Positional Adjustment

bowiac  3/4/2013 5:34 PM

I e-mailed the site about this but it may be helpful for others, so I'll post here too:

Can I ask how positional values are handled by Baseball Monster? It's not clear to me. Unlike basketball, baseball positions aren't nearly as fungible. When I look at the projections, I see positional rarity listed, but I don't see how that's being baked into the projections themselves. Can I ask how you guys are handling this?

Are you seeing how many standard deviations Player X beats their positional average by at each category, and then adding that up? Or are you seeing how much they beat the league average by, and then comparing by position? Or is there no hardcoded positional adjustment at all?

ken_admin  3/4/2013 6:27 PM

The positional rarity is currently a display-only value and does not affect the players value.  Previously, we had used Scarcity Values which was an amusement to the player Value, but it never seemed quite right.  As you mentioned, it is being used on the basketball site so I'll look at that and see why it didn't apply as well to baseball.

bowiac  3/4/2013 7:03 PM

This is a pretty helpful feature (essential really - you can't comprare production between C and OF). I would strongly suggest applying the positional adjustment after summing the standard scores however, not by creating position specific standard scores.

Here's a very good link for how to calculate positional adjustment: http://web.archive.org/web/20120413103624/ http://www.lastplayerpicked.com/how-the-price-guide-works-part-ii-positional-adjustments/

I would strongly recommend reading the whole series of course (parts 1-5). 

I don't mean to tell you your business, and I signed up for a full membership because of some features you offer that nobody else does (daily rankings and ease rankings in particular), but I think there might be a few adjustments needed to how values are computed.

Thanks.

ken_admin  3/4/2013 7:05 PM

Thanks.  We will take a look at these and see if we can get something added to the site soon.

ken_admin  3/5/2013 10:45 AM

Part I of that article looks good, and it's what I'd guess most z-score/standard deviation sites are doing.

In Part II of that article, where they address position adjustments, they find the value of the worst "top" player at each position.  Then they find the differences between these worst players, and add that amount to all players with their positions.  To me, this doesn't seem quite right.  As the worst player of the "top" players gets worse, all other players in his position get better values?  Is this how you understood it too? Thanks

bowiac  3/5/2013 1:22 PM

That's how I read it, and I think it's close to correct. What he's trying to do is define a replacement level. In a 12 team league, the 12th best shortstop will be replacement level. As in, nobody should be starting a shortstop worse than him.

The 12th best guy, is by definition, worth 0 above replacement level. He is the replacement level. So lets say the sum of his standard scores is -5.0, and the best guy has a sum of 8.

How much better than replacement level is the best guy at his position? Well, he's worth 8, and the replacement level guy is -5, so he's 13 units better. The same is true for everyone else down the chain. If the 2nd best guy has a sum of 7, then he's 12 units better than replacement level, yes?

If on the other hand, the 12th best first baseman has a sum of say -1, then a first baseman who hits exactly as well as the best shortstop (value of 8), is only 9 units better than replacement.

So 8 units of production at shortstop ends up being worth more than 8 units of production at first base.

Does that follow?

bowiac  3/5/2013 1:38 PM

I want to note that's not the only way to do it. He's assuming a somewhat normal distribution of talent, such that the 12th best shortstop isn't freakishly good or freakishly bad. Lets say for some reason there are 11 great to good to bad shortops, and then our 12th shortstop hits .100 with 5 RBI and 5 runs.

If the really awful 12th guy has a sum of his standard scores of like -50, then that'll screw up the calculations pretty badly. While it's certainly correct to say everyone else's value above replacement is now 50+their own value, it's not so helpful when comparing positional values.

Now this is more of a hypothetical concern. In my experience there's rarely a freakish dropoff at the low end. However, when I run my own values, what I do sometimes do is an average of the bottom end. So again, in a 12 team league, I average the production of 11, 12, and 13 to generate an approximation of replacement level.

You can also dispose of the entire replacement level concept. To create positional adjustments, just sum up the 12 best guys at each position, and then measure everyone by value above average instead of value above replacement. Tulo is worth 8 units, the average of the top 12 shortstops is worth 3 units, so then Tulo is 5 units above average. Our 12th guy, with a value of -5 meanwhile is 8 units below average.

Neither system is "correct." I would just use whichever one seems more intuitive to you on this point.

Feel free to e-mail me at [my username]@gmail.com if you have any other questions. I can help run you through the setup he's got there, plus I've got some ideas he doesn't cover, and to my knowledge, no public site cover (which are absolutely critical for fantasy baseball evaluations, but curiously aren't for basketball).

I've given the idea of valuations a ton of thought, but I don't have the coding skill to create these systems myself. That's why I signed up for Baseball Mogul. Your integration with Yahoo/ESPN/CBS is spectacular, as is the ability to create daily projections/ease rankings. The Last Player Picked tool does a lot of stuff right, but they also make some mistakes and don't have your ease of use.

ken_admin  3/6/2013 6:55 AM

I will contact you through email.  Thanks for the offer.