Moderately useful AI drafting info

[mhass]

I set out to prove something to myself and really didn't accomplish it, but there is some mildly interesting stuff here. It's a chart of the average adjusted bureau rating (from the html) by position and round. It covers several years of data though I'd like to continue to modify it with more years. I ran the seasons for these data with as little input of my own as possible (scouts ran draft, managed roster, I didn't do anything during free agency).

A few things I found interesting:
1. You shouldn't draft a QB in round 2 - the AI waits until 4.
2. A FL is more valuable to the AI than a SE.
3. The AI goes nuts for TE's after round 1.
4. What I always suspected is true, K & P only go in the middle three rounds.

Pos	1	2	3	4	5	6	7
C		5.2	4.7	4.3	4.5
FB		4.7	4.5	3.9		3.7	3.5
FL	5.9	5.2	4.7		4.6	4.5	4.4
FS	5.4	5.1	4.8	4.5	4.4	4.2	4.1
K			3.9	3.6	3.5
LCB	7.0	5.5	4.9	4.8	4.6	4.3	4.2
LDE	6.3	5.7	5.6	5.2	5.1	4.7	4.6
LDT	6.6	5.9	5.4	5.2	5.0	4.9	4.5
LG	6.0	5.0	4.8	4.8	4.5	4.3	4.1
LT	7.1	5.7	5.4	5.3	5.0	4.7	4.8
MLB	5.6	5.3	5.0	4.8	4.6	4.3	4.1
NT	7.1	5.8	5.5	5.2	4.9	4.8	4.6
P			3.6	3.4	3.2		3.0
QB	6.9			5.6	5.5	5.1	5.0
RB	6.3	5.5	5.2	5.1		4.9	4.8
RCB	5.6	5.3	4.9	4.8	4.4	4.5	4.3
RDE	7.4	5.6	5.6	5.2	4.8	4.8	4.6
RDT	7.1	6.2	5.5	5.2	5.0	4.9	4.7
RG	5.7	5.1	4.7	4.7	4.4	4.2	4.2
RT	6.9	5.7	5.5	5.3	5.0	4.8	4.8
SE	6.4	5.3	4.9		4.7	4.5	4.5
SILB		5.2		4.7	4.4		4.3
SLB	5.8	5.0	4.9	4.5	4.4	4.2	4.1
SS		4.9	4.6	4.5		4.2	4.1
TE	5.8	4.6	4.4	4.3	4.0	4.1	4.0
WILB			4.9		4.4	4.4	4.1
WLB	5.4	5.0	4.9	4.5	4.4	4.2	4.0
Grand Total	6.4	5.4	4.9	4.8	4.7	4.5	4.4

[nilodor]

Just curious how many drafts this is over.

[dubb93]

Originally Posted by mhass

2. A FL is more valuable to the AI than a SE.

I understand you are seeing this in your results, but from a game perspective I don't understand it. It is farely simple to change an SE to FL and I've seen the AI do it several times. I don't understand why the AI wouldn't take the more talented SE and move him to FL if they were presented with that option.

[Narcizo]

The gap between LCB and RCB in the first round is far greater than that between FL and SE. I attribute both to limited data. Interesting study though, and it will be more interesting when there's more data put in there. I've noticed the deal with QBs though.

[mhass]

Originally Posted by nilodor

Just curious how many drafts this is over.

It's either five or six seasons right now. I set out to do 10 so I knew when not to "reach". Beyond that I'm not sure how much it will change.

[mhass]

I've gone through 11 drafts now and not much is changing any more so I'll quit. But the chart has become useful to me to know how to gauge the drafts. It's as close to a mock draft as I get in SP since I can tell what the AI thinks is "value" at each round.

A couple holes filled in but that early gap is still there for QB. That's really more of an exploit than useful information.

Pos	1	2	3	4	5	6	7
C	5.7	5.3	4.8	4.3	4.4	4.0	4.0
FB		4.7	4.5	3.9	3.9	3.7	3.6
FL	6.3	5.2	4.9	4.8	4.6	4.6	4.5
FS	5.6	5.1	4.7	4.5	4.4	4.2	4.1
K			4.1	3.7	3.6
LCB	6.7	5.3	4.8	4.8	4.6	4.4	4.3
LDE	6.6	5.7	5.5	5.3	5.0	4.8	4.7
LDT	7.0	5.8	5.4	5.2	5.0	4.8	4.6
LG	6.2	5.3	4.9	4.7	4.5	4.3	4.1
LT	6.7	5.7	5.5	5.3	5.1	4.8	4.7
MLB	5.7	5.2	5.0	4.8	4.5	4.3	4.2
NT	6.8	5.8	5.5	5.2	5.0	4.8	4.6
P			3.6	3.3	3.2		3.0
QB	6.6			5.6	5.5	5.2	5.0
RB	6.2	5.5	5.2	5.1	5.0	5.0	4.9
RCB	6.1	5.3	4.9	4.8	4.5	4.5	4.3
RDE	6.9	5.7	5.4	5.3	5.1	4.8	4.6
RDT	6.8	5.9	5.5	5.2	5.0	4.9	4.6
RG	5.9	5.3	4.7	4.6	4.4	4.2	4.2
RT	6.9	5.8	5.5	5.2	5.1	4.8	4.7
SE	6.2	5.4	4.8	4.7	4.7	4.6	4.5
SILB	5.3	5.1	4.8	4.7	4.6	4.3	4.3
SLB	5.9	5.3	4.9	4.6	4.4	4.3	4.1
SS	5.7	5.0	4.6	4.4	4.3	4.3	4.1
TE	5.5	4.9	4.4	4.3	4.0	4.2	4.0
WILB		5.0	4.9		4.4	4.4	4.1
WLB	5.8	5.1	4.9	4.6	4.5	4.3	4.1
Grand Total	6.4	5.4	4.9	4.6	4.7	4.6	4.4

[GMO]

I want to understand this better.

So the AI, in the 1st round of the draft, picks (on average) LDTs of 7.0 while the average for SILBs is 5.3. Therefore the AI equates a 7.0 LDT to a 5.3 SILB, in terms of what the AI wants or needs in the 1st round?

So maybe the AI desperately needs SILBs so it chooses the highest ranked SILB, but there are many good LDTs and there's not a great, urgent need for them so the AI can be more choosy with LDTs?

[Ben E Lou]

Originally Posted by GMO

I want to understand this better.

So the AI, in the 1st round of the draft, picks (on average) LDTs of 7.0 while the average for SILBs is 5.3. Therefore the AI equates a 7.0 LDT to a 5.3 SILB, in terms of what the AI wants or needs in the 1st round?

So maybe the AI desperately needs SILBs so it chooses the highest ranked SILB, but there are many good LDTs and there's not a great, urgent need for them so the AI can be more choosy with LDTs?

Keep in mind that he's using the adjusted ratings...

[mhass]

Originally Posted by SkyDog

Keep in mind that he's using the adjusted ratings...

Exactly. I don't think an SILB ever gets as highly rated as an LDT (or LT or QB or, or, or) so we'd expect that position to be lower on the chart. Really what I was trying to establish here was where the AI sees breaks. So is the 5.3 (adjusted) rated SILB better or worse in the eyes of the AI than a 5.3 (adjusted) rated MLB? By the chart we see that a 5.3 rated SILB will likely go before (round 1) a 5.3 rated MLB (round 2). A 5.3 DE goes much later because of the adjustment in ratings. A 5.3 QB might be around until Round 7. Lots of players know this by experience. I had to make the chart.

[GMO]

Originally Posted by mhass

Exactly. I don't think an SILB ever gets as highly rated as an LDT (or LT or QB or, or, or) so we'd expect that position to be lower on the chart. Really what I was trying to establish here was where the AI sees breaks. So is the 5.3 (adjusted) rated SILB better or worse in the eyes of the AI than a 5.3 (adjusted) rated MLB? By the chart we see that a 5.3 rated SILB will likely go before (round 1) a 5.3 rated MLB (round 2). A 5.3 DE goes much later because of the adjustment in ratings. A 5.3 QB might be around until Round 7. Lots of players know this by experience. I had to make the chart.

Maybe I don't understand adjusted ratings but aren't they used to equalize the ratings, to give some idea of where they will be picked in the draft.

So, everything being equal, the highest adjusted rating players will be picked first in the draft, then the lower rated players.

So if a 5.3 adjusted rated player (no matter what position) is picked before a 7.0 adjusted rated player, then something different is happening?

[dj_morton]

nice study

[dbd1963]

Originally Posted by GMO

Maybe I don't understand adjusted ratings but aren't they used to equalize the ratings, to give some idea of where they will be picked in the draft.

So, everything being equal, the highest adjusted rating players will be picked first in the draft, then the lower rated players.

So if a 5.3 adjusted rated player (no matter what position) is picked before a 7.0 adjusted rated player, then something different is happening?

I'd like to 2nd that question.

[Narcizo]

Originally Posted by GMO

Maybe I don't understand adjusted ratings but aren't they used to equalize the ratings, to give some idea of where they will be picked in the draft.

So, everything being equal, the highest adjusted rating players will be picked first in the draft, then the lower rated players.

So if a 5.3 adjusted rated player (no matter what position) is picked before a 7.0 adjusted rated player, then something different is happening?

Adjested ratings are just a percentage applied based on the perceived value of the position in question. They don't have anything to do with the draft per se. Hence, after about 5 rounds or so the top of the draft will be stocked with quarterbacks and the odd running back. Their modifier means that they are ranked high but that doesn't mean teams are going to pick them ahead of other needs areas (particularly as 90% of quarterbacks by that stage are just chimeras who will never amount to their scouted ratings).

[DougWyatt]

Originally Posted by GMO

Maybe I don't understand adjusted ratings but aren't they used to equalize the ratings, to give some idea of where they will be picked in the draft.

I was under the assumption that the adjusted ratings just added position into the formula. So that a 6.3 QB has more value than a 6.3 Kicker or Fullback.

That should be the case, according to the help files.

Originally Posted by helpfile

Two grades are given to each player. These are assessed by a representative of the league scouting pool. The Grade is a raw score from 1.0 to 9.9, used to estimate the player's ability to compete in the league. It is based both on an impression of the player's raw skills and the annual skills competition. The Adjusted Grade is the Grade, adjusted by the player's position to give some idea of when the player should be taken in the draft.

[Synovia]

Originally Posted by dubb93

I understand you are seeing this in your results, but from a game perspective I don't understand it. It is farely simple to change an SE to FL and I've seen the AI do it several times. I don't understand why the AI wouldn't take the more talented SE and move him to FL if they were presented with that option.

I've noticed this in depth charting too. FL is more important than SE. If you ahve a 95 SE, and a 85 FL, the AI/reccomend will make the SE the starting flanker, and put the 85FL as the bench player.


It like SS over FS too. I think it just likes whatever comes first on the depth chart.

[Dutch]

Originally Posted by Synovia

I've noticed this in depth charting too. FL is more important than SE. If you ahve a 95 SE, and a 85 FL, the AI/reccomend will make the SE the starting flanker, and put the 85FL as the bench player.


It like SS over FS too. I think it just likes whatever comes first on the depth chart.

What is the positional bias set to?