## linear weights

Expand Messages
• i read as much of the group s discussion on rebounds, assists, missed shots, etc. as i could digest this evening. in general i m not immediately inclined to
Message 1 of 1 , Mar 6, 2004
i read as much of the group's discussion on rebounds, assists, missed
shots, etc. as i could digest this evening.

in general i'm not immediately inclined to change my weights given
the way i look at possessions. if you look at possessions
differently, then it is a different story from here on. i define a
shot in a possession as being the last act of that possession, so
points - missed shots * value of possession seems right- for that
part of the story. rebounding and scoring after rebounding need not
be in the valuation of the shooting because in a complete set of
linear weights it is separately and completely accounted for in the
valuation of rebounding acts(every rebound is a possession and the
value of that possession at the time that that play ended- the
rebound- is the same). the point value given thru the rebounding
linear weight score fully adds back all the points you got off
scoring from offensive rebounds (and of course this is true for
defensive rebounds as well) and in effect it corrects for what seems
like perhaps an inappropriate deduction in shooting for missed shots.
(i am sure you could do this clearer in one step as some posts here
discuss if you consider shot/rebound as being part of the same
posession which i dont, but i guess i didnt see the need to change
this somewhat linked way of thinking. between the two possession
acts and scores i believe i end up with the correct net value of the
two possession acts- shot then some % rebounded). seeing a rebound as
starting a new possession, not continuing an existing one, eliminates
the urge to give offensive and defensive rebounding different scores
based on expected future action. all future activity is accounted
for in separate possessions not the previous one. to add the value
of future actions / other categories to any one linear weight of a
possession ending/creating act is to start double counting unless you
do it for every category then you are just creating extra work (and
would need to fully consider all possible future possession acts not
just the most desired-scoring). except i then modify from that
principle in a few cases where you make a strong case that sequence
of possession actions greatly affect their point value- such as
giving more value to steals because they lead to better FG% on layups
or a block being discounted to account for a % going out of bounds or
frequently back to the other team. you can also consciously go in
and adjust linear weights to try to take care of some intangibles not
fully captured by the set of scorable actions- like giving some extra
credit for assists to try to capture some of the assumed value of non-
assist generating passing activity by assist generators. this helps
offset to a small degree my low comparatively weighting of assists as
derived from the modest extra FG% of assisted shots over nonassisted
shots as several here also have described here. i think a high
percentage of NBA credited assists really add nothing to FG%, while a
few greatly do. on average, i give .5 per assists which i recognize
is well below most other systems (except my old files suggest dean
oliver used .37 a long time ago. what is it now? the valuation for
fouls i used is a little more detailed in determination than simply
giving it a -1.

but i'll stop for now and read what might come in response later.
Your message has been successfully submitted and would be delivered to recipients shortly.