All posts by derny

Game Charters Required!

We are always looking for talented and committed people that know football and love to watch football each and every week of the season.

What do you get in return? Well, I’m glad you asked!

As a valued game-charter, you will receive free access to all the data we provide, both in CSV format and via our comprehensive API. You will also have instant access to the charted data as it is entered on a weekly basis during the season, weeks before it is repackaged for the general public.

You will be responsible for the same team(s) from the start, allowing you to gain an intimate knowledge of every aspect of their offense and defense. This is going to be important as the program evolves! We do not chart pre-season games; however, the 11 playoff games are charted.

Please reach out to us via the contact form for more details on what we chart and how the program works.

 

Twitter Feed For All Active Players Now Available!

We’ve spent the last month working tirelessly on a new and exciting feature: Twitter content for every single player currently on each team’s 90-man roster.

As far as we know, this is the only accurate record of Twitter handles and Tweets for every active NFL player.

This data will be updated weekly during the season and is also currently available via our API. Here is a quick example of how to retrieve Aaron Rodgers recent tweets:

http://armchairanalysis.com/api/1.0/test/tweets/AR-1300

Stay tuned for yet more new content coming in the next few weeks!

 

 

 

 

Roster Updates for 2017 Starting Soon!

Welcome back NFL Fans! We are kicking off the 2017 season starting next week with the commencement of weekly roster updates.

Data on every single player currently on 90-man rosters will be available, including:

  • Combine results
  • Hand/Arm size
  • Depth Chart position

And much, much more. Data will be accessible both through the website (CSV files) and our API.

Also, we are excited to announce the inclusion of the following data in September which will help give you the full picture on QB actions and events on each play. Some of the new stats include:

  • QB Pressure – # players blitzing on each play
  • QB Flushes
  • QB Hits
  • QB Throw-aways

As well as some new data on defensive formations.

Stay tuned for more news on a few other interesting projects we are working on this off-season!

Game Charted Data is Back!

Good news NFL fans: we’ve expanded our game charting program to include a total of 16 professional charters. As a result, look for more timely updates for this specialized information this coming season.

We are currently focusing on the following data points for 2016:

  • Number of Running Backs and Tight Ends on the field (aka Personnel Groupings)
  • Yards-after-catch
  • Dropped Passes
  • Missed or broken tackles

This data will be ‘official’ for 2016 after our initial pilot project went well in 2015.

The program will be expanding to include a total of 32 game-charters in 2017. Please drop us a line via the contact form if you are interested in getting involved!

New NFL Player Stats For 2016!

There is lot’s of new stuff coming your way this off-season, starting with the addition of Arm Length and Hand Size data in the PLAYERS table. This data is now available for over 1,000 current and former players and this number is going to grow to over 2,000 in the next couple of weeks.

The PLAYERS table has also been updated to include over 700 new players that can be found on the current 90 man rosters. Look for daily updates to the player data from now until the start of the season (Hardcore and Hardcore-Lite only).

Also, many of you have been asking for a ‘Schedule’ table in order to forecast results for upcoming games in 2016. This new table is going to be available in the next week – I will send out an email when it’s ready.

NFL Data Expansion: Injuries

Spring has sprung and with it comes a new data table that has been at the top of the ‘most requested new feature’ list for some time: Injuries.

We’ve gone back and recorded the 5,000+ entries from official reports in 2015 and we will be updating this data weekly in 2016 onwards.

That’s not all that’s new though. Back by popular demand: a detailed change-log (available online) that will serve to keep you fully informed on the latest expansions; updates and corrections to historical and 2016 data.

Here is a quick summary of the latest changes:

CHART: 26,058 plays have now been charted for 2015
GAME: Corrected a games ‘wdir’ and ‘wspd’ values from the 2015 season
INJURY [new]: Added new ‘injury’ table with complete data from 2015
PLAYER: Corrected draft position (dpos) values for 2 players
PLAYER: Changed player IM-0300 to have a primary position (pos1) of ‘TE”

All support files have also been updated to reflect the new Injury table. Also: 2 field types were changed in the Setup.sql file for ‘charts’. They were QBA and INCD which should be VARCHAR(8).

Stay tuned for more exciting news this Spring.

Dennis Erny
Founder
Armchair Analysis.com

2015 Season Data now Available!

You now officially have 697,176 reasons to purchase a Casual data plan: we’ve just updated our historical data set to include the 2015 season.

Much of this season was spent making our virtual statistician that much more accurate and we are pleased to announce that every single play over the past 16 years has now officially been documented.

Please also say a warm hello to the new ‘Touchdowns’ table. There have been over 20,000 of them in the past 16 years and they await your analysis.

Until next time…

Historical NFL Database (2000-2014) Overhaul

Further proof that good things come in large packages too

We’ve been pretty busy combing through 651,312 NFL plays and the results of this work was uploaded last night. Some of the major aspects of this update to our Historical NFL database include:

  • Inclusion of a couple of hundred missing shared tackles.
  • Some small errors affecting yardage listed for kneel downs and spiked balls have been fixed.
  • Instances of extra yardage gained after an “own recovery” fumble not being added have been corrected.
  • Errors on roughly 150 plays where a penalty on the offense adjusted the net rushing or passing yards, and, we did not take this into consideration, have been updated.

These errors where caught mostly thru automated processes. This stage of our analysis is largely complete, which means, the next step is a lengthy manual review of all 651,000 plays to remove the last inconsistencies.

Our goal is to have data 100% aligned with official records by the Spring of 2016. The 2015 season is already achieving 100% accuracy thanks, in part, to input from many of you.

For those of you who are interested, we are offering $0.25 US per-play for any details on an error, either with the key details of the play description or any of the fields that are derived from the description (i.e, yardage gained; passer id; pass target id; tacklers; punt/koff returners) and so on. Plays that are completely missing from our records (there are roughly 100) are also eligible. Plays that contain any form of a lateral are not.

There is a limit of $0.25US payout on each play so an error in the description that perhaps caused errors in 2 different fields still nets you $0.25US.

If $0.25US per play doesn’t sound like much — keep in mind that there are still an estimated 2,500 – 5,000 issues, from incorrect clock times to incorrectly labelled players. There is money to be made here, especially for those of you who can come up with a way to automate certain checks against official league records.

The best current official reference is the individual play-by-plays for each game, available at NFL.com (season 2001 – 2015). These play-by-play listings appear to include stat corrections (changes the NFL scorers make well after the game) which do account for some of the existing errors.

If you are interested in getting involved with this latest exciting project, drop us an email!!

Behind the Numbers: Building a Virtual NFL Statistician

A project 10+ years in the making

Without a doubt, one of the more common questions we field here is: “Where do you get your data?”

It’s a valid question on a number of fronts. The first comes from a legal perspective. Are we entitled to distribute this data? Does the NFL not hold the rights to data from it’s games? Beyond the legal rights issue, some people are simply wondering how accurate the data is in comparison with official league records.

The best place to start is with the following:

pbp_csv

This, my friends, is what the sporting-world refers to as a “play-by-play”. It doesn’t just appear out of thin air. League statisticians from every NFL stadium are tasked with watching plays as they unfold and marking down the details of each and every one. This information then makes it’s way from individual stadiums and within seconds of play completion, can be found on dozens of sites worldwide.

This small snippet of information is all we need to perform our magic. Nothing more, nothing less. A game might feature 160 such snippets. A week might total something in the neighbourhood of 2,500. Numbers aside, it’s what we do with these freely-available snippets of text that really set’s our site apart from the other unofficial sources of NFL data.

Relying on nothing more than play-by-play text means we don’t need to illegally scrape sites for data or rely on any other service to provide us with what we need. While sports leagues have tried to copyright basic written accounts of sporting events, they haven’t had the best of luck. Reporting on the number of times Brett Favre passed to Antonio Freeman in his career based on game accounts is no more of an infringement on copyright than reporting on the number of times a character spoke Harry Potter’s name in a J.K. Rowling book (16,612 if you are interested).

The real question is, how do we get from this:

pbp_csv

To this:

{
  "data": {
    "gid": 3990,
    "pid": 651831,
    "off": "PIT",
    "def": "NE",
    "type": "RUSH",
    "dseq": 1,
    "len": 34,
    "qtr": 1,
    "min": 7,
    "sec": 41,
    "ptso": 0,
    "ptsd": 0,
    "timo": 3,
    "timd": 3,
    "dwn": 1,
    "ytg": 10,
    "yfog": 7,
    "zone": 1,
    "fd": 0,
    "sg": 0,
    "nh": 0,
    "pts": 0,
    "tck": [
      {
        "uid": 430825,
        "tck": "GG-0475",
        "value": "0.5"
      },
      {
        "uid": 430826,
        "tck": "AB-2700",
        "value": "0.5"
      }
    ],
    "sk": 0,
    "pen": 0,
    "ints": 0,
    "fum": 0,
    "saf": 0,
    "blk": 0,
    "olid": {
      "lt": "KB-0750",
      "lg": "AV-0350",
      "c": "RF-0900",
      "rg": "CH-4550",
      "rt": "CW-0400"
    },
    "rush": {
      "bc": "DW-3600",
      "dir": "LG",
      "yds": 6,
      "succ": 1,
      "kne": 0
    }
  }
}

And even this:

2016 NFL Database Field Listing

Going from point A (play-by-play text) to points B, C, D and beyond required us to do 2 things:

  1. Build a ‘virtual’ NFL statistician able to digest those simple play-by-play accounts to create player and team statistics and everything else in-between.
  2. Hire our own real live ‘Statisticians’ to analyze game tape to fill in the holes and also expand on the basic play-by-play data.

The first task has taken over a decade (of off and on work) and probably in the neighborhood of 15,000 lines of code to achieve. This is not an exaggeration. Here is what’s required to properly parse the effect of Fumbles:

Sub FUMBLES()

If S1.Cells(ROW1, 6) = "PASS" Or S1.Cells(ROW1, 6) = "RUSH" Then
   X8 = InStr(X7, BOOK.Cells(X2, 2), "FUMBLES") + 8: X11 = 2
   If S1.Cells(ROW1, 26) <> "" Then S1.Cells(ROW1, 61) = S1.Cells(ROW1, 26)
   If S1.Cells(ROW1, 35) <> "" And InStr(X7, BOOK.Cells(X2, 2), "FUMBLES") > InStr(X7, BOOK.Cells(X2, 2), "pass") Then S1.Cells(ROW1, 61) = S1.Cells(ROW1, 35)
   If InStr(X7, BOOK.Cells(X2, 2), "FUMBLES (Aborted)") > 0 Then
      For X3 = 1 To 6
      If QBNAME(X3, X6 + 1) = "" Then Exit For
      If InStr(Mid(BOOK.Cells(X2, 2), X7, 40), QBNAME(X3, X6 + 1)) > 0 Then S1.Cells(ROW1, 61) = "QB " & QBNAME(X3, X6 + 1): Exit For
      Next X3
      If S1.Cells(ROW1, 61) = "" Then
         For X3 = 1 To 75
         If ALLNAME(X3, X6 + 1) = "" Then Exit For
         If InStr(Mid(BOOK.Cells(X2, 2), X7, 40), ALLNAME(X3, X6 + 1)) > 0 Then S1.Cells(ROW1, 61) = ALLPOS(X3, X6 + 1) & " " & ALLNAME(X3, X6 + 1): Exit For
         Next X3
      End If
   End If
   If InStr(X7, BOOK.Cells(X2, 2), "Aborted") > 0 And InStr(X7, BOOK.Cells(X2, 2), "FUMBLES (Aborted)") = 0 Then
      For X3 = 1 To 75
      If ALLNAME(X3, X6 + 1) = "" Then Exit For
      If InStr(Mid(BOOK.Cells(X2, 2), InStr(X7, BOOK.Cells(X2, 2), "Aborted") + 8, 25), ALLNAME(X3, X6 + 1)) > 0 Then S1.Cells(ROW1, 61) = ALLPOS(X3, X6 + 1) & " " & ALLNAME(X3, X6 + 1): Exit For
      Next X3
   End If
   If InStr(X7, BOOK.Cells(X2, 2), "Lateral to") > 0 And InStr(X7, BOOK.Cells(X2, 2), "FUMBLES") > InStr(X7, BOOK.Cells(X2, 2), "Lateral to") Then
      For X3 = 1 To 75
      If ALLNAME(X3, X6 + 1) = "" Then Exit For
      If InStr(Mid(BOOK.Cells(X2, 2), InStr(X7, BOOK.Cells(X2, 2), "Lateral to") + 11, 25), ALLNAME(X3, X6 + 1)) > 0 Then S1.Cells(ROW1, 61) = ALLPOS(X3, X6 + 1) & " " & ALLNAME(X3, X6 + 1): Exit For
      Next X3
   End If
   If S1.Cells(ROW1, 61) = "" Then S1.Cells(ROW1, 61) = S1.Cells(ROW1, 31)
End If

If S1.Cells(ROW1, 6) = "KOFF" Or S1.Cells(ROW1, 6) = "PUNT" Or S1.Cells(ROW1, 6) = "ONSD" Then
   If S1.Cells(ROW1, 6) = "PUNT" Then X11 = 2 Else X11 = 1
   If InStr(X7, BOOK.Cells(X2, X11), "MUFFS") > 0 Then
      X8 = InStr(X7, BOOK.Cells(X2, X11), "MUFFS") + 6
   Else
      X8 = InStr(X7, BOOK.Cells(X2, X11), "FUMBLES") + 8
   End If
   If InStr(X7, BOOK.Cells(X2, X11), "FUMBLES (Aborted)") > 0 Then
      If S1.Cells(ROW1, 72) <> "" Then S1.Cells(ROW1, 61) = S1.Cells(ROW1, 72)
      If S1.Cells(ROW1, 79) <> "" Then S1.Cells(ROW1, 61) = S1.Cells(ROW1, 79)
   End If
   If InStr(X7, BOOK.Cells(X2, X11), "Aborted") > 0 And InStr(X7, BOOK.Cells(X2, X11), "FUMBLES (Aborted)") = 0 Then
      If S1.Cells(ROW1, 6) = "PUNT" Then X10 = X6 + 1 Else X10 = 2 - X6
      For X3 = 1 To 75
      If ALLNAME(X3, X10) = "" Then Exit For
      If InStr(Mid(BOOK.Cells(X2, X11), InStr(X7, BOOK.Cells(X2, X11), "Aborted") + 8, 25), ALLNAME(X3, X10)) > 0 Then S1.Cells(ROW1, 61) = ALLPOS(X3, X10) & " " & ALLNAME(X3, X10): Exit For
      Next X3
   End If
   If InStr(X7, BOOK.Cells(X2, X11), "Lateral to") > 0 And InStr(X7, BOOK.Cells(X2, X11), "FUMBLES") > InStr(X7, BOOK.Cells(X2, X11), "Lateral to") Then
      If S1.Cells(ROW1, 6) = "PUNT" Then X10 = 2 - X6 Else X10 = X6 + 1
      For X3 = 1 To 75
      If ALLNAME(X3, X10) = "" Then Exit For
      If InStr(Mid(BOOK.Cells(X2, X11), InStr(X7, BOOK.Cells(X2, X11), "Lateral to") + 11, 25), ALLNAME(X3, X10)) > 0 Then S1.Cells(ROW1, 61) = ALLPOS(X3, X10) & " " & ALLNAME(X3, X10): Exit For
      Next X3
   End If
   If InStr(X7, BOOK.Cells(X2, X11), "punt is BLOCKED") > 0 And InStr(X7, BOOK.Cells(X2, X11), "FUMBLES") > InStr(X7, BOOK.Cells(X2, X11), "punt is BLOCKED") Then S1.Cells(ROW1, 61) = S1.Cells(ROW1, 67)
   If S1.Cells(ROW1, 61) = "" Then
      If S1.Cells(ROW1, 6) = "PUNT" Then S1.Cells(ROW1, 61) = S1.Cells(ROW1, 76) Else S1.Cells(ROW1, 61) = S1.Cells(ROW1, 83)
   End If
End If

If InStr(X8, BOOK.Cells(X2, X11), "recovers at") > 0 Then S1.Cells(ROW1, 62) = S1.Cells(ROW1, 61)
If InStr(X8, BOOK.Cells(X2, X11), "recovered by") > 0 Then
   If S1.Cells(ROW1, 6) = "PUNT" Then X10 = 2 - X6 Else X10 = X6 + 1
   For X3 = 1 To 75
   If ALLNAME(X3, X10) = "" Then Exit For
   If InStr(Mid(BOOK.Cells(X2, X11), InStr(X8, BOOK.Cells(X2, X11), "recovered by") + 13, 25), ALLNAME(X3, X10)) > 0 Then S1.Cells(ROW1, 62) = ALLPOS(X3, X10) & " " & ALLNAME(X3, X10): Exit For
   Next X3
End If

If InStr(X8, BOOK.Cells(X2, X11), "RECOVERED by") > 0 Then
   If S1.Cells(ROW1, 6) = "PUNT" Then X10 = X6 + 1 Else X10 = 2 - X6
   For X3 = 1 To 75
   If ALLNAME(X3, X10) = "" Then Exit For
   If InStr(Mid(BOOK.Cells(X2, X11), InStr(X8, BOOK.Cells(X2, X11), "RECOVERED by") + 13, 25), ALLNAME(X3, X10)) > 0 Then S1.Cells(ROW1, 62) = ALLPOS(X3, X10) & " " & ALLNAME(X3, X10): Exit For
   Next X3
   If S1.Cells(ROW1, 62) <> "" Then
      X9 = InStr(X7, BOOK.Cells(X2, X11), "RECOVERED by") + 13
      If InStr(Mid(BOOK.Cells(X2, X11), X9, 90), "for no gain") + InStr(X9, BOOK.Cells(X2, X11), "Touchback") > 0 Or InStr(Mid(BOOK.Cells(X2, X11), X9, 90), " for ") = 0 Then S1.Cells(ROW1, 63) = 0
      If S1.Cells(ROW1, 63) = "" Then
         If IsNumeric(Mid(BOOK.Cells(X2, X11), InStr(X9, BOOK.Cells(X2, X11), " for ") + 7, 1)) = True Then S1.Cells(ROW1, 63) = Mid(BOOK.Cells(X2, X11), InStr(X9, BOOK.Cells(X2, X11), " for ") + 5, 3) Else S1.Cells(ROW1, 63) = Mid(BOOK.Cells(X2, X11), InStr(X9, BOOK.Cells(X2, X11), " for ") + 5, 2)
      End If
   End If
End If

End Sub

Building a customized ‘rules engine’ that produces player and team data with 100% accuracy means accounting for every possible event on a Rush; Pass; Punt; Kickoff; Field Goal; Extra Point or 2pt Conversion — no easy task.

The news is good however: we are now close to 99.9% accuracy on plays from 2000 – 2014 and can say with confidence that we are at 100% for 2015. By the end of the 2015 season, we have plans to be able to mirror official league records with 100% accuracy from 2001 onwards.

You might wonder though: why on Earth would we take the time to build a complex system that can construct player and team statistics from nothing more than play-by-play text?

Firstly, we feel there should be no free lunch. Stealing significant portions of player and team stats from other sites is not something we condone and it’s not a viable business model long-term. The raw play-by-plays we work with would be largely useless to businesses and fantasy football players that need highly organized data in a relational database and this is where our “Sweat of the Brow” comes in.

Secondly, diving down to the play level gives us the flexibility to analyze the data in thousands, if not hundreds of thousands of ways. It’s a level of detail you will simply not find anywhere else.

So, next time you come across a site pushing unofficial data, be sure to ask the important question, “Where do you get your data”.

Not Your Dad’s NFL Stats

If you want to build something different it starts with innovative data

We don’t scrape and regurgitate readily available NFL data because that’s just plain boring. Our goal at ArmchairAnalysis.com has always been to provide cutting-edge NFL stats that can be used to build innovative NFL-related applications that push the boundaries of grid-iron analysis.

Take our breakdown of NFL penalties as one example.

Here’s a fun little fact: Between the 2000 and 2013 there were roughly 50,064 penalties called in the NFL with a total of over 352,500 penalty yards being assessed in the process.

That’s no small number and it adds up to a huge impact on hundreds of game outcomes.

All penalties are not created equal; however, and our goal is to provide you with the metrics that will help you determine what does, and doesn’t, matter when it comes to game outcomes; fantasy point totals — you name it.

Over the past few seasons, we’ve come up with the following 7 categories that account for 99.6% of all flags thrown in the NFL.

#1 False Start Penalties (Offense)

This is one of the most common of all penalty calls and deserves it’s very own category

#2 Offensive Holding Penalties

This category consists mostly of holding calls on the Offensive Line.

Most common flags: Offensive Holding, Offensive Pass Interference & Illegal Use of Hands.

#3 Playbook Execution Penalties

This is my favourite category. These tend to occur more often at the start of the season, and, against teams that try and skimp on headsets for the coaching staff (had you going there, right?).

Most common flags: Illegal Formations, Passes, Substitutions, Procedures and Motion; Intentional Grounding, 12 On Field, Delay of Game.

#4 Defensive Line Penalties

It’s all about the timing. And, not drinking excessive amounts of caffeine too close to game time.

Most common flags: Defensive Offside, Encroachment, Neutral Zone Infractions.

#5 Defensive Secondary Penalties

Cornerbacks and Safeties have a tough job, but, some just do it better than others.

Most common flags: Defensive Holding, Defensive Pass Interference, Illegal Contact & Use of Hands.

#6 Dumb Penalties

Teams that are overly aggressive on defense (I’m looking at you Ravens) are more likely to draw these types of flags. Emotion aside, some of these really are just plain stupid.

Most common flags: Taunting, Roughing the Passer/Kicker, Face Masks, Unsportsmanlike Conduct.

#7 Poor Fundamentals

This is a category of penalties where the quality of a team’s defensive coach and staff can be measured. Some like tripping, border on dumb. But, if you need to trip someone in the first place it’s usually because you’re not doing your job very well.

Most common flags: Illegal Blocks, Crackbacks, Tripping, Horsecollars, Clipping.

Here is our list shown in a handy table, revealing the actual number of penalties for each category and the average yards-per-call.

Category #Flags Yards Avg
False Starts 9,363 45,890 4.90
Offensive Holding 10,842 85,514 7.89
Playbook Execution 6,652 28,477 4.28
Defensive Line 5,662 22,126 3.91
Defensive Secondary 7,506 61,916 8.25
Dumb/Aggressive 7,343 85,067 11.58
Poor Fundamentals 2,493 22,442 9.00

This is just one example of our quest to innovate and present NFL data that is just plain fun to explore. Stay tuned for many more examples to come!