1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

2019

Discussion in 'Miami Dolphins Forum' started by tirty8, Aug 17, 2018.

  1. Fin-O

    Fin-O Initiated Club Member

    11,375
    11,392
    113
    Sep 28, 2015
    Actually it was a series of respectful debate on a topic that not everyone can agree on. You saw no insults, no cursing etc etc.

    So, “load of crap” Im assuming simply means you disagree strongly. :up:
     
    Redwine4all, Surfs Up 99 and KeyFin like this.
  2. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Personally, one reason I come here is to waste some time :smile:
     
  3. Fin-O

    Fin-O Initiated Club Member

    11,375
    11,392
    113
    Sep 28, 2015
    Cbrad.

    Can you help me devise a formula that will help predict the outcome of Sea @ Denver week 1?

    Thanks
     
  4. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Based on that suggestion you had weeks ago to focus more on prediction I actually scraped all game results in NFL history so I actually have some data to work with now. I only have basic stuff in that database like "year", "team", "week", "result", "score", "yards passing", "yards rushing", etc.. but it's a start.

    So let's do this one piece at a time. I mean, I don't even know basic stuff like how a team's record in the previous year predicts that team's record the following year. All that can be looked at here. So I'll update you piece by piece and maybe when it gets close to week 1 we can try putting individual components together.

    I'll work on the first piece this weekend.
     
    danmarino and Fin-O like this.
  5. Fin-O

    Fin-O Initiated Club Member

    11,375
    11,392
    113
    Sep 28, 2015
    All I can say for sure, is just about every data point is already factored into the spread itself.

    For example I can say “that altitude in week 1 is hard for a visiting team”

    That may or may not be relevant but if it is? It’s part of the reason they have arrived at the Denver -2.5 spread.

    Maybe totals are a tad more easy to manipulate mathematically. Dolphins/ Titans for example is 45.5 most books.
     
    danmarino likes this.
  6. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    That would be my default assumption too, but it's worth testing!
     
    danmarino and Fin-O like this.
  7. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    There's so much data out there I'm not really sure where you'd start. For example, the Dolphins have started slow offensively the past several years in a row....plus we've seen what the pre-season has brought. That makes me a big fan of the under there but what if my line of thinking is actually meaningless?

    At the same time, we could say that we're starting an unproven LB corp that has struggled against the run. That would want me to favor the over on that stat line. Both are good arguments but they obviously clash and give two completely different opinions.
     
    Fin-O likes this.
  8. Fin-O

    Fin-O Initiated Club Member

    11,375
    11,392
    113
    Sep 28, 2015
    Essentially Vegas has this as a 24-21 type game.

    Easily could be, easily couldn't be. This one for me anyways is a total I wouldn't touch.

    Seems inflated but Florida heat combined with a team that can run the ball as well as anyone in TEN means they could wear us down and rush for 200 yards and drop 30 on us.

    Flip side, if being uptempo WORKS for us then the same could be said for the Dolphins O vs that D.

    I actually hate the week 1 matchup in the Titans. I think we lose that game if I had a gun to my head. Ball control/run game/solid D. Yea, not real confident.
     
    Surfs Up 99 and KeyFin like this.
  9. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    Yeah, it's honestly a tough game to handicap with RT returning, the D being unproven, basically a new offense, showing nothing in pre-season, etc. We could score 6 or 50....I have absolutely no idea. I think the offense will eventually be strong this year; I'm just not expecting it early.

    As usual, I'll do a full season prediction after this week's pre-season game (I always do it after week 3). This year I have no idea though; all I can do is base my "guesses" off what RT and the offense shows this week. And we still probably won't have Gore, Parker and maybe even Stills.
     
    Irishman and Fin-O like this.
  10. Pauly

    Pauly Season Ticket Holder

    3,696
    3,743
    113
    Nov 29, 2007
    About 2 months ago I was on holidays in Peru and did a lot of hiking in the Andes.
    My experience
    1 mile high - I didn’t really notice any difference. I mean you feel the ne air is thin but it didn’t really affect walking up and down mountains as far as I noticed.
    One and a half miles high (Machu Picchu) This is where I first really noticed getting short of breath more quickly than normal, especially when climbing up the hills and stairs.
    2 miles high (Cuzco). It took me about a day to acclimitate to the altitude and even then I knew I couldn’t push myself too hard. You’d want to spend at least a week there before doing anything athletic.
    Two and a half miles high (Lake Titicaca). Oh boy, you feel it hard. On my first night I walked up two flights of stairs to my room at the hotel and had to spend 5 minutes sitting on the bed to get my breath back.
     
    danmarino, KeyFin, Irishman and 2 others like this.
  11. danmarino

    danmarino Tua is H1M! Club Member

    15,355
    20,976
    113
    Sep 4, 2014

    I never bet on or against my favorite teams. I can't trust my homerism to not rear it's ugly head.

    But yeah. If I had to bet on the Dolphins week 1 put me down for the over and I'd take the Titans -2.....But again, I'm not touching any Dolphins games. Never have and never will. (Well, except the friend $20 bet when the Dolphins are playing their favorite team)
     
  12. Fin-O

    Fin-O Initiated Club Member

    11,375
    11,392
    113
    Sep 28, 2015
    I’m the opposite.

    I think it is an advantage to know your fav team inside and out. I have bet on or against the Dolphins and Buckeyes multiple times.

    In fact I took the Gators +7.5 in that National Title game years ago vs OSU. I still thought Ohio State would win, but you could just sense the complacency. Had them winning a tight game....we all know the end result however.

    I won’t be betting on TEN week 1, too close to call for me. But if I had too????
     
  13. Dorfdad

    Dorfdad Well-Known Member

    4,052
    2,347
    113
    Dec 9, 2007
    Meh I want full contact, neck breaking hits 24/7 this is a mans game! Snowflakes step aside this isn’t college let the rookie get smacked in the mouth let em bleed on the field than they will learn! Who needs a game plan? We didn’t use one in the fields around thanksgiving he’ll we didn’t use pads and we were mostly safe!

    I joke and most everyone knows they don’t gameplan for the first two. We will have a better understanding tonight where we stand after the half. I really want to see the rookies on display this game and I’m still worried about our defense!!
     
    Irishman likes this.
  14. jdallen1222

    jdallen1222 Well-Known Member

    2,752
    1,373
    113
    May 31, 2013
    Plantation, Fl

    The game isn't til saturday night.
     
    Irishman likes this.
  15. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    In the 80's and 90's, Miami rarely ever lost at home because of the heat- that goes for the Canes and the Fins. It would often be a game in the first half and then a blowout in the 2nd half as players tired out. Of course, these we almost always 1 PM games as well. One of the dumbest things Ross did as owner was move games to 4 PM to make the fans more comfortable.

    This season, almost all our home games are 1 PM as they should be...which gives us the best possible chance of 8-0 at home. It might be a factor against the Titans and Raiders. It will probably be a factor against the Bears, Lions, Jets and Bills. Then we have the Pats Dec 9th at home- I'll always guess that one a win because of the huge temperature change for the Cheats.

    To me, that's one of the few things worth betting on in the NFL- 100 degrees with 100% humidity wears you out quickly if you're not conditioned for it. You can "get by" in cold weather as you get active and your body warms up, but players shut down in the heat because they quickly dehydrate and cramp. It's very good to see us using our 12th man again.

    The only difference today is the practice schedules and the indoor facilities- our players from out of town don't get the full benefit of the S. Florida sun like they used to. I still think it's a fairly big advantage though.
     
    Fin-O likes this.
  16. Hoops

    Hoops Well-Known Member

    1,183
    1,484
    113
    Dec 11, 2016
    You are a year early with this
     
  17. djphinfan

    djphinfan Season Ticket Holder Club Member

    111,648
    67,540
    113
    Dec 20, 2007
    Cbrad, is there any data out there that has records of teams when they lose their starting qb?, like what’s the win loss percentage..
     
    Irishman likes this.
  18. djphinfan

    djphinfan Season Ticket Holder Club Member

    111,648
    67,540
    113
    Dec 20, 2007
    It’s Mariota’s ability to improvise and make plays is what scares me..
     
    Irishman and Surfs Up 99 like this.
  19. Surfs Up 99

    Surfs Up 99 Team Flores & Team Tua

    1,950
    1,785
    113
    May 5, 2016
    Yep, I am with you. I see both defenses being gassed in the 2nd half. The Titans because of the heat and the Dolphins because the defense can't get off the field. We will probably be trailing (maybe by a lot) at the start of the 3rd Qtr. It will all come down to our offense, and I don't know if RT can take over a game and will us to victory. Of course, I hope this game and year will be different than in years past and we kick their ***, but until I see different, all I have to go on is our past performance.
     
    Dorfdad, Irishman and Fin-O like this.
  20. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Not precisely the way you phrased it (as far as I know), but a pretty good proxy is to ask what all QB's records are when they start a minimum of N games in a season vs. starting less than N games in a season. In other words, we don't care who was technically the "starter" that year, just the record of every QB when they don't start enough games etc...

    If N = 6 (i.e., we consider a QB to be a starter if they started 6 games that season.. backup if they started 5 or less), the overall win% of "starters" since 1970 is 53.67% or 8.58 games won in a 16 game season, while the overall win% of "backups" since 1970 is 40.57% or 6.5 games won in a 16 game season.

    Note that by this definition, Cutler was a starting QB last year. Still, it gives you an idea of how many extra games you expect to lose just by switching from a starter to a backup: 2 extra losses.
     
    eltos_lightfoot and Irishman like this.
  21. Fin-O

    Fin-O Initiated Club Member

    11,375
    11,392
    113
    Sep 28, 2015
    Sounds like teasing the Dolphins (+8.5) and the over (39) makes sense.
     
    Surfs Up 99 likes this.
  22. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    OK Fin-O.. here's the first piece. Actually interesting.

    The question I asked was how well does the record of a team's last N games predict whether they will win the next game. For example, if N=5 you look at win/loss record for the last 5 games. If they won more than they lost you predict they'll win the next game, and if they lost more than they won you predict they'll lose the next game. For odd N you can't have equal numbers of wins and losses, but if N is even you could. In case of equal wins and losses last N games you make no prediction and leave the data point out.

    Here are the results!
    [​IMG]
    The y-axis is the probability with which you will correctly predict the outcome of the next game, and of course the x-axis is the number of games back you are looking at.

    Main takeaway: The peak is at N=10. In other words, a team's win/loss record over the last 10 games provides the best predictive power compared to all other N.

    Note that the worst there is N=1 which suggests we should really never care about the last game's result.

    Why is that curve jagged? Because there's an inherent difference between even and odd. Even is always more powerful than odd because we're leaving cases out where the team had equal numbers of wins and losses past N games and thus the predictive power is artificially larger for even N. For example, for N=4 the worst win% for predicting a win is 3 out of 4 wins, so 75%. Compare that to N=3 and N=5 where the worst win% for predicting a win is 2 out of 3 or 66.7% for N=3 while for N=5 it's 3 out of 5 or 60%, both of which are less than 75%.

    What happens after N=100? It basically stays level out to N=200 (didn't check farther). 100 games is ~6 years back and 200 games is ~12 years back and yet you see no predictive difference! That's evidence we have a good number of teams that are consistently good or consistently bad over longer periods of time so that the graph never goes down to 50%. Given the length of tenure of good coaches (spans 1-2 decades or more), 50 years of NFL history is actually too small sample size for that graph to truly get down to 50%. Maybe 100 years might do it.

    Finally.. to your specific question of week 1. Is the predictive power less for week 1 than other weeks? Yes but not by too much. That N=10 peak for week 1 only is 58%. The average for all weeks is 59.7%, and the best weeks to predict are weeks 11 and 12 at 61.5% and 62.6% respectively (makes sense.. you're in the same season and you just had 10 games to look at). Note that by "weeks" I don't mean "game number", I literally mean "weeks" including the bye week. Thus, there are actually 18 weeks I looked at in some cases because in 1993 they had 2 byes.

    Oh, and as far as using this to predict games, note that 60% probability of winning generally corresponds to about a -150 moneyline, so this graph alone would say to bet on a team if the past N=10 games says they'll win and their moneyline is less negative than -150. Of course, these stats are only averages and we don't know more precisely when they would beat Vegas (they probably do in some specific cases because I doubt crowd sourcing is THAT optimal).


    Anyway, what I'm going to do now is to look at past N game records of both teams in a matchup. That's not as easy for me to do given the way my database is structured (games are listed by team so each game is listed twice and that just complicates stuff when you try to work backwards starting with a match instead of starting with a team), but give me a week or so and I'll have that stat.
     
    Fin-O, Irishman and KeyFin like this.
  23. Fin-O

    Fin-O Initiated Club Member

    11,375
    11,392
    113
    Sep 28, 2015

    Great stuff man!

    If I had your brain I’d count cards in blackjack, play live poker....basically live in Vegas.

    Looking foreword to more!
     
    Surfs Up 99, KeyFin and cbrad like this.
  24. djphinfan

    djphinfan Season Ticket Holder Club Member

    111,648
    67,540
    113
    Dec 20, 2007
    Why not work together on a Vegas project..
     
  25. Dorfdad

    Dorfdad Well-Known Member

    4,052
    2,347
    113
    Dec 9, 2007
    Someone mentioned Gase’s issue is that he constantly tosses screens and they are inaffective. That’s not the issue New England lives on screens as do many other top ranked teams. It’s in the execution and the players running those screens.

    I just think we have bad personal running our offense. Without a stud QB you will flounder! This has been proven over and over again. If you don’t have a stud QB you Jave to have a lights out defense to compensate. We have neither
     
    Redwine4all likes this.
  26. Disgustipate

    Disgustipate Season Ticket Holder Club Member

    31,608
    55,630
    113
    Nov 25, 2007
    I tend to think odds are what is going to happen is that the offense is pretty good and the inevitable +/- .500 season gets blamed on the defense, and Gase will get strong-armed into hiring a new defensive coordinator.
     
    Carmen Cygni likes this.
  27. Dorfdad

    Dorfdad Well-Known Member

    4,052
    2,347
    113
    Dec 9, 2007
    Should have already happened. He’s risking his coaching career keeping these project coordinators. He’s had 3 seasons to get a top rated one but he doesn’t want one that’s established or can be seen as someone who can question his authority I’m assuming. Same for the offense he needs a real coordinator and focus on the head coach role but he wants to be in control. He’s hurting this team and his coaching career if you ask me.

    He’s not going to get another three years to build this team out. He’s on the hot seat next year unless he makes the playoffs this year and odds makers are betting we won’t.
     
    Redwine4all likes this.
  28. Fin-O

    Fin-O Initiated Club Member

    11,375
    11,392
    113
    Sep 28, 2015
    If this team is still getting gashed between the tackles and can’t cover TE’s, id expect a change at DC at the bye week.
     
    Redwine4all likes this.
  29. Irishman

    Irishman Well-Known Member

    573
    532
    93
    Oct 16, 2017
    High Point, NC
    Maybe, maybe not!
     
  30. Dorfdad

    Dorfdad Well-Known Member

    4,052
    2,347
    113
    Dec 9, 2007
    Way to go out in a limb lol. You happen to be a weatherman?
     
    KeyFin likes this.
  31. Irishman

    Irishman Well-Known Member

    573
    532
    93
    Oct 16, 2017
    High Point, NC
    That depends on wither or not you buy my particular rosy view.
     
    Dorfdad likes this.
  32. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Well.. preseason is over so I might as well finish the 2nd part. So, part 1 established that we should look at N=10 games back for best predictive power. Now the question is what is the probability of the home team winning given the records of both the home and away teams over their last 10 games.

    So, I looked at 10,737 games since 1970 in the regular season where the outcome wasn't a tie and both teams had at least 10 regular season games prior to the game I was looking at (all that means is that some teams didn't exist until a certain date). For the graphs below, keep in mind we're only looking at the win probability of the home team.

    [​IMG]

    The top graph shows home team win percent (vertical axis) for every combination of win/loss records in the last 10 games. Note the axis labels: (0,0) is in the upper left so that you can see all the data, not at the bottom where you'd usually put it. The bottom graph shows you the sample sizes for each point. As you can see there are almost no cases along the edges, and finding the best-fitting plane to the data for the top graph has to take this into account.

    Technical details aside, the best-fitting plane is shown below for the data in the top graph:

    [​IMG]
    That equation in purple is the summary you want: WP = 5.4*H - 4.66*A + 54.1, where H is number of wins the home team had in their last 10 games and A is the number of wins the away team had in their last 10 games.

    For example, if H=5 and A=5 that equation tells you that the expected win probability is 57.8% for the home team. Anyway, you can play around with that equation. Note that for some extreme cases it goes above 100% because it's only a best-fitting plane, not a best-fitting surface that can curve more at the edges. Won't matter in practice.

    For betting purposes you have to convert those probabilities to moneylines. If P>50% you use 100*P/(100-P) with a negative sign in front. If P<50%, then it's 100*(100-P)/P with a positive sign in front. Or you can just look it up here:
    https://www.boydsbets.com/money-line-conversion-chart/

    When that equation predicts a probability of winning higher than the moneyline implies, then over time you should make money if you only bet on those cases (could be a LONG time though lol).
     
    Last edited: Aug 30, 2018
  33. Fin-O

    Fin-O Initiated Club Member

    11,375
    11,392
    113
    Sep 28, 2015
    I would stake you in a poker game any day of the week.
     
    cbrad likes this.
  34. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    This brings up the next obvious question- how does your calculation compare to what the Vegas odds makers use? It would be interesting to check the data against week one to see if your calculations are pretty close to theirs or if there's a decent gap in some games. I'm also wondering how much it matters from part of the "ten game results" being from the prior season- in this case, all 10. Vegas has to use something though based on last year.

    What I'm thinking here is that we chase the hypothesis and do the actual experiment. Let's place a fictitious $1 heads up bet on each of cBrad's favorites for week one ($16 total) and see where we end up. If the results are encouraging, we can pretend bet for week 2 as well and so on. I'm not personally a gambler but I'd love to see the data and figure out ways to improve the formula (w/ additional variables) even more.
     
    cbrad likes this.
  35. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Yeah I was thinking of doing that experiment myself. So right before the games begin I'll post the games that the stats in post #72 suggest one should bet on. We have to do the experiment for a long enough time though, say a full season.

    As far as the week 1 issue, we know from post #62 that the probability of successfully predicting an outcome (using only one team's record) at N=10 for week 1 is 58% while the overall average is 59.7% and the peaks at weeks 11 and 12 are at 61.5%. So week 1 is definitely harder. But the stats in post #72 combine everything so over time they should work.
     
    Fin-O and KeyFin like this.
  36. Redwine4all

    Redwine4all Well-Known Member

    1,216
    686
    113
    Feb 4, 2016
    I've been calling for Burke's jobs since last year. I don't think he inspires improved performance from players.
     
  37. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    OK.. time to put historical odds to the test of predicting future wins. Since this is mostly exploratory research and all done with fictitious money, I'm just going to make some arbitrary rules that should be useful for research purposes.

    Rules: I'll make a fictitious bet on every game where two conditions are satisfied: 1) the Vegas odds of winning are at least 5% worse than predicted in post #72, and 2) the Vegas moneyline is between +200 to -200. The reason for these conditions is because otherwise it will take too long to see how this method works (events are too unlikely). The amount of money bet should ideally be dependent on the probability of the event, which we know from that "sample sizes" graph in post #72, but since this just exploratory in nature I'll use a simple proxy: every bet aims to win $100.

    I'm going to use the current moneylines (at time of writing this post) at sportsbook.ag where I bet sometimes:
    https://www.sportsbook.ag/sbk/sportsbook4/nfl-betting/week-1-lines-nfl-game-lines.sbk

    Here are the games a purely statistical approach (looking at past 10 game records for both teams) says we should bet on, assuming the 2 conditions above are satisfied:

    Bet $115 on the Eagles (-115)
    Bet $200 on the Steelers (-200)
    Bet $80 on the Bengals (+125)
    Bet $165 on the Jaguars (-165)
    Bet $120 on the Titans (-120)
    Bet $77 on the Seahawks (+130)
    Bet $115 on the Cardinals (-115)
    Bet $57 on the Raiders (+177)

    To explain how you arrive at these choices, consider the Eagles bet. Eagles are 8-2 in their past 10 games while Falcons are 7-3. With the Eagles as the home team the equation in post #72 says their probability of winning is 64.68%. The current moneyline for the Eagles is -115, which means that if you bet $115 you win $100. The implied probability of winning for -115 moneyline is 53.5% which is more than 5% below 64.68%. Since -115 is within the -200 to +200 interval, this method says to bet $115 on it.

    It's interesting to see where historical odds are similar to Vegas odds and where they differ most.

    They differ most when the QB was different the year before. For example, the Colts are favored -145 (implied win probability of 59.1%) because Luck is back while historical odds say Bengals should win with 58.4% probability. Vegas gives the Dolphins even odds while historically it's 36.9% because of course we had that dud Cutler last year. And Case Keenum in Denver makes Denver a favorite for Vegas (-150 with implied win probability of 60%) while historically it's only 41.6%. So one interesting thing to note is whether this method tends to fail in these cases. I had to bet on all three obviously!#@!.. sorry about betting against the Dolphins, but I have to test this method.

    And for info, the closest to Vegas odds were the Bucs whose +360 moneyline implies 21.7% chance of winning while historical odds say 22.1%. Others that were close: Vikings (-250) implies 71.4% while historically it's 74.7%, Panthers (-150) implies 60% while historically it's 63.9%, Chargers (-180) implies 64.3% while historically it's 68.6%, and Lions (-280) implies 73.7% while historically it's 77.2%.

    OK, we'll see what happens!
     
    Irishman, KeyFin and Fin-O like this.
  38. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    I'm really looking forward to this! On a side note, I'm also looking forward to seeing how Cleveland does since that's obviously a team you'd never bet on (with 0-10 the last 10 games). I think they'll be decent this season with their defense...I actually thought they have a shot week 1 vs Pittsburgh.
     
    Irishman likes this.
  39. Carmen Cygni

    Carmen Cygni Well-Known Member

    2,422
    5,732
    113
    Dec 30, 2017
    It's the timing of the playcall that is most important in these cases b/c screens, draws, play-action, etc. are constraint plays that only work when a defense is leaving it's fundamentals in an attempt to exploit another area. EX: Defensive ends are continually driving hard into the backfield so you call a counter draw off the edge or dump a HB or TE screen. Or the CBs are beginning to gain more depth in their setup due to a multiple play drive, causing fatigue, so you call for a smoke or bubble to take advantage of the spacing.

    Tannehill is fine. He's a damn good QB. People need to stop trying to miraculously get a hold of the next winning lottery ticket (aka elite QB) and focus on other units of the team to round it out.
     
  40. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    I'm curious- what was your pre-draft impression of Falk? I know next to nothing about scouting QB's so I'm excited just on him being mentioned as Gase's favorite last year.
     

Share This Page