What are the best player stats for Ultimate?

[Mark]

We're in the midst of building out all our Ultimate stats
functionality on Leaguevine and in apps built on top of the
Leaguevine API. We have tons of strong opinions on how this
should be done right, but we've been living and breathing
this space for so long we might be too close to the problem
and I'd like some more fresh perspectives before we launch.
We'll launch the first version right around the start of the
AUDL, as they'll be using our stats system.

What are your top 10 favorite individual player stats
categories you'd like featured? We'll obviously have far
more data to work with than this, but we want to get a feel
for priorities. Please give some explanation as to why
certain categories are better than others.

Here are some things we have the potential to build/display
in real-time. Don't worry about how it's collected, that's
under control. I just want to hear what you want to see.

games played
points played
+/-
turnovers
throwaways
drops
catch %
Ds
blocks
handblocks
footblocks
callahans
picked up disc
pulls
inbounds pulls
out of bounds pulls
inbounds pull %
passes caught
passes thrown
complete passes thrown
incomplete passes thrown
completion %
violations called
violations against
fouls called
fouls against
fouls called contested
fouls called uncontested
fouls against contested
fouls against uncontested
(various other violations including stalls, marking fouls,
strips, travels, picks, double teams, fast counts, disc
space)

Also, do specific stats based on throw types break into your
top 10? We won't have this built into our first product, but
it's still interesting for the future. I'm referring here to
throw types (forehands thrown, forehands completed,
forehands incompleted, forehands thrown away, forehand
completion percentage, forehand catch percentage, backhand,
hammer, scoober, push, etc.), pass trajectories (inside out,
outside in, flat), and pass categories (dump, swing,
upfield, huck).

-Mark

p.s. If any developers want to get involved, go here and
introduce yourself:
http://groups.google.com/group/leaguevine-developers
--
Posted from http://www.rsdnospam.com

[Dick]

Points played
+/-
Goals
Assists
Passes thrown
Throwaways
Completion %
Receptions
Drops
Reception %
D's
(Callahan would count as an unassisted goal and a D)

I know that's 11, but that's the only sequence of stats that make
sense to me.

[Dick]

Actually, I'm immediately rethinking some of these. Here's my revised
list:

Points played
+/-
Goals
Assists

Completions (even if your receiver makes an insane grab, still counts
as a completion for the thrower)
Incompletions (includes being point blocked, throwaways - treated
exactly like a quarterback - if the receiver drops a perfect pass, its
still an incomplete pass in the thrower's statistics)
Completion %
Reception
Drops (anywhere in receiver's catching range not caught should count
as a drop)
Reception %
D's (no differentiation between a lay out D, a point block, etc. - a D
is a D in the scheme of a game)
(Callahan still count as an unassisted goal and a D)

I know that I didn't tweak them much, but I think this way makes more
sense. Also, the notes kind of clarify the way I'm imagining this.

[Bulb]

Not sure why you would want to lump all D's together. At
least split them into marking blocks and other.

Reception % is silly, as a mediocre handler could have 100%,
while some of the best receivers are going to be lower
merely because they were overthrown a few times.

Completion % is more telling, but can be computed from
passes thrown and throwaways, so you should only keep 2 of
those 3. I'd rather know the total passes thrown and the
completion %.

Drops? Seriously? Are any of the top players going to have
more than 1 or 2 drops over the entire season?

+/- is interesting, but I think it will be misleading. O
line players will be higher than D line players, on average.
I'd rather know what impact a player has on his team's
chances of scoring (or stopping an opponent) per possession.
Call them O conversion and D prevention(?) % maybe?

(Sidebar: I realize that one stat or the other will be N/A
for a lot of players, but for the guys who are playing both
lines, they might be really useful for coaches. You might
have a guy who's a great defender, but for some reason, his
D prevention % is considerably lower than his O conversion
%. Knowing this might make you reconsider his role, or bump
up another guy in the D rotation.)

Points played
O conversion %
D prevention %
Goals
Assists
Passes thrown
Completion %
Receptions
Blocks
D's

[Matt Sheehan]

The single best stat for handlers, once it can be computed,
will be:
Yards/turnover

As in, you complete ten 10-yard passes and then throw a
turnover, so you're at 100. This is the best way to
aggregate a handler's contribution to a team.

[Dan Anderson]

> Goals
> Assists

I think it's more telling to have this broken out as:

Goals Thrown
Goals Caught
Assists

Assists should be tracked like basketball/hockey assists, where you
would award them when a pass leads directly to, and aids in, a Goal
Thrown.

[Andrew]

So my 10 would be

Points played
#O points ( and maybe % O points)
+/-
goals
assits
completion
incompletion
completion %
Ds
receptions

Some more to consider just for funsies-fouls, 2nd assists
(assisting the assist), assits/completions thrown,
goals/receptions, throwaways/(layouts+skies+ds)

[Andrew]

I see that Bulb and I think alike. Does that mean I have a
great mind?

[Andrew]

I've never taken a stats class in my life and know almost
nothing about making up (deriving?) stats, but is there way
to develop an adjusted plus minus? Adjusted +/- in
basketball and hockey accounts for the effect of other
players on the court so that the stat is supposed to reflect
a players contribution regardless of all the other players
on the court/ice. Could there be an ultimate adjusted plus
minus that accounts for the other players on the field, and
specifically if a player is playing and O point or a D
point?

With an unadjusted +/-, players playing more O points will
almost always have a higher rating than D line players.
This could be helpful in rating how effective a D line is,
because a D line, and predominately D players, with positive
plus minuses are clearly doing something right. The
opposite would be true with O line players--a team could
look at an O line/O players with low/negative +/- and figure
out what is going wrong.

The exceptionally gifted ultimate players will always rate
highly in an unadjusted +/-. Many of the "best" players on
a team play predominately O points, or are called on to D
lines when a team knows they need a break. Just comparing
the number of O points against the number of D points a
player has played, and considering that with an unadjusted
+/-, would be a step in the right direction. An adjusted
+/- would help compare all players in an adjusted standard.
I just have no idea how to make this work.

My knowledge of advanced stats is extremely limited. I
think PERs and winshares are fun to look at in basketball,
but I know almost nothing about deriving them. When I did I
cursory search for the definition of adjusted +/- in
basketball and hockey, my head swam, then quickly drowned,
as I tried to figure out how they had been derived. Maybe
there already is an ultimate adjusted +/-? This may not be
a stat that will be included in the first build of your app,
but it would be cool to look at advanced ultimate stats, and
I think that the potential bias against D players built into
an unadjusted ultimate +/- makes it a good place to start.

[poaches]

what about tracking when people get scored on or when people
get broken? I realize that "getting scored on" won't always
be 100% accurate and it would be skewed for players who
mostly guard cutters, but to some extent I think it would be
useful for some teams.

points played
O points played
+/-
goals
assists
turnovers
Ds
blocks (on the mark)
scored on
broken

pull stats are also good (total pulls in, pulls out of
bounds)

[Julian Bonds]

Don't forget:

[How good a player is]/[How good the player thinks he is]

Need this to be as close to 1.0 as possible.

[joaq]

+/-? Seems biased for O players.

[Chris]

If they play both have the players stats seperated between
the two disciplines - you might find that the players is a
lot more proficient at one than the other.

I just quickly posted teh following on facebook which asked
teh same question:

"scores, assists, blocks, offensive scoring rate, offensive
turnover rate, defensive scoring rate, defnesive turnover
rate, individual offensive player completion rate when on
field, individual defensive player completion rate when on
field, 40m+ throwing completion rate, foul/violation count,
"

Someone asked why I cared so much about whether someone was
on O or D, and if that player played both O and D why do you
need to have their stats on the different lines seperated.

To which I replied:

"If they play both, have the players stats seperated between
the two disciplines - you might find that the players is a
lot more proficient at one than the other.

As for why, because if a player plays only defense then in
50% of the points he's not even going to get the chance to
throw an assit or score a goal, and you need to be able to
get that level of detail in the stats if you plan to make
any decisions based on the stats otherwise the data might
draw the wrong conclusions about the player.

The same goes for blocks, if an offensive player plays
nothing but offensive you'd expect that their block count
would be really low. When reviewing that players stats you
might want to know that.

As for noting who is on the field on either an offensive or
defensive point and whether that point was successful as a
team, I think this offers a far more accurate information
for analysis (assuming you get a big enoguh data set - ie at
least 10 games worth of data) because it will help account
for all the discrete, yet critical, things players do on the
field that don't result in stats.

So on defense this would include things like, shutting your
players cuts down, having a really aggressive thrower
marker, poaching well and clogging the lane when possible,
forcing dumps backwards, preventing breaks, having great
speed endurance - all these sorts of things lead to turns,
but none of them would be rewarded with simiplistic stats
models.

The same thing goes for offensive, some players might not be
the super long hucker, or big tall athletic receiver, but
they may have an uncanny way of always getting the disc,
alwasy getting open on the dump, never taking speculative
shots, might be a great mid cutter/link man, which might
increase an O lines chance of success, yet might not give
credit where credit is due."

[Craig Manahan]

Percentage of points a team scores when that player is on O
/ When he is on D will give some sort of statistic for how
well he plays as a teammate (ex since when you are playing
the best D and shutting down your defender you might not
(probably won't?) get a D). Obviously skewed a lot by other
players on the field but I think over the course of a season
it would give a good overall ranking of a player.

[Bulb]

Matt Sheehan wrote on Mon, 30 January 2012 12:29

> The single best stat for handlers, once it can be
> computed, will be:
> Yards/turnover
>
> As in, you complete ten 10-yard passes and then throw a
> turnover, so you're at 100. This is the best way to
> aggregate a handler's contribution to a team.

Interesting idea, for sure. But how do you account for dump
passes? Do they count as minus yards?

The concept of a hockey assist, as a statistic, doesn't seem
all that useful in Ultimate. Of course, there are
situations where a player gets off a risky break throw that
sets up an easy score, but far more frequently, the throw
before a score is just a regular pass.

Tracking when you get scored on and when you get broken
would be useful too.

[RoShamShot]

UltiStats has a relative +/- that takes in if it is an O or
a D point and does some normalizing. Your r +/- can be high
even if you only play a few points if (for instance) your
team gets 3 breaks on D and those are the only breaks of the
game and those are the only 3 points you played

I know the guy who did the port from palm to applestore
comes by occasionally and still does some work on it. It
would probably be the easiest if you guys combo'd with
ultistats because it already has an the ability to send the
stats to an e-mail address *** and then it could autoupdate
the stats. And if possible you could Bluetooth connect
devices for games/the whole tournament so stats could be
cross compared maybe?

*no idea how to do any of this/if it's possible but .......*

[RoShamShot]

Also, I thnk ultimate would benefit from the 2 pass-assist
system like hockey

[Jackson]

What about something like Augmented Yards per Throw. Here is
my idea in a raw and unrefined state.

This would be very difficult to compute, but I think would
be a really cool stat if the data set was available. You
would need to know the thrower/receiver of each pass and how
many vertical yards the pass traveled.

The idea of the Augemented Yards per Throw (AYT) is to
reward throwers and cutters for their throw as well as
subsequent throws. If you throw a difficult break throw that
sets up easy continuations that gain yards, this stat will
capture your contribution to gaining those yards. The
subsequent throws are weighted with some sort of decay
function (maybe logarithmic?) and the weights are normalized
to 1.

Augmented Yards per Throw:

Player A gets credit for the yards gained for catching the
pass (throw generation 0) from player Z with weight 1.
Player A gets credit for yards gained on his throw (throw
generation 1) to player B with weight 1. Player A gets
credit for yards gained on player B's throw to player C
(throw gen 2) with a reduced weight (for example 3/4). This
process continues for a few throws. You'd probably want to
include at least upto a gen4 throw to capture a typical
dump-swing pass with 2 more continuations.

This could also be adjusted with a multiplier so that a 5
yard pass is worth more from 10 yards outside of the endzone
than from midfield. Maybe you'd want to track that as a
separate stat that replaces yards with rating-points or
something like that.

The AYT rewards players who set up their teammates. One
issue that immediately comes to mind is that it punishes
zone offenses that relentlessly swing the disc back and
forth (do any good teams still use this strategy?).There are
probably other issues with it too, but I'd be happy to hear
feedback.

[Chris]

Twoorii wrote on Mon, 30 January 2012 17:33

> UltiStats has a relative +/- that takes in if it is an O
> or a D point and does some normalizing. Your r +/- can be
> high even if you only play a few points if (for instance)
> your team gets 3 breaks on D and those are the only breaks
> of the game and those are the only 3 points you played

No stats are valuable unless you have a big enough sample
size. You'd probably need a player to have played 30+ D
points with a mixuture of players before you even started
analysing the data.

So for example, lets say in the case of the AUDL, if they
play 6 games in thier region, ie, they played their 3
regional opponents twice, and each game had an average of 25
points. That would be a total of 150 poitns being played,
which would be spread between 20 players, so 150/20*7, means
the average player would play 52.5 points. With that sort of
sample size you could start looking at the data and start
drawing some conclusions, but I still imagine the sample
size would be too small.

In all honesty, when it comes to team based data, most teams
will probably only collect enough data in 2012 to start
making strategic decisions for their 2013 season.

[Ryan Thompson]

Goals
Assists
Passes Attempted
Passes Completed
Breaks Attemted
Breaks Completed
Hucks Attempted
Hucks Completed
====
Marking D's
Receiving D's (maybe specify in/deep D's?)
Targets of opponent
Touches by opponent
Times broken
Times hucked on (possibly for marker and downfield D)
Times scored on

[Mark]

Twoorii wrote on Mon, 30 January 2012 17:33

Chad, the creator of Ultistats is the lead developer for the
Leaguevine version of Ultistats as well. We're making this
app free and open source, so others can help and it'll be a
team effort within the Ultimate community.

Even better than sending the stats via email, we'll be
syncing them to Leaguevine so people can view them in real
time. Yes, it's a huge task, but we have a lot of smart
people on it and you can expect to see the first version in
action in about 3 months.

[Mark]

Some good things I see in this thread that are being
endorsed a lot are:

2nd assists like in hockey. This is something we hadn't
thought about, so thanks for bringing it up. Very
interesting.

Relative +/- that takes into account O and D. This is a
great one that people have talked about for years and I
think will end up being very useful.

The two suggestions that I think are most interesting are
the ones suggesting some sort of Augmented/Adjusted +/-, and
the ones suggesting yards per throw. Anyone have more
thoughts or ideas on these? I think you guys are onto some
interesting stuff here. This thread kind of feels like the
birth of Ultimate stats. Also, I should note, the stats
involving yardage probably won't be possible until later
versions of the app, but it's worth hashing out the details
early.

[cooper]

I'm excited to see these stats in action. The initial
learning curve will be steep as we discover by doing. In
other words, I wonder how much our opinions on stats will
change once we start actually getting stats to study.

My four ideas:
1) Consider how stats are grouped for presentation. To me,
stats are less effective given one at a time and better as a
basket of related stats. For example, here might be some
good ways to present combinations of stats that tell us a
fair amount at once:

Overall simple
-# points: #O points (w/ % team scores), #D points (w/% team
scores), +/-(we instantly see how many points total, how
much they are O vs D, and then we get a better sense of the
+\- as we can simultaneously see if they are mostly O or D
and see their O v D effectiveness). This would look like
this:
P7, O5(75%), D2(25%), +2 or with a header bar and then
stats
P: O(%) /D(%), +-
7: 5(75%)/2(0%), +2

Overall expanded:
-# points: #O points (% scored), #D points (% scored),
Throws, completion %, goals thrown, goals caught, total Ds.
If I'm only going to get one line of stats, I think I want
the one above (funny, that is 10, as requested). It would
look like
P: O(%) /D(%), T, C%, GT, GC, D
7: 5(75%)/2(0%), 31, 94%, 2, 1, 3.


Throwing
-Throws, completion %, goals thrown, assists. (We could
calculate, or at least get a sense, of turnover rate and
scoring efficiency while getting, to me, all the current key
thrower stats) Looks like this:
31T, 94%, 2GT, 1A or with head bar
T, %, GT, A
31, 94, 2, 1

Cutting
-Receptions, reception %, goals caught Look like:
9R, 100%, 1GC or
R, %, GC
9, 100, 1

Defense
-Opponent touches, off-mark Ds, marking Ds, goals conceded:
24OT, 2D, 1MD, 2GC or
OT, D, MD, GC
24, 2, 1, 2

2) Positions. We might be more interested in different
chunks of stats for different general positions. Obviously,
all ultimate players have the same rules and therefore the
same appropriate stats (unlike football), but I might care
more about a handler's throwing stats, cutter's receiving,
and D player's D. Indeed, we all throw, cut, and play D,
but maybe it would make more sense, if there is streaming
coverage of a game for example, to show a bar of stats under
a player that reflects what we care about most.

-Handler: throws, completion %, goals thrown, assists (see
above)

-Cutter: receptions, reception %, goals caught (see above)

-Defender: marking Ds, off-mark Ds, opponent touches, goals
conceded

3) Too many stats. OK, not a problem for the internet/those
interested, but a possible issue for general presentations.
In other words, there are stats galore for baseball, and you
can look them all up online. Awesome. But when watching a
game, they keep it simple. I think most people would rather
stick to H, R, HR, and AVG than OBS+ and GDP. For that
reason I say, keep it simple, stupid, for general
presentation. I think it would be great to look up breaks
attempted/complete, hucks attempted/completed (forgetting
about the subjective nature of these decisions), but if I'm
watching ultimate on the web, I'd rather just see: throws,
completion %, goals thrown. That, I think, willl give me a
good sense already of a handler's effectiveness. And yes,
tennis give backhand/forehand stats, which could be good for
ulty, but I'd stop at backhand/forehand/other or it gets to
be a long list.

4) If you record "discs picked up" (which I like - shows who
the dead disc handler is), why not collect "pulls received"
or even "pulls received/points started on O" to get you a %
of times player takes the pull?

[RoShamShot]

Other things my team has been wondering as features (since
there is some discussion)

A simple thing that we always wanted was pulls, and if they
were bricked

Since there is already a clock feature would it be possible
to see the average disc hold time? We always thought it
would be interesting to get real numbers for if people keep
holding it to 7 or if they move it in 1-2



On the request for yards per throw and similar....have any
of you tried doing using ultistats before? It would be near
impossible to see the thrower/receiver remotely judge the
distance and then click it in before the next set of passes
*at least on an iPad, something smaller might be easier

[RoShamShot]

Also, a different sort of +/- like in fantasy ultimate with
goals/assists being positives and throwaways and drops being
negative. We always had a problem with this.

[Bulb]

Sippy wrote on Mon, 30 January 2012 19:14

Great point - the converse can also be true. Some players
are showboaters on D, always baiting a throw to their guy
(either in or deep). So they might be leading the team in
D's, but negatively affect their team's chances of winning
the point.

Twoorii wrote on Mon, 30 January 2012 20:35

> Also, I thnk ultimate would benefit from the 2
> pass-assist system like hockey

Why?? I know several people have mentioned this, but not a
single person has provided logical justification. I am not
claiming the concept of a hockey assist is irrelevant in
Ultimate. But I am claiming that it is WAY more variable,
in terms of requiring skill/talent, than it is in hockey.
That is to say, the average assist in hockey requires a
level of skill/talent that, for the most part, only NHL
players possess. Conversely, many "hockey assists" in
Ultimate, even at the elite level, come from throws that
many non-elite players could complete.

thundercats wrote on Tue, 31 January 2012 01:05

Great post! (I especially love that you did not even
mention hockey assists.) Different stats for different
positions would be cool, but some people will have stats all
over the board - I guess we could ask the teams to assign
each player a position, though.

[Ryan Thompson]

Except he did mention hockey assists, since his Throwing category has
both goals thrown and assists.

I also agree that "hockey assists" don't really seem necessary,
especially if we track break throws and hucks and have completion
percentages for both.

[Chad]

One of the stats that I compute in UltiStats is relative
plus minus. I got it from here:
http://groups.google.com/group/rec.sport.disc/msg/fb9f0f2e03f06a79

Another is the adjusted plus minus that I got from Dar or
from a commenter on his blog... I can't remember... sorry!
(Could have been Parinella or Burrus)

I've thought a lot about these topics and I have also
received questions and suggestions since UltiStats for
iPhone (http://iphone.ultistats.com) went live.

My common response and something I hope you consider as
these discussions continue is the following:
How useful is a stat to you if it has some level of
subjectivity? What level of subjectivity is acceptable?

Let's run through a few examples.

Hucks:
1) You're running a side stack and all your cutters except
for one are out of the way. The lone cutter goes way deep,
her defender follows, then your cutter comes back under. You
complete a 40-yd pass that never clears 6-ft off the ground.
Is that a huck?
2) You're going upwind. The D is running a 4-man cup and the
deep-deep underestimates you so he's way too close. You
throw a 30-yd lead pass to your wing who receives it behind
everyone else. Is that a huck?

Breaks:
1) The other team is forcing middle... your defender
mistakenly thinks you are still close to sideline X so they
keep the force but you are really close to sideline Y and
throw it up the line. Was that a break?
2) Brodie has the disc so you play straight-up to stop the
huck. He gets a short pass off to the break side (or what
would be the break side if you were marking normally).
Should Brodie be awarded a break?

I think a lot of information that people want to get out of
tracking 'hucks' and 'breaks' can be gleaned if we know the
x,y coordinates (in field space) of every reception. This is
conceptually easy but, until we have professional
stat-takers, tracking this would be a nightmare. As a poster
mentioned above, it is difficult enough to tap a player's
name to indicate they got the disc... adding more parameters
to each pass would be nearly impossible from the sideline. I
would appreciate any design suggestions that would
facilitate keeping track of this.

Fun fact: The very first version of ultistats for iphone
(never seen by anyone but me) required users to drag the
disc around the virtual field to indicate a pass. Lifting
their finger caused a halo of options to pop up around the
disc including potential receivers and other possible
outcomes. It was a little too much.

I'm still thinking about it though. Eventually I'd like
enough data that we could replay a game like this:
http://www.sportsnetwork.com/merge/tsnform.aspx?c=sportsnetwork&page=nhl/rink/replay.aspx?id=14920

Once we aggregate the data online then we'll be able to do
some data mining and maybe come up with some more useful
(i.e. win-predicting) team-statistics and then maybe we can
figure out which individual statistics contribute most to
those team statistics. We might also be able to get some
player-rankings from experts and then see what sort of
measures we have that correlate with the expert-rankings.
There are a bunch of fun machine-learning things that are
possible once we have enough data.

[Chad]

[sorry if this is a double post]

[Dan Anderson]

On Jan 31, 6:00 am, Bulb <kevin.patrick.mccorm...@gmail.com> wrote:
> > Also, I thnk ultimate would benefit from the 2
> > pass-assist system like hockey
>
> Why??  I know several people have mentioned this, but not a
> single person has provided logical justification.  I am not
> claiming the concept of a hockey assist is irrelevant in
> Ultimate.  But I am claiming that it is WAY more variable,
> in terms of requiring skill/talent, than it is in hockey.
> That is to say, the average assist in hockey requires a
> level of skill/talent that, for the most part, only NHL
> players possess.  Conversely, many "hockey assists" in
> Ultimate, even at the elite level, come from throws that
> many non-elite players could complete.

Assists would not necessarily be awarded for every goal thrown, just
for goals thrown where the assist 'aided' the goal thrown (and
obviously this would be a bit subjective). I'm not clear on the exact
differences between a hockey assist & a basketball assist, but if I
was keeping stats for my team I would be interested in seeing this
type of assist tracked. It demonstrates the ability to think a pass
ahead, and to play selflessly within our team's system.

If someone makes a nice swing across the goal line that results in an
easy scoring pass, the current result of that is no stat for the swing
and assist for the easy scoring pass. And if you have someone on your
team who is great at creating scoring opportunities for their
teammates and can get the disc into that position, it should be
tracked.

If you're concerned about the skill/talent aspects regarding this
stat, you may want to remember that there are many goals thrown that
are just poor throws taken by selfish players and bailed out by
awesome receivers. And since there is no judgment used when counting
that stat, all of those are tracked.

[Andrew]

Part of the problem with the assist debate is that ultimate
uses the term assist differently than basketball/soccer/lax.
In Ultimate someone throws an assist to a receiver in the
goal, and the receiver scores the goal. In those other
sports, the receiver of the assist goes on to score the goal
by shooting the ball into a goal.

Our situation is more like football, where the qb throws to
a receive in the endzone, and if the receiver catches the
pass, both players are credited with a touchdown. I don't
have a problem with the way ultimate uses the term assist
but we need to take into consideration that it is not the
same as a basketball assist. The person throwing an assist
in ultimate is throwing to a human receiver who can
catch/drop the pass, not shooting into a goal.

Considering this, I think recording second assist rewards
good decision making around the goal. Simply keeping track
of touches, receptions, or completed passes do not reflect
field position. A second assist doesn't have to be the same
as hockey, where up to two players can be credited with
assists for each goal. I'm arguing for a stat separate from
the traditional ultimate assist that will show who is
distributing the disc to the two people who end up scoring
the goal.

[Reggie Fanelli]

what are the best player stats for ultimate???????

easy




MINE!

[Bulb]

Ryan Thompson wrote on Tue, 31 January 2012 09:38

> Except he did mention hockey assists, since his Throwing
> category has both goals thrown and assists.

My bad - missed that somehow.

bullale wrote on Tue, 31 January 2012 10:35

> My common response and something I hope you consider as
> these discussions continue is the following:
> How useful is a stat to you if it has some level of
> subjectivity? What level of subjectivity is acceptable?

Excellent question. Most sports have a subjective stat or
two, like errors in baseball or the number of times a
quarterback is "hurried" in football (more a stat for an O
line than for a quarterback). I think the nature of
Ultimate allows for a lot of subjectivity - moreso when it
comes to turnovers than assists. I think some subjectivity
when assigning assists would be a good thing.

Dan Anderson wrote on Tue, 31 January 2012 11:01

> If someone makes a nice swing across the goal line that
> results in an easy scoring pass, the current result of
> that is no stat for the swing and assist for the easy
> scoring pass. And if you have someone on your team who is
> great at creating scoring opportunities for their
> teammates and can get the disc into that position, it
> should be tracked.

It should be tracked if the scoring opportunity was set up
by the penultimate pass. If I throw a dump at midfield to a
guy who puts up a stall 9 huck for a score, I shouldn't be
given any credit.

Dan Anderson wrote on Tue, 31 January 2012 11:01

> If you're concerned about the skill/talent aspects
> regarding this stat, you may want to remember that there
> are many goals thrown that are just poor throws taken by
> selfish players and bailed out by awesome receivers. And
> since there is no judgment used when counting that stat,
> all of those are tracked.

Yes but you're forgetting that turnovers are also tracked.
So if you have a player who is always throwing into the end
zone, you can easily tell whether he is being selfish or
making good decisions. If his goals thrown to turnover
ratio is 9:1, he's probably making good decisions.

[Dan Anderson]

Yes, I agree that the current usage of the term 'Assist' is not
comparable to the meaning from hockey/basketball. I would change the
name for that stat to Goal Thrown.

Assists
Goals Thrown
Goals Caught

[Alex Peters]

This is only tangentially related to the tracking of stats
for yardage, but I really hope they don't use football lines
for their fields. It's bad enough that they are using 53
yard wide fields (why again?), but the lines make the field
look really bad. It's one thing in football where you are
specifically looking at the lines think "man, they just need
to pass that marker for a 1st down!" but in ultimate the
specific field position is relatively unimportant and just
makes the view messy.

Also, this thread is focusing a lot on what stats you might
use to most easily identify a quality player, but I think
from an announcing standpoint, I would love to see stats on
what kind of throws guys are hitting/turning over. "Keith
goes for the endzone with a flick... way out the back. You
know, Bob, Keith has been turning over his flick a lot
lately, he's only completing 66% of his flick hucks over his
last 9 games, while he's over 90% on his backhand. Maybe
that's why the Spinners are forcing flick this quarter, to
contain that monster backhand."

[Eric]

I don't think you should ever count break throws as a
statistic. The concept of the force-side / break-side is
based on the individual team strategy. What is a break throw
against a straight up mark? Against a progressively changing
mark?

I do, however, agree with the concept of the hockey assist.
While they may not occur on every point scored in ultimate,
a good "dump-swing-continue" for an easy goal should reward
the player who threw the swing (not just the goal thrower /
scorer).

My 10 stats:

Points Played
Goals Thrown
Goals Caught
Drops
Throwaways
Assists
Touches (possessions)
Blocks (includes d's handblocks, callahans)
Fouls
Something similar to QB rating (i guess this would be
adjusted +/-)

[Luna Tic]

Goals Caught
Goals Thrown
Assists
Personally I think the throw that leads to a thrown goal is
gigantic. Forcing the disk into the end zone via a huck,
break mark or tight pass..etc. leads to turns. The decision
to wait and make an extra pass can tell coaches a ton about
which players are making the best choices.
Not sure if it was mentioned but Red zone scoring % might be
valuable. Tons of teams can get the disc in the red zone via
hucks or shorter passes but repeatedly throw it away near
the goal line.

[Matt]

Something else that should be taken into consideration is
each stat on a per point basis. For example, 5 D's in 10
points wouldn't be as good as 4 D's in 4 points. I think
D's/point (and likewise assits/point, goals/point, etc.)
would be useful once you had enough of a pool of points to
draw from.

Once you have determined the X most useful stats for O-line
cutter, D-line cutter-marker, O-line handler, D-line
handle-marker, you could figure out a way to apply a weight
to each statistic, and eventually grade player vs. the
average in the league, similar to how ESPN developed Total
QBR prior to this football season. With enough data you
could eventually say, O-line cutter has played at an OLC of
77 (vs. average of 50) and the man marking him has been
played at a DLC of 40, so the offense might want to use the
mismatch. It would take a lot of data and fine-tuning, but
you could end up with a metric that gives a pretty good idea
of how strong a player is as a cutter, handler, etc. Many
sports have versions of this general idea, and although they
aren't perfect they do give a good picture of how good a
player is (often better than more traditional stats).

[Matt]

Breaks would be tough to keep track of, but I think it is
worth trying because they can be so important. Maybe only
track a break when it is clear (based on the mark's and
defenders' positions). It is subjective, but if the
commentators can differentiate what constitutes a break and
what does not, I think the stat can be counted fairly
accurately

For assists (Assuming the stats used are goals caught, goals
thrown, and assists), I think it should be a cross between
hockey and basketball. Multiple players can get assists, but
only if the throw is let go of before stall 2, for example.
If a break throw leads to an immediate huck and score, the
break throw deserves an assist because it directly lead to
the huck (the evidence for this would be that the huck was
let go of before stall 2). This is the same way assits are
kept track of in basketball IIRC - the player must have shot
within x seconds of catching the pass. I think the stall
count is fine for being used as the measuring stick.

Some sort of adjusted +/- would be great, but as another
poster noted, the sample size would have to be quite big to
get any info out of it. What I think would be interesting,
though likely impractical, would be unit cohesiveness. How
well the combination of players A, B, C, and D do when they
are all on the D-line together vs. ABCE or ABDE.

In order for stats to be useful (in terms of grading a
player's performance) I think stats need to be broken down
pretty far. Offensive vs. Defensive obviously, then O-line
vs. D-line, then Cutter vs. Handler.

For instance, how would you go about valuing D stats for
cutters? If you are marking their #4 cutter you will clearly
get more D's than if you are marking their #1 cutter, but
that doesn't make you the better player. I suppose this has
less importance at the highest level where there is more
parity on the field.

I think an important stat, perhaps most important for D line
players, would be look off%. If a disc gets D'ed, I think a
lot of the time it is just a poor decision by the handler
and not as much an indication of a unique individual play,
but if the handler looks off the receiver then the defender
has done his job and it should be noted.

For the O-line dump, the opposite would be true. How well he
plays as a dump has a lot to do with how well he gets open
on the first cut after he is engaged (probably moreso than
the skill level of the player making the dump throw, which
should have low variability at high levels), so a good stat
for a handler might be 1st cut dump completion%. You might
conclude that handler X might be a weaker player because of
an inability to get open right away.

I second the sentiment about tracking forehands and
backhands differently, although it would have to be a
secondary stat because you would need twice as much data to
accurately draw the same conclusions. I have no idea how
practical any of this stuff is, but if we are talking about
how to properly keep track of stats for a professional
league, these would be important. It would be great to be
able to say that 85% of forehand hucks have been completed
for the season vs. 70% backhand.

[Bulb]

Ichiroo wrote on Wed, 01 February 2012 01:25

> For instance, how would you go about valuing D stats for
> cutters? If you are marking their #4 cutter you will
> clearly get more D's than if you are marking their #1
> cutter, but that doesn't make you the better player.

You won't "clearly" get more D's marking a team's #4 cutter,
as that team will be throwing to their #1 cutter way more
often. You will get more D's as a percentage of how often
that cutter is thrown to, but that can easily be measured
(see below).

Ichiroo wrote on Wed, 01 February 2012 01:25

> I think an important stat, perhaps most important for D
> line players, would be look off%. If a disc gets D'ed, I
> think a lot of the time it is just a poor decision by the
> handler and not as much an indication of a unique
> individual play, but if the handler looks off the receiver
> then the defender has done his job and it should be noted.

Agreed, though I think there is a better way to give the
defender credit than to keep track of "look offs," as that
would probably be very tedious, as well as subjective. If a
stopped disc play calls for 3 cutters to make 3 different
cuts, and your guy doesn't get thrown to but another cutter
does, did you do your job? You may have let your guy wide
open underneath, but if your teammate got burned deep,
should you get still credit for a lookoff?

I would simply count the number of times your guy catches
the disc, and divide it by the number of completed passes
the other team throws while you are on the field. This
would give you a good estimate of your "lookoff%" as a
defender, without requiring any subjective judgments being
made. If your guy is only catching 1 out of every 20
opponent passes, you are probably doing a pretty good job,
regardless of how often he is looked off.

Heck, the number of completed passes that other teams throw
(per possession), in and of iteslf, might be useful. Some
defenders may give up more goals, but they may always force
other teams to throw 2-3x as many passes before scoring. On
a windy day, that kind of defender might be ideal.


Ichiroo wrote on Wed, 01 February 2012 01:25

> For the O-line dump, the opposite would be true. How
> well he plays as a dump has a lot to do with how well he
> gets open on the first cut after he is engaged (probably
> moreso than the skill level of the player making the dump
> throw, which should have low variability at high levels),
> so a good stat for a handler might be 1st cut dump
> completion%. You might conclude that handler X might be a
> weaker player because of an inability to get open right
> away.

That's a useful stat in theory, but would it be all that
different from touches or touch%? If a good dump is getting
open on all his first cuts, he's going to get a lion's share
of the touches. If he's always getting looked off, those
numbers will drop.

[alansmith175]

I'd also like information on what calls players
make/contest.

[Parinella]

I'd like to see these stats tracked and published if only so
that people have a realistic idea of what to expect. Just
like TV golf fans think that pros make 90% of 5 foot putts
because that's all they show (it's more like 50%), the
average ultimate fan thinks good teams routinely turn it
over only 5 times a game or hucks are completed 80% of the
time or good players never drop the disc.

[tim]

I think we are looking at the stats issue the wrong way.
This is not basketball, the scoring throw itself does not
take any more skill than any other throw on the field. This
is a yardage/field position game. Why give the guy more
credit for the last 5 yards vs the first 60 yards. As a
captain/coach I like to see the guys that are eating up all
that yardage, not who makes the last throw.

Something like this would work for me. (yardage estimates of
course).

Deep-Reset -5yards+
Reset -5 to 5 yards
+ 5-20 yards
++ 20-40 yards
+++ 40 and up.

This allows us to see how players are doing at specific
tasks. Player A is 5-5 in the ++ area but 0-3 in +++. Or
maybe someone is having problems resetting the disc and is
3-7 on reset throws. The dynamic in ultimate is that all the
players on the field have certain roles. You want the stats
to tell you how effective they are in that role. Does
someone have way too many Deep-Resets, or are they always
reseting? Or maybe someone you want to be a deep thrower is
throwing really well in the the + range, but is only 2-2 on
the +++. You can work on getting him to be more aggressive
in his decision making. You would of course pair the
catching stats. So you could see who is the most effective
target at each level.

[Matt]

Bulb wrote on Wed, 01 February 2012 09:07

I would have to agree with pretty much everything. I think
lookoff% would be easy enough to calculate, even with
multiple cuts at once, but overall touch% would be very
similar and much easier to figure out. The only question
would be, if a team is picking on a mismatch of the D's
player 1, does that make player 2, 3, or 4 strong defenders?
Probably only relative to player 1, whom is probably marking
their best cutter (as you mentioned, the best cutter
porbably gets more discs thrown to him).

Team stats would be great as well, although I suppose that
is another story altogether.

[Bulb]

Ichiroo wrote on Thu, 02 February 2012 02:23

> The only question would be, if a team is picking on a
> mismatch of the D's player 1, does that make player 2, 3,
> or 4 strong defenders? Probably only relative to player 1,
> whom is probably marking their best cutter (as you
> mentioned, the best cutter porbably gets more discs thrown
> to him).

Exactly. And after a few games, you might look at the
numbers and realize the guy you THOUGHT was your top
defender (because of some impressive but ultimately sporadic
defensive plays) might do better to cover other teams' #3 or
#4 cutters, where he'll get thrown to less often but still
have opportunities to make poaching D's.

oldstyle wrote on Wed, 01 February 2012 15:39

> I think we are looking at the stats issue the wrong way.
> This is not basketball, the scoring throw itself does not
> take any more skill than any other throw on the field.
> This is a yardage/field position game. Why give the guy
> more credit for the last 5 yards vs the first 60 yards. As
> a captain/coach I like to see the guys that are eating up
> all that yardage, not who makes the last throw.

Eh, I see your point, but I disagree with the last
statement. It *is* important to know who is giving your
team the most yards per throw. But completing a 5 yard
throw for a goal is, on average, more difficult to complete
a 5 yard throw in your own end zone, merely as a function of
how the defense plays at those spots of the field. So I
think both are important.

As the OP suggested, we should identify a short list of
stats that spectators will easily relate to. We can apply
some sabermetrics and glean dozens of different stats that
are useful for a coaching staff, but publishing two pages of
data about each player is only going to confuse the
spectators.

oldstyle wrote on Wed, 01 February 2012 15:39

> Something like this would work for me. (yardage
> estimates of course).
>
> Deep-Reset -5yards+
> Reset -5 to 5 yards
> + 5-20 yards
> ++ 20-40 yards
> +++ 40 and up.
>
> This allows us to see how players are doing at specific
> tasks. Player A is 5-5 in the ++ area but 0-3 in +++. Or
> maybe someone is having problems resetting the disc and is
> 3-7 on reset throws. The dynamic in ultimate is that all
> the players on the field have certain roles. You want the
> stats to tell you how effective they are in that role.
> Does someone have way too many Deep-Resets, or are they
> always reseting? Or maybe someone you want to be a deep
> thrower is throwing really well in the the + range, but is
> only 2-2 on the +++. You can work on getting him to be
> more aggressive in his decision making. You would of
> course pair the catching stats. So you could see who is
> the most effective target at each level.

I do think this is a cool idea for a coaching staff. Could
be VERY useful, kinda how quarterbacks have their throws
broken down by yardage, left/middle/right, and in/out of
pocket. But this is the kind of stuff that only veteran
PLAYERS will readily understand during the first AUDL
season.

[anakin gerics]

~~~~~~~~~~~~~

---so....two people are going to get credit for the goal?
i mean...when you catch a goal, it's a goal.
when you throw a goal...it's an assist.

an assist in ultimate is exactly like an assist in basketball or
soccer.

is someone suggesting that there be an assist(a pass to the person who
throws the goal)
AND another category of 'thrown goals'?
AND caught goals?

an assist.....is the throw that leads to a goal....the pass the
results in the goal being scored.

[Matt]

anakin gerics wrote on Thu, 02 February 2012 08:36

Yes, two people would get credit for the goal, similar to
football where both the QB and WR are given credit for a
touchdown. No other sport that I can think of, including
basketball and soccer, involve two players for a score.

Just because ultimate has been calling the thrower of the
goal the one who gets the assist in the past does not mean
it is the best way to do it, especially when the more
impressive/important throw may very well be the one that set
up the player who threw for the goal.

[tim]

But this just adds more confusion to your stats. You say
that the important throw may very well be the one before it,
but it might not be either. Now we have a dump-swing-score
where the dump gets the assist? You aren't really fixing the
problem.

Last year I actually considered having someone keep stats
with + and - only, and it was their discretion for what was
a + or -, but for instance I didn't think that you should
get a + for a score (throwing or passing), unless you earned
it. If someone gets a D near the endzone and you run to the
disc and basically hand it off to another player for an easy
score, frankly as a coach/captain I don't really need that
recorded....

Of course if we are talking about for the fans, it's pretty
much throws for scores and catching goals...

[Reggie Fanelli]

> Yes, two people would get credit for the goal, similar to
> football where both the QB and WR are given credit for a
> touchdown.

---HUH?
the QB gets credit for a 'touchdown pass", yes....but the center
doesn't get an 'assist' for snapping the ball to him.
in football...that 'touchdown pass' is the exact same thing as an
assist in hockey or soccer or whatever other sport has assists.

HEY!....chalk up an assist for the center for snapping the ball!
nope.....just ain't right, huh?
~~~~~~~~~~~~~~~~

No other sport that I can think of, including
> basketball and soccer, involve two players for a score.

---RIGHT...because only one person scores the goal or the basket or
the point or whatever.
same for ultimate...one person scores the goal.
there person that threw it earns the assist.
if you want to change the word 'assist' to 'goal thrown'......it'd be
weird....and unnecessary....and probably just another 'made up way' to
pretend that ultimate is different from other sports.

certainly....because we're all geniuses here....no one is considering
having and 'assist', a 'goal thrown' AND a 'goal caught'......RIGHT?!
~~~~~~~~~~~~~

> Just because ultimate has been calling the thrower of the
> goal the one who gets the assist in the past does not mean
> it is the best way to do it,

---i'm gonna have to CALL!
"JUST" because we have been......don't make it WRONG!
.....because......the thrower of a goal......IS earning the assist!
~~~~~~~~~~~

especially when the more
> impressive/important throw may very well be the one that set
> up the player who threw for the goal.

---"may very well" have been the previous pass?
that's goofy.

the first cut MIGHT have been the most impressive from 70 yards away
for a 8 yard gain....and then there might have been 13 lame cuts and
passes and miscues and mis-reads and bobbles and close calls....and
finally a weak catch for a goal....but we ain't givin' an assist to
that sweet cut 70 yards away!

"may very well" have been the previous pass......that's.......WRONG.

the assist...is the thrown goal.

don't try to make ultimate different........it's already got the disc
and.....more.

previous pass?...the assist?....INDEED!

[Reggie Fanelli]

> Last year I actually considered having someone keep stats
> with + and - only,

~~~~~~~~~

---i thought + and - were.........
a + when your team scored when you were on the field
and a - was when your team got scored on when you were on the
field....

that way...you could say, yo, i was PLUS 12 when IIIIIII was on the
field!

but if yours is different...i can get down with that too.

BUT....an assist for some throw before the REAL assist.....is wrong.

[Mark]

Thanks for all your inputs, everyone. I'm not going to
respond to every post individually, but know that I've read
and considered everything you wrote here. As Bulb mentioned,
the original intent for this thread was just to get a
glimpse of the most interesting stats for spectators, and I
believe you have all addressed this very well.

As this thing evolves and we generate some interesting data,
we'll be looking to do much more with the stats. It'd be
great if we could implement some of the advanced metrics
that you all have proposed here. When this time comes, we'll
revisit this in even more detail.

As Ichiroo mentioned, we will be also doing team stats. This
is a topic that deserves it's own thread, and I'll probably
start one of those sometime soon as well.

If any of you have more things you'd like to tell me or want
to get more deeply involved in this project, email me at
mailto:ma...@leaguevine.com. There are plenty of useful roles
you can play even if you don't have technical skills. Thanks
again!

[dan nonjax]

If I was a coach, the one stat I would most want is line
performance that tells me how each grouping of 7 players
performs.

-If they are on D, do they force a turn?
-If they are on O, do they turn it over?
-If they turn it over, do they get it back?

And then, the big questions:

-How many points do they put on the board as an O line?
-How many points do they put on the board as a D line?

NBA teams use this soft of stuff a lot.

http://www.nba.com/2011/news/features/john_schuhmann/04/15/east-lineups/index.html

Over one game, wouldn't mean that much, but over a season
would be very useful in spotting player trends. If all the
lineups featuring players A-D score a lot, might tell you
something. If every grouping that has player K in it sucks,
that might also tell you something.

[anakin gerics]

> Our situation is more like football, where the qb throws to
> a receive in the endzone, and if the receiver catches the
> pass, both players are credited with a touchdown.

~~~~~~~

---whoa...wrong.
the QB ISN'T credited WITH a TD.
they are credited with a TD Pass.......which is the ASSIST.
the receiver is credited with a TD.
huge difference, that i'm surprised genius ultimate players have been
failing, in this thread to realize.

BOTH players ARE NOT credited WITH a touchdown.
the receiver is credited with a TD...and the QB is credited with a TD
Pass....an assist.

come on....i thought we ultimate players were smarter than this simple
error that ya'll are making.

[anakin gerics]

> Considering this, I think recording second assist rewards
> good decision making around the goal.

~~~~~~~~

---ya'll just want to make up 'assist' categories that don't exist in
some weird effort to make ultimate something it's not.

[anakin gerics]

~~~~~~~~~~~~~~~~~

---WHOA.....huh?
i mean....i WANT to see your opinion...and i can ALMOST agree....
sure, if you want to call ultimate's assist a 'goal thrown' sure, go
ahead.....

BUT.....you CAN'T THEN go....and make up some earlier pass an
'assist'.

if you're going to use a different term....then you have to DROP
COMPLETELY the term 'assist'

i mean....an assist...is something that HELPS the goal be scored.
the way you guys want to change the term 'assist'.....you're going to
have to start giving an "assist" to EVERY throw PRIOR to the thrown
goal.
17 pass......hey, that's 16 assists and one thrown goal AND a goal.

come on smart ultimate players.....use your brain!

[anakin gerics]

> My 10 stats:
>
> Points Played
> Goals Thrown
> Goals Caught
> Drops
> Throwaways
> Assists
> Touches (possessions)
> Blocks (includes d's handblocks, callahans)
> Fouls
> Something similar to QB rating (i guess this would be
> adjusted +/-)

~~~~~~~~~

---these are alright....except you accidentally listed assist
twice.....
once as 'assist' and another as 'goals thrown'

they're the same thing.

[Reggie Fanelli]

> Drops?  Seriously?  Are any of the top players going to have
> more than 1 or 2 drops over the entire season?
~~~~~~~~


--drops? heck yes drops should be a stat.
are any of the top players going to have more than 1 or 2 drops during
a season?
we'd know with this stat being kept.

it's a TURNOVER stat. a lot of team sports keep turnover stats.
throw aways and drops.....both a pretty important and easy to track.

[Reggie Fanelli]

> I see that Bulb and I think alike.  Does that mean I have a
> great mind?
~~~~~~

--possibly quite the contrary

[Pat Kelsey]

Have you guys ever thought of making fantasy software? I can
just see legions of Fantasy Frisbee players installing that
on their phones or playing on your website.

[Mark]

+PK+ wrote on Mon, 06 February 2012 09:08

> Have you guys ever thought of making fantasy software? I
> can just see legions of Fantasy Frisbee players installing
> that on their phones or playing on your website.

Yep, we've thought a bunch about this, and think it could be
amazing. Assuming people latch onto stats and we see users
entering stats for a sufficient number of games within a
single league/division, we'll very likely build out a "real"
fantasy ultimate system. And by real, I mean not a pick-em
system but a full draft league like you see on yahoo or
espn.