![[DEV] ISFL Forums](http://dev.sim-football.com/forums/images/emerald/logo.png "[DEV] ISFL Forums"){kind=logo}
+- [DEV] ISFL Forums (http://dev.sim-football.com/forums)
+-- Forum: Community (http://dev.sim-football.com/forums/forumdisplay.php?fid=5)
+--- Forum: Media (http://dev.sim-football.com/forums/forumdisplay.php?fid=37)
+---- Forum: Graded Statistical Analysis (http://dev.sim-football.com/forums/forumdisplay.php?fid=153)
+---- Thread: *Attempting to Assign Defensive Value (/showthread.php?tid=17490)
Pages:
1
2
*Attempting to Assign Defensive Value - HENDRIX 2.0 - 01-06-2020
In every sport statisticians, mathematicians, and analysts do their best to assign an overall number to a player that describes their performance. In basketball there's BPM and RPM, in football there are the ProFootballFocus grades and things like approximate value.
In this chart I made my own attempt at this. Using all the box score stats the sims gives us, I slammed them into my own formula while adjusting for positions and playing time. The Defensive value added (DVA) is on a per game basis.
It's not going to win a Nobel Prize but I think it did a decent job, the true standouts of the seasons are clearly at the top. It might underrate defensive lineman a bit, though I'm not sure if that's partly due to DL being a less active position than LB, CB, and S. After the chart, I make my own picks for the S19 defensive awards.
![[Image: 59N4pGB.jpg]](https://i.imgur.com/59N4pGB.jpg)
![[Image: FUGbU6X.jpg]](https://i.imgur.com/FUGbU6X.jpg)
![[Image: NaSmU0S.jpg]](https://i.imgur.com/NaSmU0S.jpg)
1st Team All-NSFL Defense
DE- G. Kuusela
DE- M. Spector
DT- D. Wizardry
DT- B. Faux
LB- L. Grievous
LB- M. Berry
LB- R. Vans
CB- D. Taylor
CB- D. Rector
S- R. D'Attoria V
S- D. Grithead
2nd Team All-NSFL Defense
DE- M. Virtanen
DE- F. Baldari
DT- R. Leaf Jr.
DT- F. Riposte
LB- S. Bakshi
LB- J. Slothface
LB- Q. Bode
CB- A. Asipi III
CB- S. Little IV
S- W. Green
S- M. Hockhertz
NSFL Defensive MVP
R. D'Attoria V
NSFL Defensive Rookie of the Year
M. Hockhertz
*Attempting to Assign Defensive Value - Modern_Duke - 01-06-2020
Wow. I looked into doing basically exactly this once but figured it was way too hard. Great stuff
*Attempting to Assign Defensive Value - MSombrero - 01-06-2020
This is awesome, I love it. Any chance you could share the formulas you used to come up with these values?
Do you think the main reason DL are so low is due to the lack of stats (no pressures, no qb hits etc.) rather than being a less active position?
*Attempting to Assign Defensive Value - HENDRIX 2.0 - 01-06-2020
(01-06-2020, 11:18 AM)MSombrero Wrote:This is awesome, I love it. Any chance you could share the formulas you used to come up with these values?
Do you think the main reason DL are so low is due to the lack of stats (no pressures, no qb hits etc.) rather than being a less active position?
Sorry, the formula is proprietary. With DL what I did do is attempt to adjust for run defense, for example a tackle after a 2 yard run should be worth more than a CB getting a tackle after getting burned for 30 yards. The QB pressures thought did come to mind, but I figured I weighed sacks enough for the DL. What you said I agree with, they add value that doesn't show up on the stat sheet, they help others succeed. Tough to quantify.
Thinking about it more the defensive formations GMs use could play a role in it. Defensive linemen in a 3-4 defense will definitely have less counting stats, and there's obviously only 3 of them.
*Attempting to Assign Defensive Value - MSombrero - 01-06-2020
(01-06-2020, 05:16 PM)HENDRIX 2.0 Wrote:Sorry, the formula is proprietary. With DL what I did do is attempt to adjust for run defense, for example a tackle after a 2 yard run should be worth more than a CB getting a tackle after getting burned for 30 yards. The QB pressures thought did come to mind, but I figured I weighed sacks enough for the DL. What you said I agree with, they add value that doesn't show up on the stat sheet, they help others succeed. Tough to quantify.
Thinking about it more the defensive formations GMs use could play a role in it. Defensive linemen in a 3-4 defense will definitely have less counting stats, and there's obviously only 3 of them.
No worries. How are you collecting the data for this? I've been looking at the csv files but I can't see anything which would show you how many yards are made before a tackle. Are you scraping the log file?
I'm also guessing a lack of negative statistics for defensive backs helps bump them up. We have missed tackles but I think that's more likely to effect lineman and linebackers than DB's, and we're missing things like passes allowed into their coverage.
*Attempting to Assign Defensive Value - caltroit_red_flames - 01-06-2020
Now do OL
*Attempting to Assign Defensive Value - Modern_Duke - 01-06-2020
(01-06-2020, 12:31 PM)MSombrero Wrote:No worries. How are you collecting the data for this? I've been looking at the csv files but I can't see anything which would show you how many yards are made before a tackle. Are you scraping the log file?
I'm also guessing a lack of negative statistics for defensive backs helps bump them up. We have missed tackles but I think that's more likely to effect lineman and linebackers than DB's, and we're missing things like passes allowed into their coverage.
I’ve long said tackles are actually a negative stat, since it means you allowed a reception. But that would really only apply to CBs playing on the outside, not so much NBs and safeties
*Attempting to Assign Defensive Value - White Cornerback - 01-06-2020
wow it's almost like S @ LB is OP or something
regardless, good work, it's been a long time since people went this direction with media and it's great to see you back
*Attempting to Assign Defensive Value - iseedoug - 01-06-2020
Love this. Wish you would have looked at it when Ben Horne was actually alive. Mabb and I were trying to do this solely for CBs, but just got too busy.
*Attempting to Assign Defensive Value - timeconsumer - 01-06-2020
Something you might find interesting is a while back @speculadora and I were comparing different defensive statistics (sacks, picks, tackles, etc) to win rates across large sample sizes of test sims. What we discovered is that of the individual stats the only one that had a significant effect on improving win rate was sacks.