How would you weight factors in evaluating a player?

[conception]

I can't really delve my evaluation into a formula. Honestly, I just try to look at what a player does well and see if there are any things he is so bad at that make him not worthwhile. If eye, control, or defensive ability are too bad, I just don't consider them. If they are at certain positions, like first base or LF, I need some type of power production. Nobody plays if they don't have contact.

[NoOne]

exactly.

same with war, ops+, era+ etc etc... although they aren't completely useless.

if you make your own like op wants to do with something summing up ratings, then you can pick all the thresholds and weights. adjust them as you learn etc... that last part would be the biggest possible return. you can adjust better when you have a more consistent and mapped out current method.

with the new # of HRs being nomal, i defintely am weighting power more than i have i nthe past, and i already favored it in the past... if i had somethign like what the op wants to do, i could more easily see how i am weighting it more. (definited better, milled down better etc)

i've heard alot of fantasy sports players say they can just "look and know." they tend to finish near the bottom of any competitive league

for a vid game i'm not putting in that kind of effort... $$$ gets involved and i'm not leaving things up to a number of inevitable false perceptions that we all have.

[SR000]

Originally Posted by ThePretender

Sorry I meant to say 2 out and nobody on for number 3 hitter. Number 3 hitter has fewer high leverage abs than 2/4, which is why you want power in third spot and balance in fifth spot. Your number 5 hitter is more important than your number 3 hitter, believe it or not.

https://www.beyondtheboxscore.com/20...your-lineup-by explains it well.

I don't think I'd agree with that. I read the article but it doesn't source anything and frankly SB Nation will pay any writer who can fill space. What is a high leverage at bat? Common sense tells you that if your first two hitters have your highest OBPs then your third hitter is going to come up with men on base the most. Doesn't make sense to me to place your fifth best hitter in the middle of your other four best.

I do know - and Bill James researched this - that the third hitter is the least likely to lead off an inning however. So I do tend to put the slower hitter there over the 4th or 5th spots if all else is equal. OBP is less of a factor as a guy with a good OBP can keep a rally going even if he isn't hitting well. In addition your 5th hitter is most likely to lead off the 2nd inning so even if it's true about high leverage at bats it is something to consider. Most teams get one runner on base on average in the first.

[drhay53]

So in my post history, you can find a linear regression analysis that I performed to correlate attributes with various measures of on-field success. Basically, at the end of each season, I dump all player attributes and stats to a file, which I use to determine the importance of attributes in a 'runs'-like measure. Then I can use each players attributes to predict their WAR. These measures are highly dependent in the league modifiers, but for my latest season (which was pretty homerun heavy), I got the following weights (normalized to contact = 1.0):
con 1.0
gap 0.315517213459
pow 0.712698333647
eye 0.709514881061
k 0.217793959185

For pitchers (regression attempts to estimate FIP-, NOT runs):
SP
stuff 0.310857867295
movement 1.0
control 0.300874339998

RP
stuff 0.514771388665
movement 1.0
control 0.62997042866

Things make a good bit of common sense here; contact/power/eye lead most directly to runs (although I expected gap power to be higher weighted than it actually turns out to be).

For pitchers, it makes sense that walks and K's turn out to be more predictive for relievers.

FWIW, I also estimated defensive runs and baserunning runs, and I found that defensively it's basically experience at the position and range, with a ton of scatter. In any given season the scatter about the linear regression for defensive zone rating is ~5 runs, which is like half of the runs for most good defenders. Basically even a great defender will have a 'meh' year on defense pretty often.

For baserunning, it's hard to measure because it looks like the 'bsr' measure basically only takes stolen bases into account; there's no real way to measure the effect of the baserunning attribute with the statistics given by the game. Therefore I've hacked the 3 baserunning attributes to be 0.025 on the scale of contact=1.

For all position players, contact is the strongest predictor of runs and therefore WAR, followed by Power/eye and then defensive experience / range.

[Dyzalot]

What do you mean by "defensive experience"? Is that what you are calling the overall rating given for a position? If so I thought that was a composite rating that includes experience at the position as well as factoring in range, error rating and arm?

[drhay53]

Originally Posted by Dyzalot

What do you mean by "defensive experience"? Is that what you are calling the overall rating given for a position? If so I thought that was a composite rating that includes experience at the position as well as factoring in range, error rating and arm?

I don't think it includes their range, error, and arm, does it? I thought it was just a measure of experience alone. I thought it was a direct representation of the 'experience' attribute you can see in the editor in commissioner mode.

If not, there is probably significant degeneracy between that parameter and the other 3 defensive attributes; still, if it does incorporate experience, I think it's the only way to get the experience through an HTML dump that includes both attributes and stats, which is about the most work I'm willing to do

[Dyzalot]

Originally Posted by drhay53

I don't think it includes their range, error, and arm, does it? I thought it was just a measure of experience alone. I thought it was a direct representation of the 'experience' attribute you can see in the editor in commissioner mode.

If not, there is probably significant degeneracy between that parameter and the other 3 defensive attributes; still, if it does incorporate experience, I think it's the only way to get the experience through an HTML dump that includes both attributes and stats, which is about the most work I'm willing to do

I could be wrong but my understanding was that it was an "overall" rating that included all the base ratings plus experience. Think about it this way. No matter how much experience some people have at a position, they will never reach a rating of 20 on a 1-20 scale. If that's the case then it can't just be a rating for experience. Like a center fielder with a range rating of 12 is unlikely to ever get a rating at center better than 10. However if you move him to left he can get well above that depending on his arm and error ratings as well.

[drhay53]

Originally Posted by Dyzalot

I could be wrong but my understanding was that it was an "overall" rating that included all the base ratings plus experience. Think about it this way. No matter how much experience some people have at a position, they will never reach a rating of 20 on a 1-20 scale. If that's the case then it can't just be a rating for experience. Like a center fielder with a range rating of 12 is unlikely to ever get a rating at center better than 10. However if you move him to left he can get well above that depending on his arm and error ratings as well.

Perhaps, but as I said, using the options for 'view' and exporting to html, the position rating is the only way to get at the underlying experience, so it's probably still necessary for me to include both in the regression. To be honest I don't fully trust the results of the defensive analysis I'm doing at the moment, but since it seems to roughly reproduce my 'by-eye' assessment I haven't bothered to improve it.

[Dyzalot]

Originally Posted by drhay53

Perhaps, but as I said, using the options for 'view' and exporting to html, the position rating is the only way to get at the underlying experience, so it's probably still necessary for me to include both in the regression. To be honest I don't fully trust the results of the defensive analysis I'm doing at the moment, but since it seems to roughly reproduce my 'by-eye' assessment I haven't bothered to improve it.

Yeah I'm not criticizing your math or anything. I was just trying to understand what you meant. It makes sense that an overall position rating would be better correlated to preventing runs than any of the individual ratings anyways since those individual ratings are encapsulated within the overall rating. It is the same reason that OPS correlates more strongly to runs scored than either of the two statistics that it is comprised of.