Current year stats weight 50% default

fredbeene · 07-27-2020, 10:35 PM

Current year stats weight 50% default.
So at this default setting wouldn't it SKEW the ratings, evals early in the season?

SirMichaelJordan · 07-28-2020, 02:55 PM

Quote:

Originally Posted by fredbeene

Current year stats weight 50% default.
So at this default setting wouldn't it SKEW the ratings, evals early in the season?

There needs to be a certain amount of AB (or PA) and IP in order for current year stats to be fully weighted otherwise it’s taking stats from last year or even 2 years ago depending on the numbers. I believe it’s prorated.

PSUColonel · 07-28-2020, 04:57 PM

Quote:

Originally Posted by SirMichaelJordan

There needs to be a certain amount of AB (or PA) and IP in order for current year stats to be fully weighted otherwise it’s taking stats from last year or even 2 years ago depending on the numbers. I believe it’s prorated.

Correct

fredbeene · 07-30-2020, 12:07 AM

Quote:

Originally Posted by PSUColonel

Correct

Do you think it using any stats from SPRING TRAINING?

I would love to run spring training reports against whole league to see if i can find someone who has shown significant improvement

DaBears · 07-30-2020, 01:33 PM

I have never felt it included spring training stats - and should not anyway.

Lukas Berger · 07-30-2020, 04:29 PM

Quote:

Originally Posted by DaBears

I have never felt it included spring training stats - and should not anyway.

It doesn't.

Syd Thrift · 07-30-2020, 04:35 PM

Doesn't it also do some kind of rolling seasonal-average thing? The game might not literally look at the last 365 days but I know that even if you have your league set to scouts making decisions based on stats only, it will still turn to ratings, for instance, where no stats are available, and won't at least as far as I've seen suddenly blip into using stats only on like June 1 or something.

DaBears · 07-31-2020, 11:30 AM

Quote:

Originally Posted by Syd Thrift

Doesn't it also do some kind of rolling seasonal-average thing? The game might not literally look at the last 365 days but I know that even if you have your league set to scouts making decisions based on stats only, it will still turn to ratings, for instance, where no stats are available, and won't at least as far as I've seen suddenly blip into using stats only on like June 1 or something.

A long time ago - maybe 2017-18, I remember a thread where people speculated that you had to be a couple of months into the season before the game started using "current season" stats. Maybe it's a mental bias on my part, but that has seemed to be true. Best guess, it would make sense if you're using 25/25/25/25 (just as an example) and starting in 2020 - on Opening Day the ratings are coming from 25% 2017, 25% 2018 and 25% 2019 (with of course 25% pure ratings). If you get to May 31, then it's still 25% 2018 and '19, but 16.7% still from 2017 and 8.3% from 2020. What I don't know is if they have a flip-switch date. That's the other option - you have my first example until 5/31/2020, and then it all just moves forward a year or if that's a rolling move forward. The algorithm is very easy to build either way, but there's no way you want to all the sudden use current season stats as 25% of your rating on 4/1/2020. That would create all kinds of chaos with the AI.

My one caveat on this is the team has some reason it puts 65% weight in the default on pure ratings, so that's always worried me. Even there, I don't see jumping to 2020 on 4/1. Given the sophistication of what they've built, I lean toward the rolling example above but it could be there's a trigger around 5/31, 6/30 or at the 81 game mark it drops 2017 (3 years back). One of the joys of OOTP is the underlying mystery where we don't know everything.

Questdog · 07-31-2020, 12:17 PM

Current Years stats means just that. Previous seasons are not included in it at all at any point in the season. There is a threshold of PA that must be reached for them to count at all. I do not know what that is. Same goes for previous season stats. If there is not enough PA, then the season is not counted at all.

I argued unsuccessfully a while ago for all stats to be used from the past three seasons, with current season numbers weighted more heavily than last years and last years more heavily than the year before. For instance, each PA this year would count 4 times, each from last year 2 times, and each from the season before once. That way, numbers in April would count for something and the further you got into the season, the more the current season would count. But it was nixed...

DaBears · 07-31-2020, 05:44 PM

Quote:

Originally Posted by Questdog

Current Years stats means just that. Previous seasons are not included in it at all at any point in the season. There is a threshold of PA that must be reached for them to count at all. I do not know what that is. Same goes for previous season stats. If there is not enough PA, then the season is not counted at all.

I argued unsuccessfully a while ago for all stats to be used from the past three seasons, with current season numbers weighted more heavily than last years and last years more heavily than the year before. For instance, each PA this year would count 4 times, each from last year 2 times, and each from the season before once. That way, numbers in April would count for something and the further you got into the season, the more the current season would count. But it was nixed...

So you're saying that on 4/1/2020, if you're using 25/25/25/25 - it would actually be 37.5% 2018, 37.5% 2019, 0% in 2020 and 25% pure ratings?

Questdog · 07-31-2020, 05:57 PM

Quote:

Originally Posted by DaBears

So you're saying that on 4/1/2020, if you're using 25/25/25/25 - it would actually be 37.5% 2018, 37.5% 2019, 0% in 2020 and 25% pure ratings?

I have no idea if it would be as you suggest or 33% 2018 33% 2019 and 33% ratings or, as I suspect, that the ratings take the place of the missing data and it is 25% 2018, 25% 2019 and 50% ratings.

I_Got_Wood_34 · 08-01-2020, 02:35 AM

I always got a kick out of these weights. For one, they're way too loosely defined, especially as it relates to "current year". And two, most projection systems go as far back as 4 years in assessing a player.

Given the vagueness surrounding these weights I would probably do something like:

Current: 10%
last year: 35%
2 years ago: 20%
3 years ago: 10%
4 years ago: 5%

And then just put the rest in the ratings.

I don't think there's a "3 years" or "4 years" ago option though.

There also needs to be an explanation given for what "current year" means ..
No way would any modern day GM be basing their evals of a player based on a few hundred plate appearances.

And since we're kind of on the topic, it would be nice to see good statcast data implemented in the game engine somehow as it's by far the best data to make small sample size decisions with and right now, it's nowhere to be found in OOTP.

DaBears · 08-01-2020, 11:52 AM

Agreed on Statcast - OOTP typically brings things in a year to three years after it becomes a big deal in MLB, so my guess is that will be one of the big "sells" for OOTP22. They might have a license issue, but seem so closely ingrained with MLB now it's probably a doable thing. It will be exciting to get!

07-27-2020, 10:35 PM	#1
fredbeene All Star Starter Join Date: Aug 2016 Posts: 1,624	Current year stats weight 50% default Current year stats weight 50% default. So at this default setting wouldn't it SKEW the ratings, evals early in the season?

07-30-2020, 01:33 PM	#5
DaBears All Star Reserve Join Date: Jun 2013 Posts: 501	I have never felt it included spring training stats - and should not anyway.

07-31-2020, 12:17 PM	#9
Questdog Hall Of Famer Join Date: Jun 2006 Location: In a dark, damp cave where I'm training slugs to run the bases...... Posts: 16,142	Current Years stats means just that. Previous seasons are not included in it at all at any point in the season. There is a threshold of PA that must be reached for them to count at all. I do not know what that is. Same goes for previous season stats. If there is not enough PA, then the season is not counted at all. I argued unsuccessfully a while ago for all stats to be used from the past three seasons, with current season numbers weighted more heavily than last years and last years more heavily than the year before. For instance, each PA this year would count 4 times, each from last year 2 times, and each from the season before once. That way, numbers in April would count for something and the further you got into the season, the more the current season would count. But it was nixed...

08-01-2020, 02:35 AM	#12
I_Got_Wood_34 Banned Join Date: Apr 2003 Posts: 188 Infractions: 0/1 (4)	I always got a kick out of these weights. For one, they're way too loosely defined, especially as it relates to "current year". And two, most projection systems go as far back as 4 years in assessing a player. Given the vagueness surrounding these weights I would probably do something like: Current: 10% last year: 35% 2 years ago: 20% 3 years ago: 10% 4 years ago: 5% And then just put the rest in the ratings. I don't think there's a "3 years" or "4 years" ago option though. There also needs to be an explanation given for what "current year" means .. No way would any modern day GM be basing their evals of a player based on a few hundred plate appearances. And since we're kind of on the topic, it would be nice to see good statcast data implemented in the game engine somehow as it's by far the best data to make small sample size decisions with and right now, it's nowhere to be found in OOTP.

08-01-2020, 11:52 AM	#13
DaBears All Star Reserve Join Date: Jun 2013 Posts: 501	Agreed on Statcast - OOTP typically brings things in a year to three years after it becomes a big deal in MLB, so my guess is that will be one of the big "sells" for OOTP22. They might have a license issue, but seem so closely ingrained with MLB now it's probably a doable thing. It will be exciting to get!