A Realistic Model for Draft Pick Value in Madden

[DCEBB2001]

Quote: Originally Posted by Trojan Man

Love this data. It'd be great to have Madden's trade logic for players and picks governed by data gleaned from historical analysis. Obviously, once you enter your CFM, you're entering a self-contained, dynamic universe, but I'd love it if the general rules governing trading had some 10-15 year data from longitudinal studies behind it. Maybe it does already, IDK, but this data is great as a conversation starter.

And that is the idea - to start a conversation about such things. Personally, I know that when it comes to PICKING QBs in the draft, they are often very overvalued. In other words, they are often selected higher than what their grade indicates.

I would have expected the same thing to hold true when it comes to TRADING draft picks for QBs - you overpay to get one. However, that doesn't seem to be the case. As I pointed out in a previous post in this thread, of the 17 QBs included in this analysis, only 4 were deemed to be "overvalued" by the present model. That is very surprising.

Let me pull together some data for QBs selected in the draft to give you a better idea of where rookie QBs are drafted compared to where their value shows they SHOULD be drafted....I'll see what I can come up with for the sake of comparison.

[jhigginsluckow]

Just want to be clear - I wasn't attempting to criticize the potential of this sort of system. Actually, I was trying to defend it for the most part. But there are some things that you can't use statistics for, is all I'm saying. JJ Watt, as good as he is, is not the most valuable asset in the league, which seems to be the overall feel of the chart.

I think a system like this would work much better when making draft pick trades in Madden, more so than player trades. You'd have a much larger sample size to go off of, and you're still fixing a major flaw in the game.

Out of curiosity DCEBB2001, and maybe you've answered this in a different thread, what all goes into the values you have in the OP? Age, Years of experience, positional value, contract terms/length and other things like that are all factored into a team's decision in real life when making a trade. Can your models incorporate that in some way?

[ggsimmonds]

This doesn't account for other variables that go into trades.
Ngata was in the final year of his contract and the Ravens were unwilling to extend him. Other teams knew this so that brought his trade cost down.

That is one example but it happens frequently where a player is traded in part due to off field issues (contract, "cancer" status, disagreement with coaches, etc) that are not accounted for in either FBG or Madden.

When enough players are being traded for less than their value it brings the trade costs down for everyone.
The data is skewed.

[DCEBB2001]

Here is the draft data I spoke of earlier:

I pulled in every instance where a QB was drafted since 2005. Every draft class has every prospect ranked from worst to first. Their grade value is not necessary as only the selection slot is important. I will compare where the player was actually drafted as opposed to where the player's value suggested he should have been picked.

2005:
Smith (#4) drafted #1 overall
Rodgers (#5) drafted #24 overall
Campbell (#39) drafted #25 overall
Frye (#34) drafted #67 overall
Walter (#71) drafted #69 overall
Greene (#156) drafted #85 overall
Orton (#41) drafted #106 overall
LeFors (#188) drafted #121 overall
Orlovsky (#192) drafted #145 overall
McPherson (#107) drafted #152 overall
Anderson (#212) drafted #213 overall
Kilian (#478) drafted #229 overall
Cassel (#603) drafted #230 overall
Fitzpatrick (#244) drafted #250 overall

47 places higher than average


2006:
Young (#11) drafted #3 overall
Leinart (#4) drafted #10 overall
Cutler (#5) drafted #11 overall
Clemens (#126) drafted #49 overall
Jackson (#277) drafted #64 overall
Whitehurst (#153) drafted #81 overall
Croyle (#114) drafted #85 overall
Martin (#319) drafted #148 overall
Jacobs (#204) drafted #164 overall
McNeal (#223) drafted #193 overall
Gradkowski (#177) drafted #194 overall
Shockley (#293) drafted #223 overall

57 places higher than average


2007:
Russell (#2) drafted #1 overall
Quinn (#5) drafted #22 overall
Kolb (#105) drafted #36 overall
Beck (#87) drafted #40 overall
Stanton (#63) drafted #43 overall
Edwards (#75) drafted #92 overall
Stanback (#152) drafted #103 overall
Rowe (#259) drafted #151 overall
Smith (#123) drafted #174 overall
Palmer (#215) drafted #205 overall
Thigpen (#499) drafted #217 overall

45 places higher than average


2008:
Ryan (#7) drafted #3 overall
Flacco (#39) drafted #18 overall
Brohm (#19) drafted #56 overall
Henne (#38) drafted #57 overall
O'Connell (#142) drafted #94 overall
Booty (#98) drafted #137 overall
Dixon (#173) drafted #156 overall
Johnson (#84) drafted #160 overall
Ainge (#154) drafted #162 overall
Brennan (#219) drafted #186 overall
Woodson (#73) drafted #198 overall
Flynn (#263) drafted #209 overall
Brink (#341) drafted #223 overall

-1 places higher than average


2009:
Stafford (#5) drafted #1 overall
Sanchez (#6) drafted #5 overall
Freeman (#17) drafted #17 overall
White (#62) drafted #44 overall
McGee (#96) drafted #101 overall
Bomar (#90) drafted #151 overall
Davis (#156) drafted #171 overall
Brandstater (#199) drafted #174 overall
Teel (#271) drafted #178 overall
Null (#494) drafted #196 overall
Painter (#234) drafted #201 overall

36 places higher than average


2010:
Bradford (#2) drafted #1 overall
Tebow (#36) drafted #25 overall
Clausen (#12) drafted #48 overall
McCoy (#40) drafted #85 overall
Kafka (#196) drafted #122 overall
Skelton (#129) drafted #155 overall
Crompton (#118) drafted #168 overall
Smith (#437) drafted #176 overall
LeFevour (#100) drafted #181 overall
Pike (#69) drafted #204 overall
Brown (#151) drafted #209 overall
Canfield (#321) drafted #239 overall
Robinson (#221) drafted #250 overall

-2 places higher than average


2011:
Newton (#11) drafted #1 overall
Locker (#24) drafted #8 overall
Gabbert (#9) drafted #10 overall
Ponder (#32) drafted #12 overall
Dalton (#44) drafted #35 overall
Kaepernick (#62) drafted #36 overall
Mallett (#37) drafted #74 overall
Stanzi (#100) drafted #135 overall
Yates (#201) drafted #152 overall
Enderle (#244) drafted #160 overall
Taylor (#153) drafted #180 overall
McElroy (#165) drafted #208 overall

6 places higher than average


2012:
Luck (#1) drafted #1 overall
Griffin (#2) drafted #2 overall
Tannehill (#8) drafted #8 overall
Weeden (#44) drafted #22 overall
Osweiler (#83) drafted #57 overall
Wilson (#121) drafted #75 overall
Foles (#200) drafted #88 overall
Cousins (#65) drafted #102 overall
Lindley (#129) drafted #185 overall
Coleman (#159) drafted #243 overall
Harnish (#185) drafted #253 overall

-4 places higher than average


2013:
Manuel (#40) drafted #16 overall
Smith (#21) drafted #39 overall
Glennon (#77) drafted #73 overall
Barkley (#37) drafted #98 overall
Nassib (#41) drafted #110 overall
Wilson (#72) drafted #112 overall
Jones (#111) drafted #115 overall
Sorensen (#281) drafted #221 overall
Dysert (#142) drafted #234 overall
Daniels (#818) drafted #237 overall
Renfree (#276) drafted #249 overall

37 places higher than average


2014:
Bortles (#6) drafted #3 overall
Manziel (#7) drafted #22 overall
Bridgewater (#13) drafted #32 overall
Carr (#28) drafted #36 overall
Garoppolo (#47) drafted #62 overall
Thomas (#179) drafted #120 overall
Savage (#45) drafted #135 overall
Murray (#140) drafted #163 overall
McCarron (#102) drafted #164 overall
Mettenberger (#125) drafted #178 overall
Fales (#204) drafted #183 overall
Wenning (#232) drafted #194 overall
Boyd (#298) drafted #213 overall
Gilbert (#281) drafted #214 overall

-1 places higher than average


2015:
Winston (#2) drafted #1 overall
Mariota (#3) drafted #2 overall
Grayson (#95) drafted #75 overall
Mannion (#107) drafted #89 overall
Petty (#73) drafted #103 overall
Hundley (#63) drafted #147 overall
Siemian (#458) drafted #250 overall

19 places higher than average


Total Poplulation: average QB is selected 21 places higher than he should be.

As you can see, some years have QBs selected an average of a worse slot than they are ranked, but most years show that QBs are selected higher than where there overall rank. It seems that QBs are valued higher in the draft compared to their true value.

[DCEBB2001]

Quote: Originally Posted by jhigginsluckow

Just want to be clear - I wasn't attempting to criticize the potential of this sort of system. Actually, I was trying to defend it for the most part. But there are some things that you can't use statistics for, is all I'm saying. JJ Watt, as good as he is, is not the most valuable asset in the league, which seems to be the overall feel of the chart. I think a system like this would work much better when making draft pick trades in Madden, more so than player trades. You'd have a much larger sample size to go off of, and you're still fixing a major flaw in the game. Out of curiosity DCEBB2001, and maybe you've answered this in a different thread, what all goes into the values you have in the OP? Age, Years of experience, positional value, contract terms/length and other things like that are all factored into a team's decision in real life when making a trade. Can your models incorporate that in some way?

Correct, we can also do this for pick-to-pick trades as well. That would require another analysis. I wanted to demonstrate the more nebulous player-for-pick trades, however, in this examination.

The values are based solely on the scouts grades for each player at the time they are traded. I have historical scouting grades going back to 1996. I can basically look up the grade of the player to the day they are traded and compare their grade to the points they are traded for. This model does not incorporate any of that other stuff, just the value of the player vs. the value of the picks as explained in the original post.

[DCEBB2001]

Quote: Originally Posted by ggsimmonds

This doesn't account for other variables that go into trades. Ngata was in the final year of his contract and the Ravens were unwilling to extend him. Other teams knew this so that brought his trade cost down. That is one example but it happens frequently where a player is traded in part due to off field issues (contract, "cancer" status, disagreement with coaches, etc) that are not accounted for in either FBG or Madden. When enough players are being traded for less than their value it brings the trade costs down for everyone. The data is skewed.

The data is not skewed. It just solely incorporates graded player value vs. draft pick value. Nothing more, nothing less. I explained that in the original post, and want to reiterate that any absence of criteria are simply contributing to the "limitations" of the analysis.

I also want to note that FBG Ratings has absolutely nothing to do with this analysis so saying that other factors are " not accounted for in either FBG or Madden" is not a correct statement. This model is not for FBG, or derived from FBG, in any way. FBG is a system for rating Madden players. This analysis is completely separate from anything on the FBG site. Rather, it is presenting a discourse to discuss one way (of many, I am sure) whereas Madden can better model picks-for-players trades in their game.

[ggsimmonds]

Quote: Originally Posted by DCEBB2001

The data is not skewed. It just solely incorporates graded player value vs. draft pick value. Nothing more, nothing less. I explained that in the original post, and want to reiterate that any absence of criteria are simply contributing to the "limitations" of the analysis. I also want to note that FBG Ratings has absolutely nothing to do with this analysis so saying that other factors are " not accounted for in either FBG or Madden" is not a correct statement. This model is not for FBG, or derived from FBG, in any way. FBG is a system for rating Madden players. This analysis is completely separate from anything on the FBG site. Rather, it is presenting a discourse to discuss one way (of many, I am sure) whereas Madden can better model picks-for-players trades in their game.

Your analysis has omitted variable bias. There is no way around it. You are attempting to assign trade values while ignoring a variable that undoubtedly influences trade value. If I'm am being blunt, the analysis is useless.

My bad on the FBG remark, I used that as a shorthand in referencing the scouting data you used.

[DCEBB2001]

Quote: Originally Posted by ggsimmonds

Your analysis has omitted variable bias. There is no way around it. You are attempting to assign trade values while ignoring a variable that undoubtedly influences trade value. If I'm am being blunt, the analysis is useless.

The point of the analysis was not to check for other variables. I just wanted to examine a possible correlation between player grades and trade value via draft picks. The analysis showed that there is a correlation and suggests that such a correlation could be used to create a model for the game.

If you think my analysis is useless, then we will simply have to agree to disagree.