Jul 10, 2023·edited Jul 10, 2023Liked by Seaside Joe
After reading the interesting comments...I decided to dumb it down to answer the simple question "Did Seattle have a good offense last season?".
Season is singular...one season. Therefore, breaking the season into smaller sections is maybe not relevant. And it's very easy in the NFL stats world to go way too deep into minuscule numbers and lose your way...so KISS is a reasonable approach.
I'd consider "good" to mean top third of the league, compared to "average" and "poor" for the middle and bottom tiers respectively. Seattle landed in the middle tier (#13) based on yards/game offensively. However, they were 9th in offensive points scored per game, which would be in the "good category". I don't want to consider wins/losses or other metrics that are heavily influenced by the rest of the team, coaching, schedule, injuries, blah blah blah. Since these two key offensive metrics (yards/game and offensive scoring/game) conflict regarding "good" or "average" rating, I'm falling back on my gut...which is large and serves me well in times of stress (see what I did there?). Based on my gut, the Seattle Seahawks had an average offense in 2022, with both good and bad games, and and trending in the right direction for the future.
I agree with many others here that all the factors point to Seattle having a "good" offense in 2023, almost certainly. If we don't...many of us will be very disappointed. If I were going to predict...I'd say we'll have a good offense, average defense (also trending up), and win one or more playoff games.
As an aside...the comments here are always excellent and great reading...often as good or better than the stories (no offense Ken).
As you point out, it doesn't really matter why the offense was less productive in the 2nd part of the season, but I always assume, if the roster is still more or less in tact, then the problem is the league has adjusted after watching the tape, and we haven't adjusted back. I like that folks are excited about the offense going into next year, but I'm not assuming anything at this point. And I'm certainly not sure yet that Waldren is 'the guy' to lead this offensive to bigger things.
Man, I kept telling my friends, if this defense is just average, it will go a long way. Last year if they were, we'd be talking about a 12-13 win team. Pretty crazy when one of your fantasies is an "average" defense.
I enjoyed the article and wanted to comment on the offence but one BIG thing that kept jumping to the forefront in my mind was just how atrocious the run defense was from Tampa Bay til the end and I wonder how much affect that had on the way the offense had to operate, how much more pressure it felt, how many more chances they had to take. For me the offense was fun to watch but too many times were put in a difficult position to get the job done. By the change in personnel, I believe Hawks feel somewhat the same. There is quite the lull in interesting NFL info these days but SSJ continues to stimulate our thoughts. Thanks!
Jul 8, 2023·edited Jul 9, 2023Liked by Seaside Joe
I had to go work on SWMBO's new retaining wall for a couple of hours. Stuff like that only takes about 2/3 of my brain, so with the other cell I started to think that the original question might be a bit of a red herring. (Note to Ken: if I need to grovel a bit for that one, just let me know.)
Maybe the issue can't really be approached as "is the offense good?", or "is it good enough?". Maybe that sort of thing can really be asked only at the team level...which, in the context of having just looked at the team-level stats for last year's SEA-KC game, led to the following random musing:
What does the team look like that stands a reasonable chance of beating a healthy KC in the Super Bowl?
I'm thinking...let me check...yep, still just the one cell working on it...you have to do two things:
1) Deny KC's offense as many opportunities as possible; that is, hold the ball. Which, I think, means consistently converting third downs, running the ball well, and reliably completing short-ish passes. You could throw four passes, all from your own 25, all of them for untouched touchdowns...and lose 70-28 to those guys. Also, you could beat them by 20 minutes in the time of possession...and still lose 70-28.
2) You have to pressure their whirling dervish-leprechaun-Tasmanian Devil every which way but loose and simultaneously prevent any completions longer than about 2 yards (with, or without, the cloud of dust). Either one alone isn't going to be good enough.
I'm thinking that your team only has to barely survive everybody else well enough to get there. Which is in itself a pretty tall order.
Well last season was special to me. It showed character to have win the last game of the season and then to have to have(thank you) Detroit win. It made the whole season gratifying. They were expected to finish last in their division. The bottom , all washed up. Thank you Seahawks for showing that toughness. May the 12s be with you and Go Seahawks!
Geno was hot and cold, and we rarely got a good game from both DK and NoE. Walker was boom or bust, and in a way, our offense was like that. Domination over Giants, Lions and Chargers, good enough against Cardinals, struggling against the Rams and the playoffs against the 9ers (in the first half, with a meltdown in the second half), and lucky in our other wins. Bad against everyone else, especially the Niners.
One thing was how to play chess. I thought it was pretty cool that there were all manner of different openings (even tho' chess had only just been invented), and you could follow the pattern. Sadly, it turned out that you can't just memorize White's pattern & expect to win. You had to pay attention to the interplay between the two sides, and adjust each side's offensive approach to deal with how well it could defend itself.
The second thing I learned young was that I sucked at chess. Nonetheless, I have to wonder if once the rest of the league observed that the Fleahawks' defense could be gashed on the ground to consume a lot of clock, so the offense wasn't going to get on the field much...then would start to press, trying to make up for the lost ToP.
Being too cheap to pay for a high-end NFL statistical analysis site, I popped over to NFL.com's game-by-game summaries. I thought, briefly, that it would be fun to pull numbers, fire up Matlab, and run a thorough Analysis of Variance...but then, the crying started, I couldn't see the screen anymore, and my keyboard started to spark from the excess moisture. My dog (Spike) gets pretty upset when that happens, so I ceased & desisted.
I'm pretty sure the KC game was going to screw with the results, anyway. My Mark I eyeball doesn't suggest that the ToP fell off all that much in the season's second half, so...y'know...never mind.
Aside #1: I'm not sure I really know what "good" means, anyway. They were "good" enough to outscore their opponents more often than not, but not by enough or enough more often to look dominant. Peeking between the tears, the wild scatter in the numbers suggests that statistics, per se, might not mean much; team-by-team match-up might be more significant (and much harder to analyze). As we used to say in lethality assessment, you only have to be fast enough & accurate enough to whack the OPFOR before they can get off a well-aimed shot. Anything better than that doesn't mean much in practice.
Aside #2: Yes, "Fleahawks". That's what we called 'em in '76, and it stuck in my head. I don't know what the term is for "earworm, only for sarcasm", but my affliction is probably permanent.
To me, seeing a decline on the offense in the second half of the season points to a few things, but to me, one thing stands out. The team relied on a bunch of rookies. And none had played anywhere near that many games in a season. There was some wall-hitting and if that many players are having a slight bit of difficulty, it reflected in how well the team played.
While that crop of rookies should do better with the length of the season, the new crop could have similar issues this year.
A lifetime as a baseball fan has trained me to look at the entire season. Unless a team is really good or really bad, there are bound to be ebbs and flows. These are apparent in a 162-game season; less so in a 17-game season. In 2022, the Hawks had a big flow followed by a big ebb. All told, this gave them the 13th-ranked offense in the league in terms of yards gained, which puts them in the middle of the second quartile of teams. This seems about right to me.
Will the offense be better in 2023? I believe so. Here’s why:
* Geno and Waldron have a full year together
* The tackles are no longer rookies
* Center will be improved--there’s no possible way for the position to be worse
* There are three legit receiving WR options for the first time since 2013
* There should be the best depth at RB since 2013
* A TE group that combined for 109 receptions last year remains formidable and could get better depending on Parkinson’s development
Plus, there’s margin for error--Seattle doesn’t need all of the above to come true in order to have an improved offense.
I thought that getting Geno more weapons was as big an off-season need as improving the run defense. Schneider got him the consensus best WR in the draft and one of the best backs. Given how weak the NFC looks, the Hawks offense could be among the best in the conference.
Seaside, thanks for sitting back and letting this thread unfold. VI knows a lot more about this than I do—my knowledge comes from a few classes and employment as an editor for financial modeling papers written by statisticians. I have a lot of regard for real quants—I just don’t see the kind of analysis of football on Twitter that I’m used to combing through.
Honestly, it’s telling that the analyticals have chosen a 140-character platform to communicate. When Bill James revolutionized baseball, he did it by publishing articles in SABR, which was established in the early ‘70s and by writing an annual book. Just saying.
The Law of Unintended Consequences is a vicious and relentless taskmaster.
I read your post yesterday & thought "That's cool...I wonder who Vi is?" Then, because it was time, I cycled out to Carnation & back (near the north end of Lake Washington). About halfway back, another of those random musings popped top of mind: "VI...you talkin' about me?".
Now...if you were NOT talking about me, go ahead and enjoy yourself at my expense. But if you WERE, then...
It was not ever my intent to put anybody in the position of writing "the village idiot knows more about this than I do". I consider myself no expert on any of that stuff and, as we all know by now, an "ex" is a has-been & a "spurt" is just a drip under pressure.
I'm not changing my handle, though, because there will come a time when somebody...somebody who does not read what they write...literally does write that...they'll be wrong, too, but...damn, that's gonna to be a very, very good day.
"A lifetime as a baseball fan has trained me to look at the entire season."
That's something I've been trying to figure out how to separate and gain useful insights from. Obviously deconstructed data can be useful but splitting it into halves seems arbitrary because it could also be looked at in thirds and I can't figure what information can be gleaned in one scenario over another.
On the other hand, data from a single season only gives a general idea about the team because the offense could be ranked 15th overall but had a season where the first half they were 1st and a second half at 32nd which is more concerning than a team having the 15th ranked offense though the entire season.
Basically what I'm getting at is I see the data but how do I determine the useful parts based on the way it is presented?
Well yeah, I'm not arguing it isn't. I just want to understand the significance between different splits, overall stats and what I can learn from them.
Statistically, it’s harder to draw conclusions from a 17-game sample than a 162-game sample. Consider the question as such: What conclusions would you draw about a baseball team from a randomly selected 17 games? It’s not this simple—there’s a lot to be learned from 17 games of football—but suppose a football season lasted 25 games. Statistically, you wouldn’t be able to draw many conclusions about how the Seahawks offensive might perform over the final eight games.
I feel like no conclusions can be drawn until the data set is complete and at that point predictions are moot because there's always a chance for an anomaly.
Statistics can be great for helping one understand what has already happened, but it is an imperfect forecasting tool. (If it were perfect, we could read the future.) It
This is especially true in football with its short and unbalanced schedule, which contributes to making it the most conditional and situational of all sports (e.g., weather is a condition; down-and-distance is a situation). Randomness -- I.e., luck -- looms large as a result and--if you ask me--is a decisive factor in any close game. That’s why the SB loss never bothered me much--one team was going to be luckier and that turned out to be the Patriots.
I updated my latest response but to summarize, I feel like the statistics are almost fractal in how you can be more and more specific -- Completion percentage on 2nd down > On 2nd down in night games > night games with this specific oline and/or receivers > and it becomes so metaphysical. I realize I'm making this more complicated than I need to but where is the cutoff between useful information and too granular? What makes it useful or too granular?
I had to think about this for a while. "...the cutoff..." depends on how many different "factors" are involved (and how many "values" each factor can reach), and how many times each "test" was identically repeated. Crudely, you want to have lots more "tests" (and repetitions of tests) than you have combinations of factor values. It is not clear to me that a single game is a single "test"; a single play might be.
You can read the rest of this, if you're strong enough...but a nice mutton, lettuce, and tomato sandwich would probably make you a lot happier.
Truth in advertising: "it is not clear to me" means my BS detector has gone off but that I am cognizant of having been wrong before.
In this context, "factor" mostly refers to a thing that can, irrespective of anything else, be better or worse. (Lockett might have got a good night's sleep, and is very sharp today...) However, "factors" can interact with each other (...and Geno did, too, so the two of them are on fire). That interaction is, statistically, another "factor" to consider.
Which team you're playing this week is a factor. How many injuries of what severity the two teams incurred last week...and the weeks before...is another whole 'nother boat load of factors. To achieve a given level of confidence in the statistics, more factors means you need more "identical" repetitions of more combinations of factor values.
See, for example, "Statistics for Experimenters"; Box, Hunter, & Hunter; Wiley, 1978.
That MLT is starting to sound pretty good now, isn't it?
With some complicated ciphering, nerds can estimate how many "values" each "factor" can hit. That allows the nerds to add up how many combinations of values figure into the problem. Painting with a broad brush...from, like 140 feet away, with one eye closed & a log in the other...the nerds would call that number the "Degrees of Freedom" (DoF) in the problem. In a single NFL play, that number is pretty large. In a whole season for the whole league...fuggetaboutit.
There is another kind of DoF: not in the problem, but in the analysis. This has to do with how many samples you obtained, and how those samples were spread between different combinations of factor values and repetitions thereof.
Every time a nerd produces a statistic, they're supposed to "remove" (account for) the number of analysis DoF it represents & compare that to the estimated problem DoF. Contrary to popular belief, for example, an "average" (the "mean") should be calculated dividing by one less than the number of samples...because that calculation has "used" an analysis DoF. For really large sample sizes (e.g., number of repetitions), this works out OK. But for small sample sizes...I believe I recommended an MLT.
To make matters worse, most statistics are relevant only for "bell curve" distributions. There are the usual caveats, provisos, and quid-pro-quos on that statement, but characterizing the shape of the distribution takes more analysis DoF...in most cases, lots more, because in most real-world cases, there is more than just one distribution going on. Like, bazillions of them in every game.
Although the number of analysis DoF might look large, they disappear pretty fast when you slice and dice the statistics. If you do it right, and if you're strong enough.
Statistics with fewer analysis DoF than problem DoF are less trustworthy (or, more accurately, "it is not clear that they can be trusted"), but there's almost always an easier way to know when not to trust a reported statistic: if they didn't report how they accounted for the two kinds of DoF. If they just pop a number out there...run.
You'd do statistical mumbo-jumbo to determine where the "knees" are in the various curves. That's where I was headed with the ANOVA (which might well have blown up into a Multi-variate Analysis of Variance, a.k.a., "MANOVA"), before the weeping started. To do it right (-ish) you'd have to basically perform the analysis at the league level...because one team's "knee" might obscure another's. Even if I could control the flood of tears, the confidence intervals would likely be so large (because of the sample size) that I suspect you couldn't trust the answers.
So...Mark I eyeball...yours is probably better than mine!
Also, I'm not trying to call anyone out on anything, I'm genuinely interested to learn the how and why behind data/stats and not just the end result. It can just get more and more granular. YPC for the season, then per game...then you've gotta look at every players' performance and how it rated within the structure of the game plan and playcalls... what are their tendencies in night games vs morning games?
There's just so many variables that have their own variables that they may as well be fractals and I wanna know how to separate the wheat from the chaff, so to speak.
Great info and humor! Most people can only do one or the other, haha. I aprithe explanation because I've always felt sports analysis has been presented to us in a way that is technically true but there is a rabbit hole of statistics and data science that I just don't understand. It'd be nice to have a resident expert who could explain how and why any specific datum and its methodology matter
I thought the O-line play went downhill in the 2nd half of the season. For some reason people want to talk about K-9's stuff rate as if it was him trying to take everything to the house but there was waaaayyyyy too many times when he took the handoff and had a d-lineman right in his face. I think that gets better with another year under the young line's belt.
You couple that with the defense's inability to stop the run and you're going to get a lot of games where teams that aren't so good are going to hang tough.
My hope is that the defense does a lot better at putting offenses behind the chains and the offense does a lot better at not getting beat immediately at the snap. If that happens then I think this team will be a lot better.
If the D takes a significant step forward to at least be "average" that is going to also help the offense quite a bit. We know the offense is going to be better with the additions (JSN, Charbs, OL) and it needs to be--turnovers are killers, and Geno has to (and can) be "first half season" Geno for the Seahawks to do well.
But if the D can just get off the field (third and long conversions last year were gruesome to watch) that alone is going to improve Geno and Co's performance with more opportunities and possibly better field position.
After reading the interesting comments...I decided to dumb it down to answer the simple question "Did Seattle have a good offense last season?".
Season is singular...one season. Therefore, breaking the season into smaller sections is maybe not relevant. And it's very easy in the NFL stats world to go way too deep into minuscule numbers and lose your way...so KISS is a reasonable approach.
I'd consider "good" to mean top third of the league, compared to "average" and "poor" for the middle and bottom tiers respectively. Seattle landed in the middle tier (#13) based on yards/game offensively. However, they were 9th in offensive points scored per game, which would be in the "good category". I don't want to consider wins/losses or other metrics that are heavily influenced by the rest of the team, coaching, schedule, injuries, blah blah blah. Since these two key offensive metrics (yards/game and offensive scoring/game) conflict regarding "good" or "average" rating, I'm falling back on my gut...which is large and serves me well in times of stress (see what I did there?). Based on my gut, the Seattle Seahawks had an average offense in 2022, with both good and bad games, and and trending in the right direction for the future.
I agree with many others here that all the factors point to Seattle having a "good" offense in 2023, almost certainly. If we don't...many of us will be very disappointed. If I were going to predict...I'd say we'll have a good offense, average defense (also trending up), and win one or more playoff games.
As an aside...the comments here are always excellent and great reading...often as good or better than the stories (no offense Ken).
As you point out, it doesn't really matter why the offense was less productive in the 2nd part of the season, but I always assume, if the roster is still more or less in tact, then the problem is the league has adjusted after watching the tape, and we haven't adjusted back. I like that folks are excited about the offense going into next year, but I'm not assuming anything at this point. And I'm certainly not sure yet that Waldren is 'the guy' to lead this offensive to bigger things.
Man, I kept telling my friends, if this defense is just average, it will go a long way. Last year if they were, we'd be talking about a 12-13 win team. Pretty crazy when one of your fantasies is an "average" defense.
KC has an average defense and they're doing OK. I'm coming around to SJ's point of view, that you have to score points. Lots of points.
I enjoyed the article and wanted to comment on the offence but one BIG thing that kept jumping to the forefront in my mind was just how atrocious the run defense was from Tampa Bay til the end and I wonder how much affect that had on the way the offense had to operate, how much more pressure it felt, how many more chances they had to take. For me the offense was fun to watch but too many times were put in a difficult position to get the job done. By the change in personnel, I believe Hawks feel somewhat the same. There is quite the lull in interesting NFL info these days but SSJ continues to stimulate our thoughts. Thanks!
I had to go work on SWMBO's new retaining wall for a couple of hours. Stuff like that only takes about 2/3 of my brain, so with the other cell I started to think that the original question might be a bit of a red herring. (Note to Ken: if I need to grovel a bit for that one, just let me know.)
Maybe the issue can't really be approached as "is the offense good?", or "is it good enough?". Maybe that sort of thing can really be asked only at the team level...which, in the context of having just looked at the team-level stats for last year's SEA-KC game, led to the following random musing:
What does the team look like that stands a reasonable chance of beating a healthy KC in the Super Bowl?
I'm thinking...let me check...yep, still just the one cell working on it...you have to do two things:
1) Deny KC's offense as many opportunities as possible; that is, hold the ball. Which, I think, means consistently converting third downs, running the ball well, and reliably completing short-ish passes. You could throw four passes, all from your own 25, all of them for untouched touchdowns...and lose 70-28 to those guys. Also, you could beat them by 20 minutes in the time of possession...and still lose 70-28.
2) You have to pressure their whirling dervish-leprechaun-Tasmanian Devil every which way but loose and simultaneously prevent any completions longer than about 2 yards (with, or without, the cloud of dust). Either one alone isn't going to be good enough.
I'm thinking that your team only has to barely survive everybody else well enough to get there. Which is in itself a pretty tall order.
Are SWMBOS "swimbos"? People whose brains are addled by their love of swimming?
https://m.youtube.com/watch?v=WEO1QLwPceU
Not sure why but it reminds me of this
https://youtu.be/64yianfGvzc
Well last season was special to me. It showed character to have win the last game of the season and then to have to have(thank you) Detroit win. It made the whole season gratifying. They were expected to finish last in their division. The bottom , all washed up. Thank you Seahawks for showing that toughness. May the 12s be with you and Go Seahawks!
Great article!! Only Seahawks writer I know producing quality content this time of the year, keep on the great work!
Why Seaside Dynamite, Why?
In summary, there’s hope.
Short answer: no
Not so short answer: it wasn't a disaster
Geno was hot and cold, and we rarely got a good game from both DK and NoE. Walker was boom or bust, and in a way, our offense was like that. Domination over Giants, Lions and Chargers, good enough against Cardinals, struggling against the Rams and the playoffs against the 9ers (in the first half, with a meltdown in the second half), and lucky in our other wins. Bad against everyone else, especially the Niners.
I learned two things at a young age.
One thing was how to play chess. I thought it was pretty cool that there were all manner of different openings (even tho' chess had only just been invented), and you could follow the pattern. Sadly, it turned out that you can't just memorize White's pattern & expect to win. You had to pay attention to the interplay between the two sides, and adjust each side's offensive approach to deal with how well it could defend itself.
The second thing I learned young was that I sucked at chess. Nonetheless, I have to wonder if once the rest of the league observed that the Fleahawks' defense could be gashed on the ground to consume a lot of clock, so the offense wasn't going to get on the field much...then would start to press, trying to make up for the lost ToP.
Being too cheap to pay for a high-end NFL statistical analysis site, I popped over to NFL.com's game-by-game summaries. I thought, briefly, that it would be fun to pull numbers, fire up Matlab, and run a thorough Analysis of Variance...but then, the crying started, I couldn't see the screen anymore, and my keyboard started to spark from the excess moisture. My dog (Spike) gets pretty upset when that happens, so I ceased & desisted.
I'm pretty sure the KC game was going to screw with the results, anyway. My Mark I eyeball doesn't suggest that the ToP fell off all that much in the season's second half, so...y'know...never mind.
Aside #1: I'm not sure I really know what "good" means, anyway. They were "good" enough to outscore their opponents more often than not, but not by enough or enough more often to look dominant. Peeking between the tears, the wild scatter in the numbers suggests that statistics, per se, might not mean much; team-by-team match-up might be more significant (and much harder to analyze). As we used to say in lethality assessment, you only have to be fast enough & accurate enough to whack the OPFOR before they can get off a well-aimed shot. Anything better than that doesn't mean much in practice.
Aside #2: Yes, "Fleahawks". That's what we called 'em in '76, and it stuck in my head. I don't know what the term is for "earworm, only for sarcasm", but my affliction is probably permanent.
Ha! I had a sucking chess wound!
How did I NOT see that before?
If you consider last season's results against expectations they had a great offense. If they can exceed our expectations again I will be a happy 12.
To me, seeing a decline on the offense in the second half of the season points to a few things, but to me, one thing stands out. The team relied on a bunch of rookies. And none had played anywhere near that many games in a season. There was some wall-hitting and if that many players are having a slight bit of difficulty, it reflected in how well the team played.
While that crop of rookies should do better with the length of the season, the new crop could have similar issues this year.
A lifetime as a baseball fan has trained me to look at the entire season. Unless a team is really good or really bad, there are bound to be ebbs and flows. These are apparent in a 162-game season; less so in a 17-game season. In 2022, the Hawks had a big flow followed by a big ebb. All told, this gave them the 13th-ranked offense in the league in terms of yards gained, which puts them in the middle of the second quartile of teams. This seems about right to me.
Will the offense be better in 2023? I believe so. Here’s why:
* Geno and Waldron have a full year together
* The tackles are no longer rookies
* Center will be improved--there’s no possible way for the position to be worse
* There are three legit receiving WR options for the first time since 2013
* There should be the best depth at RB since 2013
* A TE group that combined for 109 receptions last year remains formidable and could get better depending on Parkinson’s development
Plus, there’s margin for error--Seattle doesn’t need all of the above to come true in order to have an improved offense.
I thought that getting Geno more weapons was as big an off-season need as improving the run defense. Schneider got him the consensus best WR in the draft and one of the best backs. Given how weak the NFC looks, the Hawks offense could be among the best in the conference.
Seaside, thanks for sitting back and letting this thread unfold. VI knows a lot more about this than I do—my knowledge comes from a few classes and employment as an editor for financial modeling papers written by statisticians. I have a lot of regard for real quants—I just don’t see the kind of analysis of football on Twitter that I’m used to combing through.
Honestly, it’s telling that the analyticals have chosen a 140-character platform to communicate. When Bill James revolutionized baseball, he did it by publishing articles in SABR, which was established in the early ‘70s and by writing an annual book. Just saying.
The Law of Unintended Consequences is a vicious and relentless taskmaster.
I read your post yesterday & thought "That's cool...I wonder who Vi is?" Then, because it was time, I cycled out to Carnation & back (near the north end of Lake Washington). About halfway back, another of those random musings popped top of mind: "VI...you talkin' about me?".
Now...if you were NOT talking about me, go ahead and enjoy yourself at my expense. But if you WERE, then...
It was not ever my intent to put anybody in the position of writing "the village idiot knows more about this than I do". I consider myself no expert on any of that stuff and, as we all know by now, an "ex" is a has-been & a "spurt" is just a drip under pressure.
I'm not changing my handle, though, because there will come a time when somebody...somebody who does not read what they write...literally does write that...they'll be wrong, too, but...damn, that's gonna to be a very, very good day.
TTYL.
"A lifetime as a baseball fan has trained me to look at the entire season."
That's something I've been trying to figure out how to separate and gain useful insights from. Obviously deconstructed data can be useful but splitting it into halves seems arbitrary because it could also be looked at in thirds and I can't figure what information can be gleaned in one scenario over another.
On the other hand, data from a single season only gives a general idea about the team because the offense could be ranked 15th overall but had a season where the first half they were 1st and a second half at 32nd which is more concerning than a team having the 15th ranked offense though the entire season.
Basically what I'm getting at is I see the data but how do I determine the useful parts based on the way it is presented?
162 games is a much larger statistical sample than 17.
Well yeah, I'm not arguing it isn't. I just want to understand the significance between different splits, overall stats and what I can learn from them.
Statistically, it’s harder to draw conclusions from a 17-game sample than a 162-game sample. Consider the question as such: What conclusions would you draw about a baseball team from a randomly selected 17 games? It’s not this simple—there’s a lot to be learned from 17 games of football—but suppose a football season lasted 25 games. Statistically, you wouldn’t be able to draw many conclusions about how the Seahawks offensive might perform over the final eight games.
I feel like no conclusions can be drawn until the data set is complete and at that point predictions are moot because there's always a chance for an anomaly.
Statistics can be great for helping one understand what has already happened, but it is an imperfect forecasting tool. (If it were perfect, we could read the future.) It
This is especially true in football with its short and unbalanced schedule, which contributes to making it the most conditional and situational of all sports (e.g., weather is a condition; down-and-distance is a situation). Randomness -- I.e., luck -- looms large as a result and--if you ask me--is a decisive factor in any close game. That’s why the SB loss never bothered me much--one team was going to be luckier and that turned out to be the Patriots.
I updated my latest response but to summarize, I feel like the statistics are almost fractal in how you can be more and more specific -- Completion percentage on 2nd down > On 2nd down in night games > night games with this specific oline and/or receivers > and it becomes so metaphysical. I realize I'm making this more complicated than I need to but where is the cutoff between useful information and too granular? What makes it useful or too granular?
You’re not being too granular. Plus, as a wise person around here once said, drilling down to this level takes the fun out of the game.
You challenge me, sir.
I had to think about this for a while. "...the cutoff..." depends on how many different "factors" are involved (and how many "values" each factor can reach), and how many times each "test" was identically repeated. Crudely, you want to have lots more "tests" (and repetitions of tests) than you have combinations of factor values. It is not clear to me that a single game is a single "test"; a single play might be.
You can read the rest of this, if you're strong enough...but a nice mutton, lettuce, and tomato sandwich would probably make you a lot happier.
Truth in advertising: "it is not clear to me" means my BS detector has gone off but that I am cognizant of having been wrong before.
In this context, "factor" mostly refers to a thing that can, irrespective of anything else, be better or worse. (Lockett might have got a good night's sleep, and is very sharp today...) However, "factors" can interact with each other (...and Geno did, too, so the two of them are on fire). That interaction is, statistically, another "factor" to consider.
Which team you're playing this week is a factor. How many injuries of what severity the two teams incurred last week...and the weeks before...is another whole 'nother boat load of factors. To achieve a given level of confidence in the statistics, more factors means you need more "identical" repetitions of more combinations of factor values.
See, for example, "Statistics for Experimenters"; Box, Hunter, & Hunter; Wiley, 1978.
That MLT is starting to sound pretty good now, isn't it?
With some complicated ciphering, nerds can estimate how many "values" each "factor" can hit. That allows the nerds to add up how many combinations of values figure into the problem. Painting with a broad brush...from, like 140 feet away, with one eye closed & a log in the other...the nerds would call that number the "Degrees of Freedom" (DoF) in the problem. In a single NFL play, that number is pretty large. In a whole season for the whole league...fuggetaboutit.
There is another kind of DoF: not in the problem, but in the analysis. This has to do with how many samples you obtained, and how those samples were spread between different combinations of factor values and repetitions thereof.
Every time a nerd produces a statistic, they're supposed to "remove" (account for) the number of analysis DoF it represents & compare that to the estimated problem DoF. Contrary to popular belief, for example, an "average" (the "mean") should be calculated dividing by one less than the number of samples...because that calculation has "used" an analysis DoF. For really large sample sizes (e.g., number of repetitions), this works out OK. But for small sample sizes...I believe I recommended an MLT.
To make matters worse, most statistics are relevant only for "bell curve" distributions. There are the usual caveats, provisos, and quid-pro-quos on that statement, but characterizing the shape of the distribution takes more analysis DoF...in most cases, lots more, because in most real-world cases, there is more than just one distribution going on. Like, bazillions of them in every game.
Although the number of analysis DoF might look large, they disappear pretty fast when you slice and dice the statistics. If you do it right, and if you're strong enough.
Statistics with fewer analysis DoF than problem DoF are less trustworthy (or, more accurately, "it is not clear that they can be trusted"), but there's almost always an easier way to know when not to trust a reported statistic: if they didn't report how they accounted for the two kinds of DoF. If they just pop a number out there...run.
Or, in my case, weep.
You'd do statistical mumbo-jumbo to determine where the "knees" are in the various curves. That's where I was headed with the ANOVA (which might well have blown up into a Multi-variate Analysis of Variance, a.k.a., "MANOVA"), before the weeping started. To do it right (-ish) you'd have to basically perform the analysis at the league level...because one team's "knee" might obscure another's. Even if I could control the flood of tears, the confidence intervals would likely be so large (because of the sample size) that I suspect you couldn't trust the answers.
So...Mark I eyeball...yours is probably better than mine!
Also, I'm not trying to call anyone out on anything, I'm genuinely interested to learn the how and why behind data/stats and not just the end result. It can just get more and more granular. YPC for the season, then per game...then you've gotta look at every players' performance and how it rated within the structure of the game plan and playcalls... what are their tendencies in night games vs morning games?
There's just so many variables that have their own variables that they may as well be fractals and I wanna know how to separate the wheat from the chaff, so to speak.
Great info and humor! Most people can only do one or the other, haha. I aprithe explanation because I've always felt sports analysis has been presented to us in a way that is technically true but there is a rabbit hole of statistics and data science that I just don't understand. It'd be nice to have a resident expert who could explain how and why any specific datum and its methodology matter
Thx.
I crack me up, too.
I thought the O-line play went downhill in the 2nd half of the season. For some reason people want to talk about K-9's stuff rate as if it was him trying to take everything to the house but there was waaaayyyyy too many times when he took the handoff and had a d-lineman right in his face. I think that gets better with another year under the young line's belt.
You couple that with the defense's inability to stop the run and you're going to get a lot of games where teams that aren't so good are going to hang tough.
My hope is that the defense does a lot better at putting offenses behind the chains and the offense does a lot better at not getting beat immediately at the snap. If that happens then I think this team will be a lot better.
If the D takes a significant step forward to at least be "average" that is going to also help the offense quite a bit. We know the offense is going to be better with the additions (JSN, Charbs, OL) and it needs to be--turnovers are killers, and Geno has to (and can) be "first half season" Geno for the Seahawks to do well.
But if the D can just get off the field (third and long conversions last year were gruesome to watch) that alone is going to improve Geno and Co's performance with more opportunities and possibly better field position.
We pretty much posted the exact same thing at the exact same time.
Yes lol... the connection between D and O is really important.