GUESS by the numbers Topic

Posted by gt_deuce on 5/6/2013 6:46:00 PM (view original):
Posted by hcazualcc on 5/5/2013 2:18:00 PM (view original):
gtdeuce - it might not be fair to say the best player at one position is AS good as the best player at another...but can't we come up with a baseline for each attrubute at each position and call that a guess score of 70 for arguments sake?

e.g. - a good qb recruit and guess score of 70:
Ath - 50
Spd - 50
Str - 55 (weighted more)
GI - 60 (" ")
Tech - 60 (" ")

fluctuations above or below the baseline would result in a change in score + or -

just a different way to skin the cat

That's not a bad idea. It also has the effect of "normalizing" across recruiting classes. Which would allow Bob to add - if he so chose - cool new features like Top N of all-time recruits (per position).
Would have to define this for all four divisions, plus come up with the weights for +/- each stat for each position for each division.

Not saying it can't be done, but folks tend to focus on the Division they're playing and ignore the rest.
5/8/2013 11:07 AM (edited)
OK , so working on the idea of recruit class scoring based on relative ranking within a position.

Let's say we give the highest-rated recruit at each position 100 points of Relative Recruit Value (RRV).  And we'll give the lowest-rated recruit 0 points. Each player in between gets a percentage RRV based on their rating  (  (Rating - Worst Rating) / (Best Rating - Worst Rating) ).  For example, DB1 (the best) has a rating of 74.89, while DB254 (the lowest in my sample set) has a rating of 50.25.  DB2, with a rating of 74.3, gets an RRV of (74.3-50.25) / (74.89-50.25) = 24.05/24.64 = 97.6055

Your recruiting class score would be the total of your RRV / # of recruits.  There could be some anomalies in extreme cases - if you needed two recruits and signed the best punter and best kicker you would have a class RRV of 100, but that's obviously not the best recruiting class.
5/8/2013 11:29 AM
That sounds pretty cool to me!

5/9/2013 11:53 AM
Posted by bhazlewood on 5/8/2013 11:29:00 AM (view original):
OK , so working on the idea of recruit class scoring based on relative ranking within a position.

Let's say we give the highest-rated recruit at each position 100 points of Relative Recruit Value (RRV).  And we'll give the lowest-rated recruit 0 points. Each player in between gets a percentage RRV based on their rating  (  (Rating - Worst Rating) / (Best Rating - Worst Rating) ).  For example, DB1 (the best) has a rating of 74.89, while DB254 (the lowest in my sample set) has a rating of 50.25.  DB2, with a rating of 74.3, gets an RRV of (74.3-50.25) / (74.89-50.25) = 24.05/24.64 = 97.6055

Your recruiting class score would be the total of your RRV / # of recruits.  There could be some anomalies in extreme cases - if you needed two recruits and signed the best punter and best kicker you would have a class RRV of 100, but that's obviously not the best recruiting class.
this almost seems too easy . we expect something more difficult for you.
5/10/2013 12:53 AM
me likey
5/10/2013 8:57 AM
bhazlewood:  First, thanks for putting together the GUESS ratings!  WIS definitely owes you one as I doubt I would have gotten hooked on GD without the guidance it gave me in player evaluation when I got started.   As for the current question, I would vote against normalization for team ratings.  Right now, the GUESS ratings are terrific for evaluating players.  I like that I can calculate the GUESS ratings of my players throughout the season and have it match your ratings on signing day.    I think the normalization of recruiting classes the way you describe would be good, as long as the player's un-normaliized GUESS scores are still listed in the roster section.  The one other change I would want is for positions like DT/DE where almost everyone is higher at DT than DE.  In that case, it would be nice to see all the players whose DT-DE value is less than the average listed as DEs. But, that is kind of nitpicky... and if I was going to be nitpicky, I'd point out that Montana State is the Bobcats, not the Northern Lights :)
5/10/2013 9:02 PM
This formula might make it a littler easier for you.

(1 - ((Rank -1) / (Number of Players - 1))) * 100

The best player would be 100, the average player would be 50 and the worst player would be 0.
5/12/2013 7:41 AM
Posted by mr_automatic on 5/12/2013 7:41:00 AM (view original):
This formula might make it a littler easier for you.

(1 - ((Rank -1) / (Number of Players - 1))) * 100

The best player would be 100, the average player would be 50 and the worst player would be 0.
That's another way of doing it, but I chose to use their actual rating instead of their rank so that players who score similarly get similar points.  In the example I gave, with three DB's, the second guy would get a 50 (if there were just those three) but he's clearly closer to the top than the bottom.
5/12/2013 10:11 AM
Posted by bhazlewood on 5/8/2013 11:29:00 AM (view original):
OK , so working on the idea of recruit class scoring based on relative ranking within a position.

Let's say we give the highest-rated recruit at each position 100 points of Relative Recruit Value (RRV).  And we'll give the lowest-rated recruit 0 points. Each player in between gets a percentage RRV based on their rating  (  (Rating - Worst Rating) / (Best Rating - Worst Rating) ).  For example, DB1 (the best) has a rating of 74.89, while DB254 (the lowest in my sample set) has a rating of 50.25.  DB2, with a rating of 74.3, gets an RRV of (74.3-50.25) / (74.89-50.25) = 24.05/24.64 = 97.6055

Your recruiting class score would be the total of your RRV / # of recruits.  There could be some anomalies in extreme cases - if you needed two recruits and signed the best punter and best kicker you would have a class RRV of 100, but that's obviously not the best recruiting class.
This seems like a great way to compare the players. The assumption of this comparison would be that as more teams chase the more popular positons such as OL, DL, DB the competiton will be higher and the best recruiting coaches will get the best players. Positions such as QB, K or P would be less competitive and easier to land the #1 or top players. So consider the following modifier:

Calculate the average number of recruits per position - (total recruits/number of positions). Divide the total recruits for each position by the average number of recruits per position. This would give a weight to the competition for those recruits - above 1 for the more recruited positions and below 1 for the least recruited positions - call this RPV - Recruit Position Value. Apply that to each recruits - RRV - Raw Recruit Value  (Relative Recruit Value above). So by combining the RRV and RPV, this would provide a score which would reflect the players relative value as a recruit based on player rating and player need - Team Recruit Value - TRV. Add the TRV values together and divide by recruits to get a Team Recruit Rating - TRR.

So an example: Coach ONE gets the #1 QB, #1 P and #1 K for a RRV of 300. Coach TWO gets the #1 OL, #1 DL and #1 DB for a RRV of 300. But since the number of QB, P and K recruited are all less than the average of all positions recruited their RPV is less than one, Coach ONE's TRR is now say 240. Coach TWO has the #1's for very competitive recruits. The total recruits for the positions he recruited is above average, so the RPV is above one and Coach TWO's team may score 360.

This wouldn't change the individual player scores, as they would be 100 for the top player and descend, only to the team ratings for comparison.
5/12/2013 3:08 PM
Posted by bhazlewood on 5/12/2013 10:11:00 AM (view original):
Posted by mr_automatic on 5/12/2013 7:41:00 AM (view original):
This formula might make it a littler easier for you.

(1 - ((Rank -1) / (Number of Players - 1))) * 100

The best player would be 100, the average player would be 50 and the worst player would be 0.
That's another way of doing it, but I chose to use their actual rating instead of their rank so that players who score similarly get similar points.  In the example I gave, with three DB's, the second guy would get a 50 (if there were just those three) but he's clearly closer to the top than the bottom.

DB1 = 100
DB2 = 97
DB3 = 0

I don't think you want that either.

If you want a rating's system that anyone could look at and know that a rating of 50 meant average, then MrAutomatic's way is the way to go.

Even if all of the DBs in the world were rated from 90 to 100, there are still "above average", "average" and "below average" players in that group.  MrAutomatic's method tells you which ones fall within what group.  Another advantage of MrAutomatic's method is that it would not be skewed by SIM AI recruits.  The worst player is still the worst player.

5/12/2013 3:48 PM
what does it matter really, if everyone's qb, p or k recruit is similarly "de-valued"...winning the recruiting class rank not as important as winning the overall rank when they are seniors, or their growth over that period, no?
5/12/2013 7:04 PM
Here's why I don't like using RANK instead of "Relative Rating" when calculating Recruit Class Value.  These are from the latest Yost D1A recruiting class.  The rating numbers are from my "beta" spreadsheet, so they are using the base position values in the GD 3.0 engine, not current GUESS ratings, but the principle is the same.

RATING  RANK           RRV "Rating"         RRV "Rank"

Using Rank makes no differentiation between player scores, so in this case, a guy who is a full five points lower in rating is still pulling in a 95% Rank value.

One "problem" with Relative Ratings, of course, is that AC-signed players skew the bottom numbers, so the range between the top score and the bottom score is larger than it would be if no walk-ons were involved.   The Median (middle value) is almost always higher than the Mean (average value) because the AC signings skew the average down.  Here's the middle 10% of that same sample.

To answer jmcgarry's question : None of this REALLY matters at all, since it's all for fun anyway.  No rating formula is "right", since the game doesn't work on 10% of ATH + 30% of STR + whatever when it comes to determining results anyway.  Still, I want the ratings to be as good as they CAN be, and penalizing a recruiting class because they had to sign a QB or a K - or worse, two QB's and a K, as I had to do this season - just doesn't seem to make sense.
5/12/2013 10:59 PM
that's cool - i use the yatzr & guess scoring myself all the time & think its great - i just didnt think that the inclusion of a lower current score for qbs or punters or kickers matters all that much in terms of listing a classes "value". i do always look to see where we rank and its fun - but my concentration in recruiting is finding the high current score recruit with the right upside ratings in the scouting reports & right work ethic, which i have found to be a much bigger indicator of future growth & "value" as seniors, and there really (i don't think) is a way to include that. but i get it, its nice to see your efforts highly ranked, & when the gb is generally 10 points lower than your average player its a downer on the score. i'm just more inclined to look at the current overall as more important than winning the recruiting score.
5/12/2013 11:52 PM
to illustrate; westchester in rockne d2....i actually recruit a qb every single class....i have 5 on the team & they all red shirt as frosh & only play as sr's so that i always have a "boss" qb every year & wouldnt suffer cuz of a devastating injury to the starter. we usually score in the top 15 of the frosh rankings, and a qb every year drags that down....but i have also had 5-6 sky the limits in the past 4 yrs and have never taken a less than "room to grow" recruit and min 45 or so work ethic, even if some higher current score low potential guys turn green for us....the result is top 4 overall class 4 out of last 5 yrs. but again.... i get it & would welcome any positive changes to the system
5/13/2013 12:02 AM
◂ Prev 1234
GUESS by the numbers Topic