Personal Opinion: Using Zips is the worst
 

So this will likely garner a lot of negativity but it's an opinion really. OOTP 19 started the use of ZIPS and I personally saw a major difference in game with the use and not in a good way. ZIPS is a fantastic in depth projection system but it serially underrates certain players depending on park factors. It heavily penalizes offensive park teams. During the past year I have followed some statistics in an online league I co-run and my team is the Colorado Rockies. Here are a couple stats that to me draw a conclusion that OOTP 19 had major issues likely due to ZIPS.

We ran 4 seasons in prior OOTP versions and the Rockies BABIP in those versions were as follows

2017 - .295
2018 - .302
2019 - .304
2020 - .297
Switched to OOTP 19
2021 - .317
2022 - .313
2023 - .312
2024 - .312
2025 - .337
2026 - .321
2027 - .330

Prior to 2017 the Rockies team BABIP had ranged from .279-.316 with the average being .298. Since OOTP 19 came out the average is .320... Using the ZIPS system for forecasting ratings/stats has been a negative for OOTP when it comes to parks that are extremes whether it favors hitters or pitchers.

Is that why Milwaukee always sucks in OOTP...?

Milwaukee, Colorado, and Texas all seem to suck in every OOTP 19 standard game I've played. This might explain it a bit.

long story short -- go autocalculate modifiers before the next season... once is enough for 'now'... may need more if it is a transition context that requires a full league turnover before new ~equillibrium is reached.

---------------

this is almost certainly Stats and AI settings related. this isn't necessarily zips fault at all. very unlikely, the fault of zips converting RL stats to ratings for players, actually.

resulitng BABIP is controlled by the stats and ai settigns that relate to babip calculation.

now, if all the league stats are ~normal and everythign else is normal, there still may be an elevated babip for any particular team, i guess... not everything can be callibrated perfectly, is what i mean. something is bound to be off if you look deep enough. (this is where you can argue about zips... the subtle deviations as you let the model play out could be caused by talent being out of proportion or miscalculated etc etc... but could be more than just proportions and distribution of players too as the cause... impossible to be perfect, relative to a video game. close is good enough.)

e.g. you can either focus on individual career results or league averages.. one does not give you the other in the same proportion as RL. if your league averages are spot on, then likely something won't reach the same levels as what you see in the RL baseball record book. it's not perfect, but it's close, either way you choose to go on that.

in this case, i'd wager you just need to auto-calculate your modifiers.

you can help determine cause by looking into things too... is league avarege BABIP off? if so, then modifiers will almost certainly help solve the problem.

be forgiving. like you said it ranges from 279-.316. +/- .020 may not be off... even if it happens for a few years in a row. time would tell, of course. clearly the average has shifted, but if you make an adjustment and see ~.317 again, let it ride for a bit to determine actual change you effected.

is this an import from ootp18 or before? players may not be created in the exact same proportions as before (fictional, of course). this will impact stats as teh league progresses after importing to new release. anytime you import to new release, you likely have a transition period for ~20-25 years as different players fill the league than how they were created before.

LTMs likely need to be autocalculated every few years to 5 years while you turnover league with new way of creating players... as they fill the league, the stats will shift, otherwise, because they are different as a whole (on average over time).

this is the same suggestions i make for people transitioning from Real 2018 players to fictional players after the included RL amatuers run out. for the same exact reasons, you want to autocalculate the modifiers to keep stats somewhat near baselines as you transition. don't do it every year... that will make stats unnaturally flat, if you autocalculate too often or without a good reason to do so. (like this context)

It's not zips fault - we only really use them for creating the current player ratings. Anything that happens after that is the player creation/player development engine at work.

Why your team BABIP is higher than before could be many factors. Maybe you simply have better players, or better BABIP players. Maybe the other teams in your division are playing terrible players at SS against you giving you a lot of extra hits. Maybe league stats have spiraled up, or league offense has increased, or there's something weird going on with your park factors.

There's a lot of possibilities, but if you're anything past the first year of a new game, it's not Zips at play. And anyways, we've been using zips for ratings for the last 2-3 years now (since OOTP 16 or 17, I think).

Thats the issue I, and most likely others, have. ZIPS seems to underrate certain teams at the jump, for example this past season, Milwaukee would usually finish under .500 in sims when in reality they finished with the best record in the National League. And tbh, it really wasnt surprising at all, everyone knew Yelich would do better being in Miller Park and everyone knew Cain can still play.

To be quite honest, I think basing ratings off of projections from a single source makes it prone to stupid inaccuracies because of human bias, whereas going off of 3-4+ projections or past season stats would be a lot better.

EDIT: As another example, I just ran another 2018 sim, all default settings and this is what I found:
Sim record / Real record
Milwaukee: 71-90 / 96-67
Colorado: 71-90 / 91-72
Texas: 66-95 / 67-95
Cincinnati: 90-71 / 67-95
Philadelphia: 94-67 / 80-82
Atlanta: 76-85 / 90-72

Don't see the problem especially with the new live update feature.

Nothing will predict real life

Well, it's pretty easy to say that a projection system has underrated a team AFTER the team has performed well. Of course, if you ran the Brewers' 2018 season hundreds of times, you might well have just as many times that they win 73 games as 96.

This article from by Jeff Sullivan (at Fangraphs, before he was cruelly snatched away by the Rays) looks at the projections entering 2018: https://blogs.fangraphs.com/here-is-...projections-2/. ZIPS saw them as a 79-win team; the fans saw them as an 82-win team. No one saw them as a 96-win team at the time.

This article by Sam Miller (then at Baseball Prospectus) goes into all the bizarre things that might come to pass when you simulate a season (2016) a million times: https://www.baseballprospectus.com/n...-city-1000000/

The fact that OOTP leads to results that differ from what we later see happen in real life strikes me as a feature and not a bug. It wouldn't be much fun if everything turned out the same way every time!

First off, 79 wins is laughable when you really look at it. Milwaukee won 86 in 2017, added Cain and Yelich, and they think they'd go down to 79? With their GM hailing from an Astros franchise that led a complete rebuild to a world series? Chicago last year was probably a 90 win team, Milwaukee was probably a 86-90 win team, St Louis was probably a 85 win team. I think the problem here is these projections are based on matchups, so they say "oh, this team wont do well because they have to play this other team a lot." No, let real life and the sim engine decide that. If you project a top hitter to hit .265 because he has two teams with in his division with ace staffs, then let the sim decide that. RATINGS ARE CURRENT ABILITY, NOT STAT PREDICTORS.

Second, the randomness happens throughout the season because of non-ability factors like injuries, roster moves, etc. If you turn off injuries, turn off roster moves, and just run a 162 game season with each player playing at natural current ability level, yes i do expect it to play out similarly each time, because the law of averages says so. the sim engine itself will naturally provide some randomness to it, but common sense tells you each sim should provide a similar result most of the time.

----------------------------------------------

What ZIPS does is give projections on what will happen based on their simulations. Take Yelich, they project him at .298, but we know his ABILITY is a .300-.330 hitter. They say he ends the year under his ability because he's in a tough division which makes up the majority of his schedule.

Whats happening, is his ability is around .315, his projections are at .298, and OOTP takes the projections and says his current ability is .298.

They're basing their ratings off of simulations and then running further simulations.

1st Yelich is a career .297 hitter thats 790 games.
2nd zips, steamers, atc all of the predictions based tools are mostly identical to one another.

But lets take wrc+ and woba

career - woba :.363 wrc+:130
Zips 2019 prediction : woba:.384 wrc+:141

So Zips has hit offensive production rated significantly higher than his career numbers.

790 MLB games last years 147 games out of that 790. Meaning 2018 was career best across the board. If you were to graph it you would have 2018 out in no mans land not near any other of his years. So 2018 looks like a huge anomaly until he does it 2 more times.

Most of his Zips 2019 numbers are higher than his career totals. He had 1 insane season last year. Other than that he was putting up consistent numbers for 650 games of his career.

In fact Yelich has only hit 330 once and that was in A+ ball with a large sample size being 100+ games. Thats it. No one in the world will have him close to hitting .330 in 2018 and not even in 2019. Maybe he does, maybe he doesn't.

That's what makes baseball great. No one knows what will happen on the next pitch.

Oh how the rays have hurt us :( Fangraphs and Effectively wild will never be the same

Looking at Yelich's 2019 ZiPS projection, I see nothing wrong with it. He's projected for a 144 wRC+, and he's projected as the 7th best hitter in the game this season by WAR. 144 would be his second best season by that measure only behind last season. He had an obviously outstanding season last year, but would anyone expect him to reproduce that in 2019? Outside of Mike Trout i dont think theres any player that you can look at and say "his current ability is a 7.6 win player"

As for Milwaukee as a whole, they are low in the projections because of their starting pitching.

You guys arent getting the point. Im not saying the projections are terrible, im saying his actual ability level is a bit higher, at least in terms of BA. He has the contact ability of a .310 hitter, but his BA this season will probably be around .298 because he will face a lot of tough opponents.

The problem is OOTP take the projections and saying his ability is of a .298 hitter, so when you run the simulations, he will actually hit under that because he's facing tough opponents in the sim, even though it was already accounted for in the projections.

OOTP has to start somewhere, and on the whole Zips does a better job than Markus sitting in a chair and dictating ratings for every player in organized baseball.

Also if anyone dislikes ZIPS they are welcome to re-edit their game to whatever projections they prefer.

I'm not saying that to be a smartass. I'm simply saying the game offers that. No matter what projection system is used, someone will be unhappy.

I remember when OOTP started using DIPS and the uproar that created over a decade ago.

But we have so much control so we can do whatever we want.

Sent from my Pixel 2 using Tapatalk

Except in his career he doesn't hit .300 or above .300 except one time last year. Anomaly until he does it multiple times.

Again career 790 games he is a .297 hitter. Not sure what the problem is honestly. Especially once ootp20 has live updates

where are you getting that his "contact ability" is a .310 hitter? what makes you think that? i dont even care about batting average but he has batted greater than .300 exactly once in his career. and for what its worth, his babip was a career high .373 last season, second highest in the league, a number that suggests regression. also, is the difference between .310 and .298 really all that different? one sure looks better, but really they might as well be the same as far as im concerned

It does appear that generating ratings based on projections, particularly with an unbalanced/divisional schedule, would tend to favour strong teams in weak divisions and hurt weak teams in strong divisions (at the extremes). Is there not a "double dipping" of the strength of schedule effect? I guess Cleveland and the AL Central would be at the other end of the scale?

I don't know this as fact [I'm an Aussie so MLB projections are unheard of here] but one would assume the projections factor a number of variables such as ballpark, own team strength (think a batter being protected) and strength of schedule (ie division).

I guess it depends on whether OOTP uses the final variable adjusted data or the raw independent data (based on prior performance and growth/decline curves, etc) for their ratings generation?

It would seem that projections result in any given player having completely different projected stats, and therefore potentially OOTP ratings, based on which team they play for.

So do Manny or Bryce, for example, enter OOTP20 with ratings determined by who they signed for? Whereas, in theory, shouldn't they be entering OOTP20 with the same ratings regardless of who they signed for (and then the OOTP simulator applies the ballpark effect, his own teammates apply the own team strength effect and the schedule applies the divisional bias effect... along with any other effects)?

As the OP mentions, I also wonder how the Rockies are treated? Does OOTP generate ratings for Rockies players based on the ZIPS projections that are already park factored, then does the OOTP simulator apply a further park factor when simulating?

I guess this could all be a moot point if OOTP ratings are generated based on the raw projected data and one would hope that's the case.

the difference between hitting .298 and .310 is 7 hits over a 600 at bat season

literally one hit a month. if you watched every game you couldn't tell the difference between a .298 and .310 hitter if the stats weren't on the screen for you. if that's the margin of error then things are looking alright, hardly "the worst"

Thats the exact point im trying to make, yet everyone always gets stuck on the damn example. If projections are based on all variables considered, and then those factors are applied again in the sim, then youre applying those factors twice.

If its just raw projections not taking into account any season-specific factor though, then yeah its probably fine to just take those projections as ratings.