Fun with Stats - Defense Edition, warning, may lack sound methodology

Submitted by Sebastien Belanger on Wed, 2017-02-08 16:02

Forums:

There are 2 types of successful defenses you can play, Glory D and Solid Coverage D. Glory D is measurable and memorable in that it is recorded as a D-Block and results in a game changing turnover. Solid Coverage D usually goes unnoticed but when incorporated in the culture of a team, it wins championships. In this week’s “Fun with Stats - Defense Edition, warning, may lack sound methodology” I will attempt to highlight some of the best and worst individual performances of 2016/17 Parity League Session 1.

For my analysis, I looked at two key statistics which were D-Blocks and Points Against (sum of OPointsAgainst and DPointsAgainst).

D-Blocks are a great measure of Glory D and can be used to identify those players that create turnovers with superior athleticism, anticipation, reaction and/or experience relative to their match-up. Of course, if everyone chose a favorable match-up, there would be nobody available to cover the top tier players OR your name is Owen Lumley. D-Blocks may also reward gamblers who choose to poach a lane or go for hand blocks at the risk of leaving themselves open to getting beat downfield or broken.
Points Against can be used as an indicator for Solid Coverage D only if it is adjusted for an individual team's Points Against relative to the league average. A player who is good at Solid Coverage D is almost always positioned on the force side, gives up no free bails while taking away Berkeleys and prevents back-breaking deep throws (this is more relevant indoor). Usually, it takes years of playing against people of equal speed to acquire this knowledge/technique.

Methodology (please skip if you don’t like math or you don’t care)

     Normalized D-Blocks per Games Played:
    Glory D = (D-Blocks / GP)*
    *Normalized out of 100 (i.e. highest league D-Blocks / GP = 100 and lowest = 0)

    Normalized Adjusted Points Against per Games Played:
    Solid Coverage D = ((((OPointsAgainst + DPointsAgainst) / GP) - ((Team Goals Against - Average of         Teams Goals Against)) / Total Number of Games)* - 100) x -1**
    Normalized out of 100
    ** Since the higher the Goals Against, the worse the Defensive measure, therefore the highest scores         need to measure the calculations furthest from 100   

    D Score = Average (Normalized D-Blocks per Games Played, Normalized Adjusted Points Against per         Games Played)
    Sample Size for inclusion in this analysis is 3 Games Played

Now that the awkward math part is out of the way, let’s get to the naming and shaming, blaming, listing. As always, let’s celebrate the good. Here are the top individual D-Block performances of the session.

Scott Higgens (Week 1) - 5
Kevin Barford (Week 2) - 5
Krys Kudakiewicz (Week 3) - 5
Ariel Untiveros (Week 5) - 6
Megan Robb (Week 7) - 5
Megan Robb (Week 8) - 5
Jonathan Pindur (Week 8 ) - 5
Justine Price (Week 8 ) - 5
Matthew Schijns (Week 10) - 5

If you made this list, congratulations, getting at least 5 D-Blocks in a game and making sure that they were obvious enough for the stats keepers to notice is no easy feat. Of course, if you are getting that many D-Blocks in a game, you’re probably covering people who are much slower than you or who are getting thrown to by Al (sorry Al, I couldn’t get through an entire post without giving you a shout-out).

While anybody can get lucky and put up some monster numbers for a single game, the next list highlight’s players who put up some mad D over a longer period of time. Because defense is so dependent on what team you’re on (i.e. the disc, like water, often finds the path of least resistance to the endzone) my methodology looks at players on specific teams for a minimum of 3 games which means players are listed more than once. Top 10 players with D-Blocks / GP are:

Ariel Untiveros    (F Bombs) - 2.50
Jessie Robinson (Like a Boss) - 2.25
Scott Higgins ((owen's) Basket of Deplorables) - 2.10
Martin Cloake    (F Bombs) - 2.09
Owen Lumley    ((owen's) Basket of Deplorables) - 2.08
Sebastien Belanger (Katy Parity) - 2.00
Megan Robb (Like a Boss) - 1.91
Krys Kudakiewicz (Katy Parity) - 1.80
Frederic Caron (Kaboom) - 1.70
Ariel Untiveros    (Katy Parity) - 1.60

Bottom 10 players with D-Blocks / GP are:
An Tran ((owen's) Basket of Deplorables) - 0.25
Sandra Hanson (Kaboom) - 0.22
Richard Gregory (F Bombs) - 0.22
Julia Laforge (Hindsight Hooligans) - 0.20
Andrew Spearin (Mike and the Milburys) - 0.17
Trevor Stocki ((owen's) Basket of Deplorables) - 0.14
Darlene Riley ((owen's) Basket of Deplorables) - 0.00
Rachel So (Katy Parity) - 0.00
Tanya Gallant (Like a Boss) - 0.00
Jason Fraser (Hindsight Hooligans) - 0.00

Most of the people on the first list are experienced players, tall, athletic and/or are very good at choosing to cover players with limited experience (that’s my strategy). I’m also certain that there is much poaching and hand blocking, which while risky, can be effective if not overused. Special shout-out to Ariel who was able to put up monster numbers on two different teams on opposite ends of the standings. The second list has a few players that have no D-Blocks credited to their names and is a better reflection of how good players are in this league across the board.

After generating my completely unscientific attempt of calculating a Solid Coverage D Score (out of 100) and combining it with a Glory D Score (out of 100), here are my final lists.

Top 10 D-Score:
Martin Cloake (F Bombs) - 82.79
Scott Higgins ((owen's) Basket of Deplorables) - 79.50
Ariel Untiveros (F Bombs) - 77.08
Owen Lumley ((owen's) Basket of Deplorables) - 77.08
Ariel Untiveros (Katy Parity) - 74.78
Sebastien Belanger (Katy Parity) - 74.03
Jamie Wildgen (Kaboom) - 73.80
Jessie Robinson (Like a Boss) - 71.39
Krys Kudakiewicz (Katy Parity) - 68.78
Kevin Barford (Hindsight Hooligans) - 68.23

Bottom 10 D-Score:
Rob Ives (F Bombs) - 31.94
Kindha Gorman (Katy Parity) - 30.28
Tim Kealey (Mike and the Milburys) - 30.14
Janet Ibit (F Bombs) - 23.75
Chris Tran (Katy Parity) - 21.67
Graham Brown (Like a Boss) - 20.38
Laura Chambers Storey (Mike and the Milburys) - 19.86
An Tran ((owen's) Basket of Deplorables) - 18.54
Ashlin Kelly (Katy Parity) - 13.33
Trevor Stocki ((owen's) Basket of Deplorables) - 10.60

This post is already long enough so I won’t add any more analysis than what is already here but I would like to point out that many of the players on the “Bottom 10” lists end up covering players that are more athletic than they are but they (unlike myself) are willing to sacrifice their pride for the good of the team by taking on that challenge. Others are just not great defenders. You can decide for yourself which one is which....

Christopher Keates

Wed, 2017-02-08 16:20

How can we check your work, if we can't check your work?!

Sebastien Belanger

Wed, 2017-02-08 16:23

I'll see what I can do...

Hadrian Mertins-Kirkwood's picture

Hadrian Mertins...

Thu, 2017-02-09 09:28

Cool breakdown! As a coach, the D blocks stat is one of the most frustrating because it doesn't usually reward the player who actually created the turnover. Often, a good mark and good downfield defense force a bad throw to a third best option and then that defender gets the block.

Mike Lee only had 5 blocks in session one and Ashlin only had 9, but it's not because they play bad defense. They just shut their player down and therefore have fewer opportunities to get Ds. Other solid defenders with few blocks in session one include Keates (8), Ben Piper (8) and myself (8).

Mike Lee

Thu, 2017-02-09 09:38

Don't forget I have t-rex proportioned arms and am unable to put on a good mark. 

Sebastien Belanger

Thu, 2017-02-09 13:33

Just because a player is good on offense and is athletic, it does not automatically mean that the same players is a good defender. While I agree that D Blocks alone should not be used to assess a player's defense, I'm proposing that other stats can be used to accomplish this assessment and this is what I've attempted in my analysis. While I agree there are some surprising names on the Bottom 10 lists, I am interested in using session 2 to validate my methodology by watching some of those players (i.e. the eye test).

Rob Ives

Thu, 2017-02-09 11:00

Very interesting Seb.

One flaw - a player starting all their points on O being scored on 5 of 20 points is going to appear better at D than a player starting all their points on D being scored on 10 of 20 points.

If you do it by defensive opportunities I think it would better reflect the D (i.e., turns per opponent's possession).

Just a suggestion (likely driven by self-interest).

 

Christopher Keates

Thu, 2017-02-09 11:39

There is significant variance in the number of D and O points a player has played (Kevin Barford, 80-126, Maya Popovich, 97-144 are examples), but you are using absolute numbers not relative rates.

Of note, the league as a whole seems to convert on offense about 61% of the time, and on D about 39% of the time. Above and below average teams skew away from this significantly (look at F-Bombs players from session 1 v. Like a Boss players from session 1).

Public Sheet - 2017 Totals gives raw O and D point numbers, CK - Derived Stats has some conversion % numbers.

Sebastien Belanger

Thu, 2017-02-09 13:21

I assume that the variance in the number of D and O points is correlated to a team's GA and GF. In other words, if a team is getting crushed, most of the shifts will be on O to begin with until they turn it over. To compensate for this, I used the average Team GA for the league divided by Games Played (12) and compared it to an individual Team's GA per GP. I then applied this difference to individual players GA per GP which increased or decreased the value depending on what team they were on. Using your suggested approach of applying it to individuals instead of teams may be more accurate. If I do this analysis again, I will definitely try it out.

Christopher Keates

Thu, 2017-02-09 13:39

A good example of how skewed this can be:

Let's look at F Bombs, and two players that were with them for the duration of session 1.

The team was 246:185 in GF:GA, or 57%/43%. 431 points total.

Rob Ives played 241 points, starting 96 on O (40%) and 142 on D (60%), which is skewed slightly over expected.

Sully played 246 points, 121 on O and 125 on D (50/50), which is far removed from the team's expected distribution of O v. D starts.

What you've done is good and interesting and descriptive, but I think still too tied up into team effects.

Sebastien Belanger

Thu, 2017-02-09 13:43

You're right, I assumed that teams just rolled lines (which is one of the unwritten rules of parity), but obviously they didn't. God bless Rob Ives for volunteering for so many D points.

Christopher Keates

Thu, 2017-02-09 14:07

There's a Sully joke in here somewhere.

Owen Lumley

Thu, 2017-02-09 14:21

I've been doing this sort of analysys, but using the play by play data to do it by defensive attempts (every time the opposition gets a new posession of the disc).  Also to address Hadrians point, I've subtracted turnovers caused by Dblocks to get a sense of what players seem to have a greater effect on getting the other team to throw away or drop.  So the calculation is something like this

D% = (Dturns - Dblocks) / Dattempts

where Dblocks are not just yours, but any while you were on the field playing D.

As has already been mentioned, the result greatly affected by the other players on your team (4 out of the top 5 are Fbombs).  But not all Fbombs are near the top.  To try and eliminate a bit of the team bias, I calculate the weighted average of D% for all the players you've played D with - then subtract some percent of this from your own value:

D%(estimate_of_you) = D%(you) - (Factor)*avg_of_linemates(D%)

This still has lots of problems, but there is clearly a difference between those at the top and bottom of the list.  Interestingly, those at the top are not necessarily those thought of as top defenders, but more likely the best defenders against the level of matchup they normally face.  And not best defenders in terms of getting Dblocks, but in terms of not letting easy breaks, covering dumps, and covering the force side cutting lane.  Mike Lee, even with his Jurasic arms, is in the top 10.  I'm somewhere in the middle.  Some fast, athletic players who are good at offense are near the bottom.

 

 

 

Sebastien Belanger

Thu, 2017-02-09 14:21

Maybe this is why Sully plays so much O http://ftw.usatoday.com/2017/02/tom-brady-pick-6-missed-tackle-sad-new-e...

Chris Sullivan

Thu, 2017-02-09 15:56

ouch

If i could tie my shoes without the help of a friend i'd rebut that slag.

As a Steelers fan I'd prefer to be compared to the pear shaped weeble wovil that is Ben Roethlisburger 

http://www.foxsports.com/nfl/laces-out/ben-roethlisberger-pittsburgh-ste...

Hadrian Mertins-Kirkwood's picture

Hadrian Mertins...

Thu, 2017-02-09 16:31

I think this discussion highlights the fact that on defense the team matters a lot more than the individual. Even if you're the strongest individual defender in the league, your defensive performance is basically determined by your teammates. You can only do so much on your own.

In contrast, two or three good players can run an effective offense even if the rest of their team is much weaker. Individual skill is rewarded more on offense (especially on the stat sheet).

I think we all know this intuitively. A weak player on offense is a non-factor, whereas a weak player on defense is a liability.

Kevin Hughes's picture

Kevin Hughes

Sun, 2017-02-12 20:33

Seb, remember we have 3 years worth of data for some players in this league. For stats like this where the team has an affect using multiple years will help reduce the affect the team has.