no content
AdBlocker active?
It seems you are using software to block advertisements. You could help us if you could switch it off when visiting redzoneaction.org.
The reason is very simple: Advertisements help us running the site, to offer you the game in a good quality for free. So if you like the game, please support us by purchasing a Supporter Account or disabling the AdBlocker on this site.
Thank you very much!
| Main / Discussions / DVOA - Defensive Value Over Average Search Forum | |
| Navigation: |< 2 > >| | |
| Poster | Message |
|
wiesengrund
|
posted: 2012-10-13 22:48:10 (ID: 61534) Report Abuse |
|
I’ve long been tinkering with a way to compute a rating for defensive performance. It’s a bit of a hobby of mine. I’ve made some fine, usable formulas that grade the performance of a defensive player over a longer period, basically in search for a way to find out what defensive player of mine played best in any particular season. Now, the main aspects I was looking for were:
- As a rating of players, it should somehow relate to other performances, i.e. you should be able to compare two players. - As a comparison, it should disregard one-off performances in favor of bigger sample size. I.e. just because a player had one great game he should not be rated above another player who played worse in that game but better in 10 others. - The big sample size practically means that the rating should be based on averages, i.e. it should incorporate a per play metric that works off percentages, not of absolute values. (Notice that that also means we can grade players with different numbers/snaps and games played.) - Averages alone are not enough, though, since different positions have vastly different usual performance rates. For instance, you can never compare the missed tackle rate of a DT to one of a SF simply because of the schematic differences between those positions. That means that there has to be a relation to a baseline, an expected performance of a player at a certain position. Now, I have made what could be called DVOA Version 0.1. It stands for Defensive Value Over Average and it is of course a direct rip off of Football Outsider's DVOA Metric, because it shares those same values outlined above. (Their “D” stands for defense-adjusted, since it was primarily an offense-based metric, but that’s not our concern here.) DVOA 0.1 basically consists of two parts: A base play metric grading tackle efficiency in relation to the importance of the player, and a far less important big play metric. Both values work with an expected model of the position and seasonal averages. DVOA 0.1 is fairly okay grading Defenders relative to their positions, but such a rating is only as good as its baseline. DBs are no problem. I can easily compile a huge database of DB seasons to see how they typically perform, since they are easily accessible via the stats pages. But the Front 7 is a bit trickier, since DVOA distinguishes between DTs, DEs, MLBs and OLBs. Now, I do have my own players as a sample, but I guess the rating would be much better if I had access to more. So, my question is, if someone here would be willing to “share” the stats of their defensive players in order to help me generate better baselines for DVOA. I was thinking of including seasons of front 7 players of the last two seasons, only from the 1.1s (since the level of competition is still very much different in the 2.x leagues) and only from league games without playoffs (since the same argument applies also to most Supercup divisions). Basically I can find the stats myself, what I don’t know is, if any particular DL was playing mostly as a DE or as a DT. Or if a LB was in the middle or outside. So if anyone would be willing to tell me their stats with a position, that would be very helpful. Please keep in mind, that the more “typical” a player-season is, the better it is suited for the baseline. So if a DL played 10 games as a DE and 6 games as a DT he is far less valuable as a baseline for either position than guys who really played 16 games in one place. (Also keep in mind, that backups are totally fine too, if they stayed in their position for a whole season. RZA is still full of substitutions, so it is merely correct to represent backups in the baseline, since they really do constitute a part of what their position brings to the table in terms of defensive performance.) The long term goal would of course be to actually get some predictive value out of DVOA, but we are still a long way from there. First steps first. 
Any help is very much appreciated. Once there is a stable version of DVOA working – thanks to your input – it will of course be posted and made public so that everyone can use it to grade their players. |
|
| Quote Reply Edit | |
|
devastations
|
posted: 2012-10-13 22:53:52 (ID: 61535) Report Abuse |
|
Sounds like a very ambitious plan, good luck with it, unfortunately I'm still in 2.2, otherwise I would certainly have sent you my stats.
|
|
| Quote Reply Edit | |
|
canonico
|
posted: 2012-10-13 22:56:38 (ID: 61536) Report Abuse |
|
Sounds like a cool idea, and I would gladly share my players (that played in the same position) stats with you (and everyone else for that matter), but since my team is still in 2.1 division and that's not what you are looking for, it wouldn't help you out.
But surely others will, since that sounds like a cool idea. I am more familiarized with baseball sabermetics (but only to a certain point), but it seems that that might be somewhat similar. |
|
| Quote Reply Edit | |
|
espen
|
posted: 2012-10-13 22:58:13 (ID: 61537) Report Abuse |
|
Great plan! Another div.2 team who'll be happy to help in the future.
|
|
| Quote Reply Edit | |
|
wiesengrund
|
posted: 2012-10-13 23:01:00 (ID: 61538) Report Abuse |
|
devastations wrote:
Sounds like a very ambitious plan, good luck with it, unfortunately I'm still in 2.2, otherwise I would certainly have sent you my stats. Thanks, I was thinking of including the 2.x players, but I saw some stats from, for instance, Draft-DEs that just blow up all reasonable averages because of the level of O-Lines and RBs that they face over the course of a season. RZA still tilts to the defensive side, that's why I reasoned to keep the rating based on 1.1 averages since the level of offensive performance is at least somewhat competitive... |
|
| Quote Reply Edit | |
|
wiesengrund
|
posted: 2012-10-13 23:02:25 (ID: 61539) Report Abuse |
|
canonico wrote:But surely others will, since that sounds like a cool idea. I am more familiarized with baseball sabermetics (but only to a certain point), but it seems that that might be somewhat similar.
Exactly, this DVOA, just like the original one, is of course greatly inspired by this school of thought. |
|
| Quote Reply Edit | |
|
scottishbronco
|
posted: 2012-10-14 10:54:35 (ID: 61570) Report Abuse |
|
we have a small depth chart here in Oban Bay and we would love to help with this.
Any and all positions you need just drop me a PM and ill forward you all the details you need. This is a great idea  |
|
| Quote Reply Edit | |
|
TombKing
|
posted: 2012-10-14 13:51:39 (ID: 61589) Report Abuse |
|
The Sinna Ladd would also like to contribute to this research.
PM me for what you need. |
|
| Quote Reply Edit | |
|
bwadders76
|
posted: 2012-10-14 17:11:10 (ID: 61598) Report Abuse |
|
I sent my details over last night and had my feedback this morning. Amazing insight into some of my players. One of those I thought was amongst my bottom 3 is actually one of my best DLs.
Excellent work there |
|
| Quote Reply Edit | |
|
phareux
|
posted: 2012-10-14 17:17:45 (ID: 61600) Report Abuse |
|
let me know when you would like to try the 2.2 teams this sounds interesting
 |
|
| Quote Reply Edit | |
| reply Mark this thread unread | |
| Navigation: |< 2 > >| | |
| Main / Discussions / DVOA - Defensive Value Over Average | |
