http://dailydolphin.blog.palmbeachp...ieve-top-5-pick-than-playoffs-in-17-via-espn/ Anyone who is connected to our player via social media, I say we send them this article... We need to play angry this year anyways, why not start the party early...
Yeah.. this is ESPN saying "well.. if our non-transparent proprietary QBR system is getting panned, let's up the ante and create something even MORE complex called FPI that we'll still keep proprietary and see if that gets any cred!" As with QBR they have their usual pronouncements of how objective their approach is: http://www.espn.com/nfl/story/_/id/13539793/espn-nfl-football-power-index-debuts http://www.espn.com/nfl/story/_/id/...ootball-power-index-was-developed-implemented including this quote: Yet.. as usual they keep it proprietary lol. ESPN!! Publish the damn code so we can look at it! In any case, let's just see how good this FPI is at prediction. I couldn't find the comparable predictions prior to the 2016 season (seems to require ESPN insider subscription) but I did find their updated FPI after week 5 last year: http://www.espn.com/blog/statsinfo/post/_/id/124728/espns-football-power-index-for-oct-4 What's the correlation to win% between FPI after 5 weeks in the 2016 season to the final win totals? It's a respectable 0.506! Ahh.. then the hammer hits. I looked at the correlation between win% in 2015 to win% in 2016 (no damn model here.. just using actual win% from 2015 to predict win% in 2016) and guess what THAT correlation is: 0.698!!! lol so just using past season win totals to predict next season win totals does much better (at least for 2015 to 2016) than ESPN's hideously complex FPI that adds in more stuff than QBR (an "uber-QBR" so to say) in predicting win%. Summary: no need to respect this FPI. EDIT: I mistyped something in my code (happens.. which is why I double check afterwards lol).. that 0.698 is correlation between ESPN's FPI after 5 weeks in the 2016 season to the 2015 win totals! .. meaning their FPI for 2016 predicts results for 2015 better than for 2016. OK.. so it's not that it does worse than predicting from past win totals.. it's that it's mostly BASED on past win totals. Sorry about that mistake.
More garbage from ESPN to push their usual agendas. Thank you, cbrad. Any model that involves using subjective and biassed views of which teams and players are better than others as part of its formula is bunk. And the idea that the Dolphins are worse than every AFC team outside the division other than Cleveland and Jacksonville is just trolling us, frankly.
The players won't go that deep with it, they'll just take it as disrespect..that's what I want. In this case analytics and stats do not tell the real story..
Well, I'm guessing that there's some strength of opponent in there, margin of victory, etc. to spit out what they consider an accurate number. And when we look back at last year, we had some really ugly wins against some really bad teams. As a fan, I say, "Who the hell cares...our team rallied and won ten games after starting 1-4. We did that with mostly second stringers...so screw what anyone's performance metrics might say." But as an analyst, a math nerd or whatever you want to call it, I know we crapped the bed against Seattle, New England, Baltimore, etc. Then I see that we won most of those 10 wins by 7 points or less- Moore's Jets game was the only real blowout we had all year long. And truth be told, we should have blown out the Jets along with the Browns, Bengals, Rams, 49ers, etc. Yet we didn't....we barely squeaked by in most of those games and even lost one of them. So as a fan, I think with a better defense and lots of returning players off of injured reserve, we should repeat our 10-6 performance and maybe even beat most of the better teams in the AFC. But as a skeptic, I'm thinking that we need to do a whole heck of a lot better than what we showed last year. The outcome shouldn't be in doubt with 2 minutes left in a contest when we're playing some of the worst teams in the league (Browns, 49ers, etc) and there's big problems on both sides of the ball. On defense, we have to stop the run and our corners/safeties can't keep playing solid 90% of the time and then giving up big plays. On offense, we can't keep taking the first half off and then clawing our way back in contests. I mean, we'll take it as fans any day of the week...who doesn't love a big comeback? But that's not the recipe to being a dominant playoff team. I guess what I'm trying to say is that although I think ESPN's prediction is bull****, it also accurately represents how we played last year.
Thats a fact that we all know too well, and yet so far this offseason, the national media seems to be totally ignoring.
The very basic stats, such as win differential put us around 8-8 or 7-9, basically what Vegas has us at. I respect the Vegas books far more than ESPN because they have money riding on the outcone. I respect FO's DVOA and PFF's ratings more than ESPN because their business model is built around getting their stats right. Again they have real money riding on being right. Who would you trusy to repair your car? The unqualified mechanic who's had his shop for 20 years or the trust fund kid who's really into cars and has been to the most expensive school?
None of the above, I guess? I'm guessing none of them are throwing out the first five games of 2016 when they're making these predictions, and they need to in order to begin to get an accurate assessment of the 2017 Fins. Those games have no bearing on who and what the team now is, and they really drag down the overall statistics. And as DJ said, we lost an enormous amount of games to important players last season, and yet still won. An intelligent human being can look at those and come up with a relatively good picture of who the 2017 Miami Dolphins are going to be. I'm guessing that a lot of computers cannot.
Here's an interesting site that archives Vegas odds! http://www.sportsoddshistory.com/nfl-win/?y=2016&sa=nfl&t=win&o=t So.. how accurate is Vegas? Looking at the 5 years 2012-2016 Vegas' predictive power is extremely stable. The average deviation above or below predicted win totals lies between 2.05 to 2.27 wins (REALLY small variation in mean deviation eh?). This suggests we should expect whatever they predict +/- 2 wins. In our case that means it should be 5 to 9 wins. Interestingly 4 out of 5 years the "under" won more often. Things balance out only when you group "under" and "push" into one group and compare to "over". I wonder if that bias holds over time.. Oh, and Vegas during Philbin's tenure was scary accurate. 2012-2014 they got the win totals exactly right! 2015 they overestimated by 3 points and last year they underestimated by 3 points.
yeah they know the teams,I often wonder who works for them? however, there are always exceptions..I believe they will be wrong this year, including the deviation.
While that is bad by Vegas standards, how does it compare to other prognosticators and the media generally? The question also isn't about how accurate Vegas is in one slice of the data, the question is hiw accurate is Vegas over the entire league over longer time scales. My original point although is that Vegas has a financial incentive to get their projections right. ESPN has an interest in gerting the most clicks (outrage always gets more clicks) for their projections.
Oh yeah no doubt. ESPN doesn;t care if they are right, but Vegas does and that was particularly bad of them, IMO.
Well as ESPN is the final word on everything sports I guess I can skip the season and wait until next year. Thanks for the extra free time Bristol. By the way how's the balance sheet doing these days? Funny how that analytics department couldn't see that coming.