Trying to imitate PLindsey

[biju]

Ok guys, for those of you who have been around this neck of the woods a long time you'll undoubtedly remember that PLindsey88 was the DST guru. For a few years he showed us magic by using data to explain which DSTs (however unlikely) would dominate in the upcoming week and being very successful in doing so.

This is not a repeat of that.

This is my very first attempt with the numbers. I'm not hiding in the shadows testing my method out and waiting until it looks good. I'm putting it out there very raw. The reason for doing so is so folks can challenge how the numbers fall--if they believe these rankings look accurate or not. I'll give my explanation of why the numbers are saying what they are saying to back up the numbers, but much like PLindsey I'm going to keep the method hidden. If I happen to strike gold, then the Cafe strikes Gold as well and that's good for everyone here. Normally these would also go into one of the "Routine Plays" sub-forums, but for traffic and the need for feedback I decided to post it here. Mods, please move it if you feel this is in the wrong place.

And with that, the numbers...

Code: Select all

Pittsburgh Steelers      7.43
Green Bay Packers        5.68
Seattle Seahawks         4.71
San Diego Chargers       3.83
Chicago Bears            2.02
Tampa Bay Buccaneers     1.57
Dallas Cowboys           1.39
Tennessee Titans         0.41
New England Patriots    -0.11
Atlanta Falcons         -0.76
Detroit Lions           -0.78
St. Louis Rams          -0.78
New York Jets           -1.65
Kansas City Chiefs      -2.07
New York Giants         -2.24
Oakland Raiders         -2.39
Minnesota Vikings       -3.27
New Orleans Saints      -3.51
Cleveland Browns        -4.54
Philadelphia Eagles     -4.86
Miami Dolphins          -5.41
San Francisco 49ers     -6.02
Indianapolis Colts      -6.41
Baltimore Ravens        -9.19
Washington Redskins     -9.60
Jacksonville Jaguars   -10.72
Houston Texans         -11.46
Denver Broncos         -13.75


The first thing veterans will note is that my method doesn't go into the 1,000s. This system will stay roughly in the +10 to -20 range with the higher score being the better defense to start. For reference I'm hoping to also put down the "ideal" and "worst" possible defensive matchups each week as a baseline for people to know the absolute scale. This week's best *possible* defensive matchup would be the Steelers (+9.07) vs. the Vikings. This seems counter-intuitive from the perceptions the Vikings still get, but on paper they look like an error prone offense that doesn't score. This obviously doesn't take into effect the Moss trade, but I'm not sure how much that will help short term to fix this. The worst defensive matchup is interesting as well: the Texans (-15.56) vs. the Jets. According to the numbers the apparently would get crushed defensively. With the maximum range established of 9.07 to -15.56 you can see how the defensive matchups look.

The most attractive of the matchups based on which defenses are likely available free agents is the Seattle Seahawks (no, I didn't backfill the data so the Seahawks would be great in the first week of trying this ). It should be warned however that the Seahawks tend to perform very poorly on the road, but at the same time the Bears offense has a high tendency for INTs and sacks.

At any rate, this is a rough draft at best. Please note I'm not a statistician, just someone who likes numbers a bit and has been wanting to have some sort of replacement for PLindsey's weekly numbers for a few years now. Enjoy.

[m16a]

Great work my friend. I do miss Plindsey's stuff. He was the best at it. Big shoes to fill, but if you can do half as well, you'll be a great asset to the cafe. Thanks for the hard work.

[Azrael]

Good job on this man.

[petedog9]

"This week's best *possible* defensive matchup would be the Steelers (+9.07) vs. the Vikings"

Am I reading this right? The Steelers play the Browns this week @ home and the Vikes get the Cowboys @ home....

[biju]

"This week's best *possible* defensive matchup would be the Steelers (+9.07) vs. the Vikings"

Am I reading this right? The Steelers play the Browns this week @ home and the Vikes get the Cowboys @ home....

Keep in mind that this is the best possible defensive matchup based on how they fare on paper, not by the schedule. I wanted to know what the best and worst matchups possible were as a baseline to see what the maximum values could be. Obviously these best and worst matchups aren't available this week, but showing that the best possible score is 9.07 you get a sense of how good the best *real* matchup (PIT/CLE) is. Without this information, it might be easy to assume that the top matchup is great, when in reality it is average at best. In this particular case, the best possible matchup of 9.07 isn't too far off the Steelers actual DST score of 7.43, meaning it's an extremely good matchup.

Does that make sense? I knew I should have spent a little more time on that paragraph. Maybe for next week I'll normalize the numbers so it's on a graded scale and we'll have absolute top and bottom ends. That way when (and it will happen) the top defense is actually a very mediocre matchup it will be rated in the 70s instead of just being on top of the list.

Let me know if you have other ideas on how to make it more readable. It was basically the only thing I didn't like about PLindsey's system so I'm trying to fix that piece of it.

[dream_017]

Nice job Biju

I will be keeping an eye on this. For the past 2 years we have had a couple of guys step up and create a defense thread that I stickied in the help forums. I hope to see those guys back at it, but the more the better. If you feel pretty comfortable with these, I can make sure that they get a sticky for this week - while you still get the discussion and tweaks worked out here

[biju]

Let's skip the Steelers/Vikings/Texans/Jets thing for a moment and I'll show this a little more eloquently. Here's the raw numbers plus their normalized (to a scale of 0-100 with 100 being the best matchup possible) rating:

Code: Select all

Pittsburgh Steelers      7.43 (93.34)
Green Bay Packers        5.68 (86.23)
Seattle Seahawks         4.71 (82.30)
San Diego Chargers       3.83 (78.72)
Chicago Bears            2.02 (71.37)
Tampa Bay Buccaneers     1.57 (69.55)
Dallas Cowboys           1.39 (68.82)
Tennessee Titans         0.41 (64.84)
New England Patriots    -0.11 (62.73)
Atlanta Falcons         -0.76 (60.09)
Detroit Lions           -0.78 (60.01)
St. Louis Rams          -0.78 (60.01)
New York Jets           -1.65 (56.47)
Kansas City Chiefs      -2.07 (54.77)
New York Giants         -2.24 (54.08)
Oakland Raiders         -2.39 (53.47)
Minnesota Vikings       -3.27 (49.90)
New Orleans Saints      -3.51 (48.92)
Cleveland Browns        -4.54 (44.74)
Philadelphia Eagles     -4.86 (43.44)
Miami Dolphins          -5.41 (41.21)
San Francisco 49ers     -6.02 (38.73)
Indianapolis Colts      -6.41 (37.15)
Baltimore Ravens        -9.19 (25.86)
Washington Redskins     -9.60 (24.20)
Jacksonville Jaguars   -10.72 (19.65)
Houston Texans         -11.46 (16.65)
Denver Broncos         -13.75 (7.35)


[biju]

Nice job Biju

I will be keeping an eye on this. For the past 2 years we have had a couple of guys step up and create a defense thread that I stickied in the help forums. I hope to see those guys back at it, but the more the better. If you feel pretty comfortable with these, I can make sure that they get a sticky for this week - while you still get the discussion and tweaks worked out here

Thanks. This isn't supposed to be a replacement for anyone else's hard efforts by the way. I happen to be in a unique position however because in discussions directly with PLindsey he did let me in on a very old version of his formula. I've tweaked it, but it needs plenty more tweaking.

For the record, my confidence is low. Primarily for two reasons: (1) I really haven't tried this before in previous weeks--this was created yesterday afternoon while I was bored waiting for a doctor's appointment; (2) As PLindsey himself would state, the numbers get more accurate as the season progresses because there is more data to help deflect anomalies in particular games (a good example being the Seahawks DST is helped greatly by the two KRTDs against the Chargers in week 3).

But I'm posting them primarily because I wanted to plant a stake in the ground and let the community help with any suggestions or comments. I've already received one crucial piece of information in doing this so the numbers will likely change for the better next week (I probably won't have time to do this before this weekend).

Here's hoping we can make something great.

[dream_017]

Nice job Biju

I will be keeping an eye on this. For the past 2 years we have had a couple of guys step up and create a defense thread that I stickied in the help forums. I hope to see those guys back at it, but the more the better. If you feel pretty comfortable with these, I can make sure that they get a sticky for this week - while you still get the discussion and tweaks worked out here

Thanks. This isn't supposed to be a replacement for anyone else's hard efforts by the way. I happen to be in a unique position however because in discussions directly with PLindsey he did let me in on a very old version of his formula. I've tweaked it, but it needs plenty more tweaking.

For the record, my confidence is low. Primarily for two reasons: (1) I really haven't tried this before in previous weeks--this was created yesterday afternoon while I was bored waiting for a doctor's appointment; (2) As PLindsey himself would state, the numbers get more accurate as the season progresses because there is more data to help deflect anomalies in particular games (a good example being the Seahawks DST is helped greatly by the two KRTDs against the Chargers in week 3).

But I'm posting them primarily because I wanted to plant a stake in the ground and let the community help with any suggestions or comments. I've already received one crucial piece of information in doing this so the numbers will likely change for the better next week (I probably won't have time to do this before this weekend).

Here's hoping we can make something great.

Sounds good. I will leave here for discussion this week. I'll add into the help forums next week. I have no problem adding them in with the other guys as we had one then another stepped up. Both are different, but get to the same results...trying to help the Cafe folks pick a good defense.

I too was a part of the PLindsey data sharing before he left. I haven't had the time to work with it(not sure at this point where I even put it). It was and still is a great tool to have here at the Cafe, so anyone that can put something together (and won't get bent out of shape if it fails a week or two) is an asset in my book and should be posted for all to see.

[petedog9]

"This week's best *possible* defensive matchup would be the Steelers (+9.07) vs. the Vikings"

Am I reading this right? The Steelers play the Browns this week @ home and the Vikes get the Cowboys @ home....

Keep in mind that this is the best possible defensive matchup based on how they fare on paper, not by the schedule. I wanted to know what the best and worst matchups possible were as a baseline to see what the maximum values could be. Obviously these best and worst matchups aren't available this week, but showing that the best possible score is 9.07 you get a sense of how good the best *real* matchup (PIT/CLE) is. Without this information, it might be easy to assume that the top matchup is great, when in reality it is average at best. In this particular case, the best possible matchup of 9.07 isn't too far off the Steelers actual DST score of 7.43, meaning it's an extremely good matchup.

Does that make sense? I knew I should have spent a little more time on that paragraph. Maybe for next week I'll normalize the numbers so it's on a graded scale and we'll have absolute top and bottom ends. That way when (and it will happen) the top defense is actually a very mediocre matchup it will be rated in the 70s instead of just being on top of the list.

Let me know if you have other ideas on how to make it more readable. It was basically the only thing I didn't like about PLindsey's system so I'm trying to fix that piece of it.

Ok .. I took it as you said the Steelers were playing the Vikings.. Godd job biju.. Thanks for making the effort...