was able to measure his game speed using tracking data, it determined that it should have listed him as a midround pick. Other teams in the market for linebackers have homed in on what kind of closing speed elite tacklers need. For instance, Zebra Technologies, a company whose MotionWorks service collects game-day data, found that the Cowboys’ Leighton Vander Esch reaches 16 to 17 miles per hour on his best plays.
“Teams can go deep on rosters to leverage the tracking data to scout players for the future, maybe in free agency,” said John Pollard, vice president of Zebra Technologies.
I’ve been saying for a bit now that this is the reality. And I’ve been criticized for bringing up age in relation to this, but this article does as well, and it is generally true that younger people are more interested in incorporating new ideas. I’ve brought up Zebra several times because I do think that is one of the keys, but it is honestly just a small part of what it is I think we need to do. This arms race and articles like this are also important to note. If you have an original football model or ideas what are the chances you shoot an email to a guy like Gettleman vs. Ernie Adams or Howie Roseman? This is a problem that if we don’t start addressing today, we ARE doomed to fall in a deeper and deeper competition chasm. We need to be known widely as a team that embraces new ideas and technologies not scoffing at it. And the fact remains, and I don’t think this can be emphasized enough, even if we compete the next few years with Gettleman he could be causing damage in competitive advantage that lasts multiples of years longer. This is why I really am trying to emphasize that while the jury is out on Gettleman’s ability to compete today, there is enough evidence that we need to bring a strong analytics mind near the top of our organization ASAP if we want to prevent a serious long-term competitive disadvantage.
Reading between the lines of these threads I’d say that it might be the case that Bill2 either knows the Tisch family or is one of the people the Giants talk to about these issues. So despite the fact that I’ve resisted doing this given the fact that I perceive this to be valuable information that I could be paid for I’m going to go ahead and share it because if they aren’t already implementing a system like this, they should be. And I’ll be the first to say, great if they are but their internal staff does not suggest they have the knowhow to do this and I will explain that even if they are working with an outside firm this is not optimal to gain or maintain any kind of performance advantage.
It’s best to start with how to frame these problems from a high level. People that know me as a big optimist, I think you need to be one in order to make good forecasting systems. Every time I begin to try to predict something new the first take is going to be nothing compared to what I end up with after a few months of work. I’ve predicted NPV of cash flows for every securitized mortgage in the US, huge college and corporate loan packages, stock outcome probabilities and index addition probabilities, trade predictions for a global banks entire fixed income portfolio (22K ISINs paired with 13K clients), healthcare classifications and at one point I even dabbled in Football. I’ve done this for many years, each time I improve my process and the technology platform I use gets better and smarter. This is all to say even if we hired the best person in the world at making football predictions or our staff learned really well on the job there are things you pick up in academia trying to solve many different signal processing issues or in the working world dealing with many different problems that we are still most likely falling short in. This is why I brush back at the idea that Ty Siam or Buffalo IT consultant guy can accomplish something like what is needed to provide value in football. There is nothing to suggest that they can even leverage open source tree-based algorithm code, let alone boosting, bagging or neural net code. They would have to have a deeper understanding of this before they could build bespoke systems that can try to optimize around different hyperparameters and time series modifications within the training of algorithms which requires writing or changing code within these algorithms. Situational ensembling is also quite necessary for these complex non-linear equations IE if I want to predict the outcome of the next play is a single neural net might be best or random forest combined with a few weak learning XGboost algos and a recurrent neural net. And maybe, maybe when trying to predict the simple Buffalo Bills you need just a regression but when predicting the Pats you do need a complex amalgam of algorithms for the most efficacy. It could also be true that different constructions could be better if a team is playing in the rain or snow or a dome. What I’ve just outlined is really the difference between a machine learning system and an effective AI system. And with that, my plan:
1) I stopped working on my fantasy football prediction project because I concluded I needed play by play data which was out of the scope of my core business and too expensive. I first tried to forecast individual player performance but there was too much volatility in the touchdown models (which I separated from yardage models). Next I tried forecasting scores in games individually using team based numbers and then distributing the touchdowns to players individually which yielded better results but I realized I did need better data because when I tested my system live my eye test vs the results told me that I wasn’t accounting for the individual CB v WR and line play matchups well enough. IE if you just take a team’s rankings vs. opposing WRs you are grouping too much because some teams have 1 shutdown corner that can take a top WR out of the game and others have 3 great ones that is better neutralizing a pass attack with an even distribution of WRs. I suspected the problem would be too complex for the data I had, but you always learn things by trying out a solution with whatever you have, and I determined these to be the next step in building something that might have more efficacy.
a. A play type forecaster - this is how you solve your little timeout problem with game theory. Feed in game scores and timestamp every play with amount of time left, quarter, down, distance and then start with something I’d say as simple as timeout, run offensive play, FB, punt, run defensive play, FG block, punt block, punt return then 2-point conversion as well. Each of these will have a win probability associated with choosing that type in a vacuum. This isn’t a very hard system to build and something that either we should have built after Shumur butchered not taking the timeouts in the Saints game with 1:44 left and if not just subscribed to EDJ Sports at the very least, people ask why are you so sure the Giants don’t have good in house analytics? This is Exhibit ABCDEFGHIJ not getting this kind of simple game theory-based prediction correctly in game should drive any technologist crazy. So it’s not really a leap at all to suggest that we don’t have a good technologist in house.
b. An outcome forecaster - First you would need one of those handy dandy autoencoders that the Patriots have. This would convert film into signals that could be processed with timestamps. I will not get into how to make one of these (must hold something back) but there are many different ways to make these learn better IE do you want to put a label on what a play is? Do you want to label if there appears to be an audible? Do you want to label if your receiver is on an option route? The more you teach the machine, the better it understands what it is seeing and the better it will predict). Throw in the Zebra data. As well as, as much individual information on the players as possible attached with player IDs. This can give you the beginnings of an excellent per play forecaster where different types of plays and play outcomes are assigned probabilities and these can be rolled up into yardage and expected points for quarters, games, then seasons.
c. Combine a and b the better you label the better the more you can get out of it. IE run off guard to the right vs. off tackle to the right, what are the different win probabilities deltas given the situation?
d. Improve labeling efforts for better results IE was there a star on this play? Was a player victimized? These are the foundations of matchup exploitation.
This is also just one segment, the approach for a draft model, injury model and even personality model weighing psychological questions as well as social media data would could all be nested into a broader ROI / team construction model. Best practices on this is not to have a fixed algorithm or analysis structure but instead a dynamic one which also considers feedback from the system which is also engineered. IE encoding after the fact for certain situations what various staff members would have done. Or forming this feedback in different ways to send different information into the system like asking who would you have picked in the draft is a very different question than rank your top 10 picks in that spot or project stats for all of the following players or a complex draft simulation methodology where all of your scouts individually pick players given many random simulations of who is available. It looks like we are deficient in various areas needed to build this system on staff as well as an understanding of the kind of way an organization needs to position itself in order to attract the right kind of people.
A final note on reinforcement learning. Google it, understand it. I posted about Chess Centaurs and the biggest part that is needed to understand the value reinforcement learning is that the data created by subject matter experts interacting with predictive data is most probably an even more valuable data set than the predictions themselves. There is nothing better for a reinforcement learning system than various opinions as well feeding in other sources like an EDJ model which has it’s own win probability data which is only additive. All that being said, using an outside firm or outside data is not only delaying the creation of our own data set but it is hurting our competitive advantage overall and I’ll tell you why. This architecture I laid out, different teams, different architects would all end up with different code and different results even if both systems were built by world class talents there are so many ways to “skin the cat” here. Even if we have Shurmur or Gettleman leveraging these systems (which there is no evidence whatsoever of and evidence to the contrary) using outside firms or people we are giving away our institutional knowledge instead of encoding and storing it ourselves where only we can leverage it. If we believe we have the best people, it is quite dumb to engage any outside company to build our system. Even if we are working with a provider exclusively for now, unless it is a lifetime exclusive, we are essentially saying hey, let us help you train your algorithms with our experts, so you can then sell it to other people.
Again, I want to say that I have no ax to grind with anyone here, especially not Gettleman who I want to succeed. I do want him to start hiring the right people though. And also I think it is important as a fanbase that we hold the team accountable to this which is why people like Fatman bother me because even if he convinces one person this isn’t a problem when there is plenty of evidence that it is a legitimate one, from a probability standpoint, there is a lesser chance that a move is made because the Mara’s actually do care about public opinion.
Other teams might have systems like this built out already or the team in place to do so as shown in the broad analytics article I shared. This industry is billions of dollars we should be able to spend the $1-2M a year needed to build out a proper implementation internally even if we want to continue to outsource in the short term. We need to see that Giants page updating with real technical minds soon, there is absolutely no reason to be secretive about it, if we have something going it's time to spill the beans so we can begin to organically attract many more of the kinds of people needed to win the technology arms race already underway.
TL;DR - You're Welcome, Shock.
there is little doubt the Giants are behind at some aspects of analytics. no doubt.
But I dont know if the mistakes of the past wont tighten up the talent evaluation side of decision making. that's usually the side of things we would know the least about.
all we would see is generally sharper decisions.
lastly, right now, imo, they dont have enough talent to execute game analytics.
and I sure hope someone does something about better game time management.
that's been a problem from the TC era and they cant seem to adopt best practices
one thing we might not be aware of is who ( it might not be the GM) is responsible for a greater investment in analytics? I have no idea whose portfolio that falls under. So I have assumed the coach on his stuff and the GM on his and the FO on contract risk management
Of course - although with some comfort that there were other options evaluated, and that maybe one of those might not have made the same mistakes.
Without having done so - all you can do is wonder if the Giants painted themselves into a corner by not exploring all options.
Fair enough.
I think you should know that I hold you and your opinions in high regard. IMO you are truly a valued BBI contributor.
Cheers and hope your holiday season is/has been a great one.
The feeling is mutual Dan, and I thank you for taking the time to articulate your views even if I don't agree with them. I have said before and I'll say again - I hope I'm proven wrong with my feelings about the Giants management and their competence, because I'd much rather watch meaningful football than be right about this.
Happy holidays and a happy and healthy 2019 to you and yours!
It depends who they spoke with to a large degree. In other words, we they just kicking the tires, doing courtesy interviews, meeting the Rooney Rule requirement, etc.
So maybe, but doubtful because they still played it safe.
This organization needed to be bold for a change. Reinvent themselves and their view of the football landscape. At least fail forward for a change...
I’ve actually been in loose contact with the ownership group of another team for years so the work I’ve done here isn’t entirely benevolent I’m going to send it over to them soon. Will definitely incorporate your comments, I’ve suspected you were involved in some powerful analytics since it was just a glimmer in my eye in terms of a subject matter I very much wanted to gain some mastery in. Certainly on the way you’ve discussed some geopolitical issues in the past it was pretty clear.
Unfortunately, I suspect that these criticisms are true. So I actually do hope some folks on the Giants read some of this stuff, and maybe it results in some hard self evaluation which leads them to make some changes.
I have no way of knowing but I think Mara was living in a bubble sent upwards from Reese and Ross and McAdoo that he finally got the downside of delegating to "football men"
I dont pay any attention to what they say. I am hopeful that the speed at which decisions and moving on past mistakes are made are a positive sign that this may be 2 to 3 years instead of the hopeless horror of 1966 to 1983.
I have no way of knowing but I think Mara was living in a bubble sent upwards from Reese and Ross and McAdoo that he finally got the downside of delegating to "football men"
I dont pay any attention to what they say. I am hopeful that the speed at which decisions and moving on past mistakes are made are a positive sign that this may be 2 to 3 years instead of the hopeless horror of 1966 to 1983.
The problem as I see it, is that a guy with much better instincts and management capability such as Gettleman vs. Reese (I am giving Gettlemen the benefit of the doubt here, but I truly believe it) can make the brod or as you put it crude decisions now and make them effectively.
As the team ascends, and the obvious issues are addressed, the remaining issues become increasingly nuanced. There are many variables to reaching a championship all they way down to a single bad call by an official which you obviously have no control over. But the probability of reaching the mountaintop is tightly intertwined with sustained success across a broad front.
The decisions that drive sustained long term success, should be deeply rooted in analytics. The decision to cut bait a year early rather than a year late, making all kinds of adjustments. Even driving overall scheme due to the strengths and weaknesses of your players. WIthout rooting the decisions in cold hard data, you see the team ascend, have a window of opportunity, the win or lose, descend and start over. I much prefer ascending and staying near the top for long periods. ;)
Fooled by randomness in a book I read early in my investing career before I programmed my first algorithm. Even if Gettleman is successful it doesn’t mean we have set up the infrastructure to succeed long term. Ignoring all the other teams I think it’s pretty clear that the Pats have that infrastructure. Why even emphasize secrecy when that’s out on the open and it puts us at a competitive disadvantage when it comes to recruiting.
I think it’s really a shame because I feel like one call
from Gettleman to Brian Cashman could totally change the trajectory of this team. And ultimately that’s the problem with a personality type like Gettleman when he speaks the way he does. Could he call someone like Cashman and learn how to delegate analytics to the proper
people? Could he call Christian or Dan in the springs or Bill2 or McL or me and say go build me something I can use. (Again not saying any of us are the right people but there are enough like us that we shouldn’t be getting timeouts wrong) He is overdue in doing that. And anyone that doesn’t think at least that conversation is worthwhile isn’t looking out for the best interests of the team whether they know it or not.
For the love of god if you don’t want to pay for the analytics just read the a) part of my post. That is all you need to do and would take a week and about 10K to build with overseas developers, you have the data.
We might not be asking the right questions.
We might not be interpreting the data correctly.
We might not have the appropriate data to answer a key questions.
If you are trying to interpret the data, and are asking questions and you fail. Then you learn from it, and get better at asking questions, and get better at interpreting the data.
If you don't ask the questions, you don't try to interpret the data, you don't get better.
The data is NEVER wrong!!!