Live card rating updates appear very strange this time

[allenciox]

So, the way I have looked at live card rating updates is to look at zip projections for a player for the year both at the beginning of the update period and at the end, subtract the difference on the "key" stat (I figured wOBA for batters, FIP for pitchers), and use that to project who will increase/decrease, etc.

That worked well in projecting results in 2020, and in the first update this season: my results tracked pretty closely to the actual gains/drops (or at least were in the same direction, anyway). When a player underperformed expectations, they usually dropped. When they overperformed expectations, they went up.

Until now. This update almost appeared random, with very little relationship to how a player actually performed during the period in question.

To illustrate this, I will focus on two players in question: one pitcher and one batter. Both players had a fantastic month, but dropped from diamond to gold in the latest update.

To illustrate this, I will use Fangraphs for each player. For each one, I did a search for the player on Fangraphs, clicked on the "Game log" tab, and then chose to look at the period from 4/26/2021 to 5/30/2021 (i.e. the exact period of time since the previous update, which came out on 4/26).

So, first here is Yasmani Grandal (C): Here is the link:
https://www.fangraphs.com/players/ya...1-05-29&type=0

You can compare his actual season results at the top with his period-specific results below. For the rating period, he had 84 PA and a wOBA of .393 (wRC+ of 157). This compares with the season long result of .345 wOBA (so he did much better over this period than he did during the first rating adjustment period). He also did much better than his projections for the remainder of the season, which is a wOBA of .342. So, how was he rewarded for his stellar performance over that period? By a drop in his rating from diamond to gold. I had him projected to actually rise to 95. No, he fell.

The second example is Liam Hendriks. Here is his fangraph page for that period: https://www.fangraphs.com/players/li...1-05-30&type=0

So how did he do over that period? Well... fantastic. , he had 14 appearances and 12 2/3 IP. His era was 0, his FIP was 0.43, he gave up zero home runs, 14.2 k/9, 1.4 BB/9. The result: they drop him from diamond to gold.

These are just two of quite a few examples I could quote. Yes, I lost out on these players' declines because they were so unexpected. But, I, must admit, I am confused at the logic behind it all. If the declines and increases are NOT based on their performance over the period, what are they based on?

[Lukas Berger]

Grandal for the year is overall putting up a .135/.393/.344 line. He also had 23 K's in 41 ab's in just May (or primarily the update period).

That's not going to get him any boosts.

Even if he improved during the update period, you also have to take into account that each update period increases the amount that the overall season's line is factored into the ratings.

And so his putrid performance from the first update period actually still drags him down, as his overall season line has more affect on his ratings the further we get into the season, according to the algorithm.

[Lukas Berger]

The same would be true for Hendriks as well, but to a lesser extent. Those 4 HR's he allowed in April are still hurting him now, as they are counted as more of his total ratings than they were in just the first month.

[BennytheKid]

Originally Posted by Lukas Berger

The same would be true for Hendriks as well, but to a lesser extent. Those 4 HR's he allowed in April are still hurting him now, as they are counted as more of his total ratings than they were in just the first month.

So I'm confused, stats from the previous month incorporate into this months update as well?

[chazzycat]

have there been any other changes to how things work, other than now apparently updating defensive ratings?

A lot of us who typically have somewhat of a handle on these things, are totally scratching our heads this morning. (we are the ones usually defending the updates when folks don't understand...but this one seems different)

[Lukas Berger]

Originally Posted by chazzycat

have there been any other changes to how things work, other than now apparently updating defensive ratings?

A lot of us who typically have somewhat of a handle on these things, are totally scratching our heads this morning. (we are the ones usually defending the updates when folks don't understand...but this one seems different)

Other than the defensive ratings, as far as I know there weren't any other adjustments to the algorithm.

What are some of the examples you've seen that look especially strange?

[allenciox]

Thank-you so much for posting this, Lukas. It explains a lot. I was under the impression that the ratings changes were done in an "independent" basis from month to month. I did not know that there could be a lag effect on ratings, but it helps to adjust the approach that I use to evaluate. Rather than looking at Zips(Update) end-of-period to beginning-of-period, would it be better to use difference of Zips(update) to Zips(beginning of season) to account for this accumulation?

I do see that both players are down slightly in comparing ZiPs(update) to ZiPs(beginning of season)

[chazzycat]

Ozzie Albies, Gleyber Torres, Giancarlo Stanton come to mind on the hitting side. Guys who had fairly bad Aprils but significantly turned things around in the 2nd update, basically aren't seeming to get any credit for that improvement.

For pitchers I'm sure you'll hear a lot about Aroldis Chapman, again. Granted he wasn't quite as amazing as the first update, giving up ONE homer. but still he has been so fantastic on the season overall, it's hard to explain any drop.

[Lukas Berger]

Originally Posted by BennytheKid

So I'm confused, stats from the previous month incorporate into this months update as well?

The whole season's performance factors in to each update. Obviously the current months performance has the vast majority of the effect on any changes though.

[rdt24]

Originally Posted by Lukas Berger

Other than the defensive ratings, as far as I know there weren't any other adjustments to the algorithm.

What are some of the examples you've seen that look especially strange?

I by no means have done any extensive research, but Max Fried surprised me a bit. Wheeler moving +17 is pretty impressive too

I’m also a little confused about the ‘bad performance carrying over’ thing. Are you saying that a player that does poorly in the first time period, but then hits projections in the second period will still go downwards in the second update?

[Warfsu]

This is my first year playing the game. Noticed the the Live Bieber went from 99 to 100. Also noticed that even though the L10 sale price hasn’t gotten much past 7k, the quick sale value jumped from 4k to 20k. Statistically, the Special Edition Scherzer is better,l. Seems like an easy sell, no?

[BennytheKid]

Originally Posted by chazzycat

Ozzie Albies, Gleyber Torres, Giancarlo Stanton come to mind on the hitting side. Guys who had fairly bad Aprils but significantly turned things around in the 2nd update, basically aren't seeming to get any credit for that improvement.

For pitchers I'm sure you'll hear a lot about Aroldis Chapman, again. Granted he wasn't quite as amazing as the first update, giving up ONE homer. but still he has been so fantastic on the season overall, it's hard to explain any drop.

All these are great points

[allenciox]

So, I was looking at some players based on the premise that (ZiPs update) would apply to (ZiPs beginning of season), and as I mention, for the two players I listed that would explain results nicely.

However, for other players it doesn't work as well. Perfect example is Aroldis Chapman. He started the season with a ZiPs projection of 2.63 FIP. After his phenomenal first three weeks, he had a negative FIP, and his rating did go up, from 98 to 99. This caused his projections to markedly increase going forward from that point (projecting his remaining season FIP to be about 2.3).

During the rating period, his FIP did normalize to an extent. As seen here: https://www.fangraphs.com/players/ar...1-05-28&type=0 his FIP was 2.27 over the ratings period. Still, that was better than projected.

Total results this season for him are FIP of 1.31, far better than the FIP originally projected. His ZiPs update shows him forecast at a full-season FIP of 1.87, more than .75 FIP "better" than originally forecast. Yet his rating dropped two points to 97, meaning he has lost a point from his original rating this season even though he has had a phenomenal season overall. So that would still be a head-scratcher in the revised way of thinking of things.

[Lukas Berger]

Originally Posted by chazzycat

For pitchers I'm sure you'll hear a lot about Aroldis Chapman, again. Granted he wasn't quite as amazing as the first update, giving up ONE homer. but still he has been so fantastic on the season overall, it's hard to explain any drop.

Chapman's stuff is rated so that he is expected to keep up a K/9 ratio of over 17! During the May period, he 'only' had 13.5 K/9.

He's rated to have around a 3.5 BB/9, during May he had 4.5.

He's 'supposed' to only allow 1 HR every 23 or so innings, during May he allowed 1 in 12 innings.

So in all the ratings that matter, his May was subpar (for him, even if not for a normal mortal).

[Lukas Berger]

Originally Posted by rdt24

I by no means have done any extensive research, but Max Fried surprised me a bit. Wheeler moving +17 is pretty impressive too

I’m also a little confused about the ‘bad performance carrying over’ thing. Are you saying that a player that does poorly in the first time period, but then hits projections in the second period will still go downwards in the second update?

Not necessarily. In most cases, probably not.

It all depends just how bad they were and just how good they are in the next period. Sometimes there can be borderline cases though, and guys who were 'so' bad that their ratings even ran up against the cap of how much they can fall in one update.

[Lukas Berger]

Originally Posted by chazzycat

Ozzie Albies, Gleyber Torres, Giancarlo Stanton come to mind on the hitting side. Guys who had fairly bad Aprils but significantly turned things around in the 2nd update, basically aren't seeming to get any credit for that improvement.

Albies put up a .242/.321/.474 line. Our projections had him significantly higher, even as a potential slight improvement on his best year so far in 2019, at .298/.351/.540

So compared to his baseline, he underperformed.

I don't have time right now to go into the other guys you mention, but very likely it's something similar with them.

[allenciox]

Originally Posted by Lukas Berger

Chapman's stuff is rated so that he is expected to keep up a K/9 ratio of over 18! During the May period, he 'only' had 13.5 K/9.

He's rated to have around a 3.5 K/, during May he had 4.5.

He's 'supposed' to only allow 1 HR every 23 or so innings, during May he allowed 1 in 12 innings.

So in all the ratings that matter, his May was subpar (for him, even if not for a normal mortal).

So I think this is a light bulb moment for me. So let me see if I am correctly interpreting what you are saying. It isn't that the overall is adjusted based on a statistic (like FIP, wOBA, etc.), it is that the "individual" ratings are adjusted based on how they are doing based on projections, then the OVR rating is "computed" based on the new individual ratings.

So, for example, suppose that for Aroldis, K/9 is worse than projections, but BB/9 is better and HR/9 is better. His STU goes down, his MOV and EYE both go up. But you plug the new STU, MOV, and EYE numbers into whatever calculator you are using to generate OVR, and boom, he could do better than projected in overall FIP, but still end up with lower OVR.

Is this a correct summary? Thanks for entertaining these questions. I cannot imagine where else you could find the founder and CEO of a company willing to go on discussion boards and clarify things.

[chazzycat]

Originally Posted by Lukas Berger

Albies put up a .242/.321/.474 line. Our projections had him significantly higher, even as a potential slight improvement on his best year so far in 2019, at .298/.351/.540

So compared to his baseline, he underperformed.

I don't have time right now to go into the other guys you mention, but very likely it's something similar with them.

Albies ZiPS pre-projection for the year was a .349 wOBA.

During the first update, he put up a .275 wOBA and understandably dropped 2 points.
During the second update, he put up a .382 wOBA but got no update.

It just seems like guys who struggled early, then turned things around in the 2nd update are not getting credit for it.

Stanton is much more egregious.

First update period: .283 wOBA (-3 points)
Second udpate period: .437 wOBA (no change)

[Lukas Berger]

Originally Posted by allenciox

So I think this is a light bulb moment for me. So let me see if I am correctly interpreting what you are saying. It isn't that the overall is adjusted based on a statistic (like FIP, wOBA, etc.), it is that the "individual" ratings are adjusted based on how they are doing based on projections, then the OVR rating is "computed" based on the new individual ratings.

So, for example, suppose that for Aroldis, K/9 is worse than projections, but BB/9 is better and HR/9 is better. His STU goes down, his MOV and EYE both go up. But you plug the new STU, MOV, and EYE numbers into whatever calculator you are using to generate OVR, and boom, he could do better than projected in overall FIP, but still end up with lower OVR.

Is this a correct summary? Thanks for entertaining these questions. I cannot imagine where else you could find the founder and CEO of a company willing to go on discussion boards and clarify things.

Haha, I'm not Markus, so not the founder or CEO

And yeah, with the proviso that I'm not actually doing the coding myself either, and so there may be some nuances or details I'm not aware of, what you describe above is essentially correct, according to my understanding of how the process works.

[Lukas Berger]

Originally Posted by chazzycat

Albies ZiPS pre-projection for the year was a .349 wOBA.

During the first update, he put up a .275 wOBA and understandably dropped 2 points.
During the second update, he put up a .382 wOBA but got no update.

It just seems like guys who struggled early, then turned things around in the 2nd update are not getting credit for it.

Stanton is much more egregious.

First update period: .283 wOBA (-3 points)
Second udpate period: .437 wOBA (no change)

Like I mentioned above, it all depends how much they struggled early and how much they turned things around and what their overall line ends up looking like, even after turning it around.