Then it is worthless, like I said. Player populations have long been shown to follow normal distributions (bell curve) so any rating system that ignores that fact is therefore worthless by definition. OOTP should hire me to unfuck this for them.
-
Registration is disabled due to constant spammers. Email [email protected] and we will temporarily re-enable registration for you.
WBL Rule Change Thread
- Thread starter OU11
- Start date
Then it is worthless, like I said. Player populations have long been shown to follow normal distributions (bell curve) so any rating system that ignores that fact is therefore worthless by definition. OOTP should hire me to unfuck this for them.
How would you do it then? Would you throw out the outliers (the 80s and the 20s)?
How would you change the rating system from bat+defense? Would you fudge the numbers to make it normally distributed?
So the 20-80 system was fundamentally designed to be fit to a normal distribution in such a way that each 10 points represents a 1 standard deviation from average. Thus roughly 68% of players at any position, regardless of how bad/good the top or worst players are, should be within the 40-60 OVR range. 40 OVR is 1 standard dev worse than average 60 is 1 standard dev better than average. Whether the best player at a position is a 70 (2 standard devs from average) or an 80 (3 standard devs from average) shouldn't matter to the population as a whole.
-3 = 20
-2 = 30
-2 = 40
0 = 50
1 = 60
2 = 70
3 = 80
An 60s are going to be very positive WAR players and probably borderline all stars. 70s would be the normal best player at a position in any given year. 80s are GOATs
20s are typically only in the league by some horrible accident of fate, lol. Something like nepotism maybe.
-3 = 20
-2 = 30
-2 = 40
0 = 50
1 = 60
2 = 70
3 = 80
An 60s are going to be very positive WAR players and probably borderline all stars. 70s would be the normal best player at a position in any given year. 80s are GOATs
20s are typically only in the league by some horrible accident of fate, lol. Something like nepotism maybe.
By its nature a position won't be normally distributed, that's what I'm saying. There are 18 starters and then there are a ton of backups. The backups vary wildly in talent level. I'm not seeing why positions should have a normal distribution. There's always going to be more backups than starters at any position aside from 1B where other positions fill in usually, so I just don't see how that would work.
I think you'd have to scrap their bat+defense rating and create some hamfisted metric that would normally distribute them, but doing that would probably make it even more unreliable since it is based on something different than ability.
I think you'd have to scrap their bat+defense rating and create some hamfisted metric that would normally distribute them, but doing that would probably make it even more unreliable since it is based on something different than ability.
So the 20-80 system was fundamentally designed to be fit to a normal distribution in such a way that each 10 points represents a 1 standard deviation from average. Thus roughly 68% of players at any position, regardless of how bad/good the top or worst players are, should be within the 40-60 OVR range. 40 OVR is 1 standard dev worse than average 60 is 1 standard dev better than average. Whether the best player at a position is a 70 (2 standard devs from average) or an 80 (3 standard devs from average) shouldn't matter to the population as a whole.
-3 = 20
-2 = 30
-2 = 40
0 = 50
1 = 60
2 = 70
3 = 80
I'm going to come to colorado and shoot you with an arrow. Stop it with this. I took math, I know what a bell curve is. Explain how it would work with a position with little talent that only has a few at the top end and a lot at the lower end
I think a lot of the individual ratings are probably on something close to a normal distribution. If you sort the league by contact, for instance, I'm sure you'd find the majority of players between 40 and 60. Since the individual ratings are adequate representation of a normal distribution (or at least close enough to not offend my sensiblities) I decided to simply ignore the Overall star rating and make my own based on the individual ratings, assuming they were indeed normally distributed.
Gooksta
Well-Known Member
Is this suppose to be breaking news?So the 20-80 system was fundamentally designed to be fit to a normal distribution in such a way that each 10 points represents a 1 standard deviation from average. Thus roughly 68% of players at any position, regardless of how bad/good the top or worst players are, should be within the 40-60 OVR range. 40 OVR is 1 standard dev worse than average 60 is 1 standard dev better than average. Whether the best player at a position is a 70 (2 standard devs from average) or an 80 (3 standard devs from average) shouldn't matter to the population as a whole.
-3 = 20
-2 = 30
-2 = 40
0 = 50
1 = 60
2 = 70
3 = 80
An 60s are going to be very positive WAR players and probably borderline all stars. 70s would be the normal best player at a position in any given year. 80s are GOATs
20s are typically only in the league by some horrible accident of fate, lol. Something like nepotism maybe.
Right, the star system is based on position not overall talent. So if OF talent is much heavier, other positions will have skewed curves.
Maybe this would help you
I would explain that your view of "talent" is such that you are desiring to grade the worst player a 20 and the best player an 80 and then fit everyone between on a linear scale and I don't understand why. Sometimes a position can have 0 players in the 20 or 80 range. It can even have 0 players in the 30 or 70 range since we only have 18 teams.
Why not take the average players at the position (whatever they may be) and normalize the scale to fit the average. You might then end up with more outliers on the top end, which is fine, but the rating system would still tell you how each player compares to their piers rather than how they compare to a random GOAT player.
How can you throw out players from your scale? Their rating is what it is. Bat+Defense. The only way to change that is to make it fit to what you want to look at. As long as the overall talent pool fits the bell, it's fine. Certain positions can be skewed. In sports there are strong positions and weak positions, the talent gets skewed when you take a subset that is specific.
Not every sample size of every population will be perfectly distributed when you aren't allowed to excuse outliers.
My guess is that what OOTP did was average a bunch of ratings together to come up with the overall rating, but you can't simply use a mathematical average (mean) calculation on numbers that were based on a normal distribution. You will have lost all the value that the information contained in the first place if you do that.
For instance. Lets assume the "Overall" rating is an equal combination of only the contact and power ratings.
Player A has contact rating of 20 and power rating of 80
Player B has contact rating of 50 and power rating of 50
Both of these ratings are based on normal distrubtions which means Player A's contact rating is SOOOOO bad that they are 3 standard deviations from the average player. This means they are basically blind. But he's paul bunyan strong, so every ball he hits (he never does) goes for a home run because his power is 3 standard devs off average. These ratings are not linear scale, which is important to remember.
Player B is average in both categories.
If you make an "overall rating" based on the mathematical mean of these two ratings, you've then assumed a linear scale when there isn't one and fucked up all the information the original rating contained by adding them together and dividing by 2. Now you think you have two 50 OVR players which isn't at all true as Player A will be a disaster.
For instance. Lets assume the "Overall" rating is an equal combination of only the contact and power ratings.
Player A has contact rating of 20 and power rating of 80
Player B has contact rating of 50 and power rating of 50
Both of these ratings are based on normal distrubtions which means Player A's contact rating is SOOOOO bad that they are 3 standard deviations from the average player. This means they are basically blind. But he's paul bunyan strong, so every ball he hits (he never does) goes for a home run because his power is 3 standard devs off average. These ratings are not linear scale, which is important to remember.
Player B is average in both categories.
If you make an "overall rating" based on the mathematical mean of these two ratings, you've then assumed a linear scale when there isn't one and fucked up all the information the original rating contained by adding them together and dividing by 2. Now you think you have two 50 OVR players which isn't at all true as Player A will be a disaster.
Last edited:
Karl Hungus
Here to fix the cable
I think the ratings for catchers are fucked- defense isn't valued highly enough. Ask @ZackMills
Gooksta
Well-Known Member
You have other things you can derive from to create edges.. example speed or defense
ZackMills
Have mercy
It's based on their hard numbers behind the closed doors. You don't have access to those, but they are the real ratings not a 40-60 deal.
Based on not being able to throw out outliers I don't see a way that you can unskew a skewed talent pool. What are you going to do if there are 5 70+ guys and 40 30 and less guys?
What's the point in lying about their ratings? Those 40 guys are shown to be very substandard. I'd rather not have a system where shitty guys have pumped up stars because you throw the best 5 players out of your calculation.
If judging based on position I much prefer a skew because it shows those guys aren't very good. I don't care if they are average if you take out the top guys, I care if they are a 50 on the 20-80 scale calculated by bat+defense.
Gooksta
Well-Known Member
Pitching/defense/ballpark should be seen as one entity imo..
What I'm saying is this. The Overall rating is obviously based on some mathematical equation that we don't have access to and it combines various ratings to come up with 1 number. By doing the mathematical combination on 20-80 scale ratings in a way that assumes they are linear scale (they aren't) you fuck everything up.
The easy way to fix this would be to take the overall ratings and re-normalize them (this would be akin to grading a test by merely shifting the average to where you want it)
The better way to fix it would be to convert the individual ratings to a linear scale before doing mathematical operations on them, then taking that combined value and re-normalizing it (this would be akin to actually grading a test on a curve)
The individual ratings for those players look pretty normally distributed to me. My guess is that the overal computation is fucking this up for the reason I highlighted above and the resulting number is no longer normally distributed.
That's what i thought. Your system would make it worthless since you'd be grading players better/worse than they are.
Use any calculation you want for bat+defense and a lot of positions will be skewed. Id rather have that skewed number to show me how they are. So if they grade out as a 40, their star should be 2. I'd support that and honestly from looking at these numbers I'm pretty sure that's what's going on. Ill have to look more after i get these ribs going.
But to make the guy who grades out a 4* because the talent is skewed down sounds ridiculous
Last edited:
I think we are at an impasse because you aren't understanding what I'm saying about the math here. In order to do mathematical operations on a normal distribution, you must first convert it to a uniform distribution. Then you do the math based on the uniform distribution to get ur Overall rating, then convert it back to a normal distribution and everything should work fine. If you take ratings that are based on a normal distribution and you try to perform algebraic mathematical operations on them as if they were based on a uniform distribution, ur gonna have a bad time. I believe that's why the OOTP "overall" star ratings are so fucked. It is about as valuable analytically as a 1-10 scale for women. Having different scales for each position only further complicates this.
Take 20-80 scale for each individual rating and do this to it:
http://en.wikipedia.org/wiki/Probability_integral_transform
then do the maths however you want to get the overall rating for the player
Then convert back into normal distribution and apply to 20-80 scale. Ur population should be normally distributed and you have a valuable tool for comparing players.
fwiw, the BTT system only considers fielding ratings relative to position (and weights them differently depending on the position, but I'm still playing with that). Batting is based on the player population as a whole and is independent of position. It is worth noting that everyone already does this subjectively. That's why people can look at Andy Bobandy and say he's desirable even though he only shows up as a "2 overall out of 5" on OOTPs scale. Everyone knows he isn't worse than average.
Take 20-80 scale for each individual rating and do this to it:
http://en.wikipedia.org/wiki/Probability_integral_transform
then do the maths however you want to get the overall rating for the player
Then convert back into normal distribution and apply to 20-80 scale. Ur population should be normally distributed and you have a valuable tool for comparing players.
fwiw, the BTT system only considers fielding ratings relative to position (and weights them differently depending on the position, but I'm still playing with that). Batting is based on the player population as a whole and is independent of position. It is worth noting that everyone already does this subjectively. That's why people can look at Andy Bobandy and say he's desirable even though he only shows up as a "2 overall out of 5" on OOTPs scale. Everyone knows he isn't worse than average.
Last edited:
You won the trapping today, my friend. I am caught.
No, I completely understand what you're saying. You'd normalize the data set so it will fit what you want. That's fine for a lot of things, but this is not one of them. I don't want a guy to be a 2.5* just because he is in the middle of the range after you normalize it. That's where you're losing me. We want different outcomes. I want them to be shown how good they are at catcher, you want them to be shown how good they are in relation to the catcher talent pool after you normalize it. Honestly I thought the stars were more based on what you're wanting which is why I disregarded them as useless. I'm glad we did this because I will actually pay attention to them now that I know they are just a grade of the player at that position.
You joining the league is awesome, I would have never figured that out if it wasn't for this
Last edited:
What would happen if we didn't have a catcher with a defense over 40? Would you still grade the best bat with 40 defense as a 5* catcher? That's the best example of why I don't like your forced bell curve and prefer the skewed curve for position. I'd rather that guy be graded on being a catcher and leave it at that.
Long story short. I vote no for the change to 20-80 scale for Overall and Potential because they are arbitrary scales and some fucking gay looking stars are just fine in that case. If a genie granted me one OOTP wish, I'd change it to a scale between 7 and 11 dongs, which would properly show my disdain for this fuckshit. Presumably that's not an option doe.
Gooksta
Well-Known Member
Why not do your distribution based off positional value.. then combine all positions together for player overall value
Gooksta
Well-Known Member
Now yo have me wondering about that.. I always ignored stars
The real question is if the stars change if a guy has the same defense at two positions.
Lloyd Carr
Well-Known Member
I still want the stars to indicate overall player rating. I don't give a shit about positional rank.
But that's just me.
But that's just me.
Lloyd Carr
Well-Known Member
Well so far it looks like I'm the only one who wants it. I dont want you to change something unless everyone (or the vast majority) want it.
Anyone else want the change? Or am I alone?
The answer is yes
The real question is if the stars change if a guy has the same defense at two positions.
So either ootp values different bats for different positions or its normalized. That blows, it's back to useless. I really wish it was just how good they are at the position
Gooksta
Well-Known Member
Defensive value on the defensive spectrum I would assume
Lloyd Carr
Well-Known Member
See, I'd rather know how good they are overall.
A 3.5 star 3B based on positional rank may only be a 1.5 star overall. I don't want them 1.5 star scrubs on my team.
I want that too
I'll need at least a full season to finalize my BTT ratings, but I have been having fun developing a system to evaluate batting ratings and each position's fielding ratings. Probably want more input on it eventually and will share results with others. I'll also do some cool statistical analysis of the data from the season when we get there!
Gooksta
Well-Known Member
That would probably mean there aren't many 3b to choose from
Lloyd Carr
Well-Known Member
Well I'd rather know that upfront.
I don't want some 1.5 star chump trying to fool me into thinking he's a 3.5 star talent. That's horseshit. He's trying to bamboozle me about how good of a player he is, and I won't put up with that.
For a $39.99 subscription to the BTT system, you can have a BTT "true talent" rating for every player on your roster that will predict essentially what his WAR will be at a given position. I will also show you where they relate compared to all players in the WBL and all players at their position in the WBL.
For $59.99 OU will give you all 5*s