OOTP + Analytics + AI = Great Success!

[markprior22]

This is such a cool idea!

[JAF373]

Think we're just scratching the surface of how we can use this technology in the game.

I'm in year one of a long-term Giants sim. Used it extensively during the Draft as a co-pilot/scouting director.

Here's the after-the-fact summary:

[LansdowneSt]

I'd be great if there was a big brightly colored button to push that was "Draft Prep" and in a folder generated the csv files for the draft class, your top X prospects, your team needs, etc. Everything you'd just upload to your preferred AI to prep. A week out from the trade deadline there is a "Trade Deadline" button and the resulting csv output is a csv of the Trade Blocked players, Team needs in the league, who is Win Now and who is Rebuilding, Positional ratings to see if a prospect is pressuring a starter's job, Who is a FA next year, your budget, etc. Just everything you need instead of a screenshot of this and a csv of that that you make on your own. There will always be things you want to add but i'd like them to have standardized and complete outputs arranged for various typical situations be those I mentioned or heading in to the Winter Meetings, etc. Even a few csv files that generate the stats of the year and the standings at various points so you can have an Almanac year written up from the csv outputs generated by a click. Or put it into some podcast LLM that I can give the voice of Mel Allen to. I know a bunch of this can csv generation be done but I've never exported a csv from a sim I cared about (the csv's that I dig into are on the front end of a save). But making it user friendly and integrated and logical instead of the process that it seems to be to me (who again has never done it) would seem a way to get things that the game can already do... to just be more accessible. Those csv's seem to have finally found their moment.

[luckymann]

Originally Posted by LansdowneSt

I'd be great if there was a big brightly colored button to push that was "Draft Prep" and in a folder generated the csv files for the draft class, your top X prospects, your team needs, etc. Everything you'd just upload to your preferred AI to prep. A week out from the trade deadline there is a "Trade Deadline" button and the resulting csv output is a csv of the Trade Blocked players, Team needs in the league, who is Win Now and who is Rebuilding, Positional ratings to see if a prospect is pressuring a starter's job, Who is a FA next year, your budget, etc. Just everything you need instead of a screenshot of this and a csv of that that you make on your own. There will always be things you want to add but i'd like them to have standardized and complete outputs arranged for various typical situations be those I mentioned or heading in to the Winter Meetings, etc. Even a few csv files that generate the stats of the year and the standings at various points so you can have an Almanac year written up from the csv outputs generated by a click. Or put it into some podcast LLM that I can give the voice of Mel Allen to. I know a bunch of this can csv generation be done but I've never exported a csv from a sim I cared about (the csv's that I dig into are on the front end of a save). But making it user friendly and integrated and logical instead of the process that it seems to be to me (who again has never done it) would seem a way to get things that the game can already do... to just be more accessible. Those csv's seem to have finally found their moment.

Too true, bud. The almanac needs a few report templates added to it and I'm sure more people would use this great but somewhat half-baked in-game feature.

The next little QOL improvement I am crossing everything for is to be able to export CSVs for individual player hit / pit / def stats. Hoping it's on the devs' list to at least look at, as it would really allow a deeper dive into how a player's career has progressed and is likely to play out from that point in time.

G

[Thundercrack]

Originally Posted by low

LLMs... gross.

Sorry not sorry.

Indoor plumbing and paved streets …. gross

[low]

Originally Posted by Thundercrack

Indoor plumbing and paved streets …. gross

Never mind. Carry on.

[WooBallFan43]

The point is that not all AI use is bad. This is a fun and potentially useful way to enhance and enjoy your OOTP save. Just because something's AI doesn't mean it's inherently bad.

Yes, there are numerous bad examples of AI being shoved down our throats, you said it yourself, "the potential of misuse", but how those differ from using AI essentially as your scouting director is that AI plays a useful role, where as, for example, an AI vacuum, doesn't need AI, and, in fact, is actively harmed by shoving AI in places where it just doesn't need to be.

So both of your "arguments" are hasty generalizations, throwing the entirety of a rising sector (for better or worse) into a single box. Is medical AI bad when it's used to discover, find, and treat diseases faster than a team of doctors working alone? No. That AI's used as a tool and a helper. Same thing as what luckymann's doing. He's not outsourcing the entire process to AI, merely using it as a tool, just as advanced stats and ratings are used.

[WhiskyTango]

I think AI is taking OOTP to the cusp of a new era -- where the stats will be brought to life in ways unthought of. Am surprised no one has as of yet incorporated AI into generating entire scripts of play-by-play options. May even be possible to get AI to write the xml necessary for incorporating new pbp. For example - give me 100 pbp examples of a 6 4 3 double play. Or - give me 25 example of Vin Scully calling a 6 4 3 double play, etc. Could create an entire library easily.

[crashsports1]

Does anyone, who did any of these aforementioned posts, have any sort of saved templates or steps they took that they can share for someone who wants to get started on trying this quickly but had a whale of a time trying to get their own detailed one launched through chatGPT pro lol

[luckymann]

I'll readily concede that those questioning the validity of these stats may well have a point, and that is undoubtedly the true benchmark of how useful this sort of analysis is, if at all.

That notwithstanding, I still regard them as a great novelty and will continue to incorporate them at the secondary level for my roster management.

I have just finished the first full season of the league I had these reports generated for, and went back to the agent for an update, as well as an introspective analysis of how the earlier reports held up at the end of the season. I thought I'd share them, without really having digested them myself.

First, here's its full end-of-season roster analysis.


Catchers

Frank House is one of the more valuable position players in the organisation because the bat and the position still line up favourably, even if the total package is not flawless. In 507 PA he has produced a 110 wRC+, .350 OBP, .462 SLG and .353 wOBA for 2.5 WAR, which is comfortably above-average offence for a catcher. The underlying contact quality supports that line rather than undermining it: his .522 xSLG and .350 xwOBA are almost dead on the actual output, while the 10.9% barrel rate and 33.1% hard-hit rate confirm real thump. He is not just selling out for power either, with a manageable 29.9% chase rate and a solid 80.8% Z-Contact rate that keep the offensive floor stable. Defensively, the picture is more mixed. The .999 fielding percentage and only one error in 914 total chances are strong on the surface, and his 2.9 ZR suggests playable value, but the -2.0 framing mark means there is some hidden defensive giveback behind the plate. The front-office read is that House is still a net asset because offence like this at catcher is hard to replace, but the framing number matters and stops him from looking like a truly complete cornerstone.

Infielders

Freddie Freeman is giving you substandard first-base offence, and the defensive contribution is not enough to rescue the overall profile. In 662 PA he has posted an 88 wRC+, .325 OBP, .397 SLG and .319 wOBA, which is simply too light for a bat-first position. The expected line offers almost no relief: .381 xSLG and .287 xwOBA say the current results are largely deserved. His 84.2 EV, 5.6% barrel rate and 23.1% hard-hit rate tell the same story: not enough damage on contact. He is not striking out excessively, and the 29.6% chase rate is acceptable, but the 74.0% Z-Contact rate is weak for a player without carrying power. Defensively, he is actually a negative at first, with -3.2 ZR and a 0.985 efficiency mark despite a clean-looking .995 fielding percentage. That matters because first base only works if the bat clearly carries the spot or the glove is a plus. Right now he is providing neither. This is a below-average regular profile, and the position makes the weakness more damaging.

Jolbert Cabrera looks like depth rather than a player to rely on. In 154 PA he has an 81 wRC+, .312 OBP, .384 SLG and .307 wOBA, and the quality-of-contact indicators do not point to hidden upside. His .380 xSLG and .273 xwOBA essentially validate the weak offensive line, while the 81.7 EV, 4.9% barrel rate and 17.1% hard-hit rate show limited impact. He does make enough contact to avoid a collapse, with a 77.0% Z-Contact rate, but the 33.3% chase rate and lack of on-base skill keep the profile light. Defensively he is passable but not a separator. The 0.5 ZR and 1.002 efficiency are fine, but they do not elevate him beyond reserve value. The overall read is that Cabrera is playable depth who does not bring enough offence or enough defensive edge to claim a larger role.

Enrique Hernández remains a useful supporting piece because the profile is broad rather than loud. In 589 PA he has delivered a 95 wRC+, .347 OBP, .391 SLG and .329 wOBA for 2.2 WAR. The expected metrics are a little kinder than the actual line, with .435 xSLG and .319 xwOBA, which suggests the bat may have a touch more in it than the surface numbers show. His quality-of-contact profile is respectable rather than explosive, with 86.0 EV, 8.3% barrel rate and 27.1% hard-hit rate, and the approach is sound with a 27.0% chase rate and 77.1% Z-Contact rate. The key extra value comes from versatility and competence in the field. At second he is close to neutral by ZR at 0.1, his efficiency is a positive 1.006, and the arm grade is a solid 1.7. He is not a standout defender, but he is not giving away value there either. This is a classic support player: roughly league-average offensive process, multi-area usefulness, and enough defensive stability to justify everyday deployment if needed.

Jonathan Schoop is one of the clearest liabilities in the infield mix. In 340 PA he has produced just a 54 wRC+, .282 OBP, .305 SLG and .266 wOBA, which is well below playable regular level. The expected line is slightly better, with a .346 xSLG and .270 xwOBA, but not nearly enough better to change the conclusion. His 83.2 EV, 3.4% barrel rate and 21.3% hard-hit rate show very little impact, and while the 81.5% Z-Contact rate is actually good, it is mostly empty contact because the damage is not there. Defensively he does not compensate for the bat. The -0.7 ZR and 1.013 efficiency make him roughly serviceable, but not enough to carry a sub-replacement offensive line. The front-office takeaway is that this is a replaceable bat with limited hidden value.

Travis Jackson is still the most important developmental bat in the infield because there is real power here, even though the overall offensive line is not yet fully formed. Over 599 PA he has produced a 94 wRC+, .306 OBP, .478 SLG and .328 wOBA for 2.2 WAR. The power is real: .457 xSLG, 19 home runs implied by the extra-base profile, 8.6% barrel rate and 26.6% hard-hit rate. The issue is offensive completeness. His .262 xBA and .306 xwOBA show a hitter whose slugging carries too much of the profile, and the 31.1% chase rate with only 74.5% Z-Contact explain why he does not convert more of that raw impact into overall value. Defensively, though, he adds a lot. The 8.9 ZR and 1.034 efficiency are excellent for third base, even if the 20 errors and .956 fielding percentage look messy. That split matters. The glove may not be clean in the traditional sense, but the range is clearly creating value. This is exactly the kind of player whose full worth is easy to underrate if you look only at batting average and errors. He is not yet a finished hitter, but the bat-glove combination already looks like a real long-term asset.

Rick Renick is a fringe profile at present. In 115 PA he has managed a 73 wRC+, .296 OBP, .398 SLG and .295 wOBA, with expected metrics that are only modestly better at .419 xSLG and .283 xwOBA. The shape of the offence is narrow. He has some impact in the bat, with an 8.8% barrel rate and 23.8% hard-hit rate, but the 36.9% chase rate is a major issue, and without enough overall contact quality the power cannot carry the line. Defensively the picture is poor. His -3.3 ZR, 0.938 efficiency and .944 fielding percentage at third base all point in the same direction. Unlike Jackson, he is not buying back those errors with plus range. The result is a player who is currently below water on both sides of the ball. There is enough raw damage in the bat to keep monitoring him, but the present value is weak.

Dave Bancroft is the standout all-around position player in the organisation and the clearest example of a player whose defensive value meaningfully lifts the offensive line. In 686 PA he has produced a 122 wRC+, .369 OBP, .489 SLG and .371 wOBA for 5.7 WAR. Offensively he is the cleanest complete infield profile in the file. The expected metrics are a touch lower at .435 xSLG and .325 xwOBA, so some overperformance is possible, but the process is still strong: 86.0 EV, 7.3% barrel rate, 25.8% hard-hit rate, just a 27.2% chase rate and a healthy 80.1% Z-Contact rate. He is creating real offence through a mix of contact, discipline and enough playable power. Defensively he is a major asset. The 8.7 ZR and 1.070 efficiency at shortstop are elite markers in this context, and they outweigh the 19 errors and .965 fielding percentage. The traditional fielding line makes him look more erratic than he really is. In reality, the range is driving substantial value. This is a genuine core player because he gives you above-average offence at shortstop and plus run prevention at the same time.

Billy Hunter is a reserve-only bat, but the defence gives him a more legitimate floor than some of the other fringe infielders. In 105 PA he has produced only a 38 wRC+, .238 OBP, .324 SLG and .242 wOBA, which is obviously non-viable if you are looking for offensive contribution. At the same time, the expected line is notably better than the results: .372 xSLG and .261 xwOBA are still weak, but not catastrophic. The offensive process is peculiar. His 86.6% Z-Contact rate is excellent, but it comes with almost no impact, just 78.7 EV and a 2.2% barrel rate, so he is putting balls in play without doing enough with them. Defensively he has real utility. The 3.4 ZR and 1.067 efficiency at shortstop are good, and that matters for a reserve middle infielder. The front-office read is that he should not be evaluated as a bat-first option at all. He is a glove-leaning depth infielder whose offensive line is too poor for regular work but whose defensive value makes him rosterable in the right role.

Outfielders

Dusty Baker is a promising young corner bat whose early line should be taken seriously, but with some caution because the sample is still limited. In 169 PA he has produced a 120 wRC+, .373 OBP, .465 SLG and .368 wOBA for 1.0 WAR. The expected line is a bit lower at .445 xSLG and .312 xwOBA, which suggests the current production may be running slightly hot, but not outrageously so. The contact quality is decent, with 84.8 EV, 5.9% barrel rate and 27.4% hard-hit rate, while the 28.4% chase rate is manageable. Defensively, though, he is helping himself. The 1.9 ZR, 1.048 efficiency and positive 1.2 arm mark in left field are all encouraging. This is not just a bat-only prospect getting hidden in a corner. The profile suggests a player who may settle as a useful everyday left fielder rather than a star, but there is enough here to remain optimistic.

Al Smith is one of the clear lineup drivers, and unlike some high-end offensive lines, this one is not purely smoke. In 384 PA he has put up a 136 wRC+, .392 OBP, .516 SLG and .393 wOBA for 2.5 WAR. The expected numbers are lower at .442 xSLG and .347 xwOBA, so there is some overperformance in the current line, but the process is still robust enough to support a strong hitter: 86.1 EV, 7.6% barrel rate, 24.8% hard-hit rate, and a solid chase-contact mix of 28.6% OS% and 77.9% Z-Contact. He is not an explosive tools monster, but he is a very functional offensive player. Defensively he is closer to average than plus. The 0.5 ZR and 1.011 efficiency are fine, but the -1.1 arm value limits the overall fielding contribution. The read here is that Smith’s value is driven primarily by the bat, but the glove is stable enough not to create problems. He is a real asset, even if the current offensive line may be a little ahead of true talent.

Lorenzo Cain is a good example of a player whose overall value exceeds the offensive line because the centre-field defence is carrying real weight. At the plate, in 497 PA, he has been a below-average hitter with a 79 wRC+, .312 OBP, .376 SLG and .306 wOBA. The expected line is similar, with .350 xSLG and .270 xwOBA, so there is no hidden breakout case here. He hits the ball softly at 80.0 EV, with just a 3.6% barrel rate and 18.2% hard-hit rate, and the 73.6% Z-Contact rate is too low for a low-power hitter. Offensively, he is a bottom-half bat. Defensively, however, he is outstanding. The 10.3 ZR and 1.048 efficiency in centre field are among the best marks in the file, and they transform the overall player value. This is not a hitter you want to evaluate in isolation. He is a defence-first centre fielder whose glove makes him playable despite the weak bat.

Phil Cavarretta sits in the middle ground as a playable but unspectacular centre-field option. In 391 PA he has a 99 wRC+, .327 OBP, .461 SLG and .336 wOBA, which is basically league-average offensive production. The expected line is a bit lighter at .420 xSLG and .298 xwOBA, suggesting the current line may be slightly ahead of the true process. His contact quality is decent but not special, with 84.2 EV, 6.4% barrel rate and 24.2% hard-hit rate, while the 30.3% chase rate is a little loose. What keeps him from looking more attractive is the defence. The -4.6 ZR and 0.964 efficiency in centre field are poor, which is a serious issue at that position. So while the bat is more playable than Cain’s, the defensive giveback is substantial. The overall takeaway is that Cavarretta looks better as a corner-capable outfielder or depth piece than as a true centre-field solution.

Leon Culberson is a fringe offensive profile whose value depends heavily on how much you trust the defence. In 176 PA he has a 64 wRC+, .290 OBP, .345 SLG and .281 wOBA, with expected metrics of .324 xSLG and .229 xwOBA that say the weak line is deserved. He has very little impact contact, with 80.0 EV, 3.6% barrel rate and 15.1% hard-hit rate, and the 32.3% chase rate undercuts an already light offensive base. Defensively, however, he is a useful centre-field contributor, with 4.6 ZR and 1.067 efficiency. That means there is some real run-prevention value here, even if the bat is poor. He fits as a glove-forward reserve rather than as a starter, but the fielding competence gives him a reason to stay in the conversation.

Felipe Alou is a fairly stable everyday right fielder whose value is spread across both sides of the ball rather than concentrated in one elite tool. In 637 PA he has produced a 79 wRC+, .306 OBP, .402 SLG and .305 wOBA for 0.7 WAR, which is underwhelming offensively for a corner outfielder. The expected line is modestly better, with .432 xSLG and .294 xwOBA, and the contact quality is not bad at all: 86.4 EV, 6.4% barrel rate and 26.5% hard-hit rate. The issue is that the overall offensive approach is too thin, with a 31.7% chase rate and only middling on-base skill. What pushes him back toward playable value is the glove. His 7.2 ZR, 1.062 efficiency and 1.5 arm mark in right field are all strong, and they meaningfully improve the total player picture. The front-office view is that Alou is better than the batting line looks because the right-field defence is real, but the bat still needs to be better to make him a secure long-term regular.

Heavy Johnson is one of the clearest offensive cornerstones in the file, but the defensive component is a real drag that has to be acknowledged. In 618 PA he has hammered his way to a 144 wRC+, .371 OBP, .610 SLG and .405 wOBA for 4.1 WAR. This is legitimate impact offence. The 88.1 EV, 14.0% barrel rate and 32.5% hard-hit rate are the loudest raw power indicators on the roster, and although the 31.5% chase rate and 68.2% Z-Contact rate create volatility, the damage is so strong that it still works. The expected line is lower at .495 xSLG and .328 xwOBA, so he is overperforming to some degree, but even the expected profile still describes a dangerous hitter. Defensively, though, there is a major problem. He is being used outside his natural comfort area, and the file shows it: -1.6 ZR, a dreadful 0.784 efficiency, -0.3 framing, and a slightly negative arm value. Whether that positional code reflects an awkward deployment or just broad defensive weakness, the takeaway is the same: the bat is carrying the player. He is still a core asset because the offence is that strong, but he needs to be protected defensively wherever possible.

Front-office summary

There are four clear core contributors here. Bancroft is the best all-around player on the roster because he combines a 122 wRC+ bat with plus shortstop defence. Heavy Johnson is the premier impact bat, even if the glove gives some value back. House remains a meaningful advantage bat at catcher, and Al Smith is a reliable middle-order support piece whose offensive line is strong enough to matter even with some regression risk.

The next tier is your stable support group. Jackson is not yet a finished hitter, but the power-plus-defence combination makes him a high-value regular already. Hernández is the classic support player who does a lot of things competently and holds the roster together. Cain and Alou are both more valuable than their bats look because the outfield defence is carrying real weight. Dusty Baker belongs in this group as a promising developing regular, though with less certainty because of the sample.

The main regression or role-risk group is smaller here than in some previous files. Smith is probably the clearest candidate for some offensive pullback relative to the current line, though not a collapse. Cavarretta is playable, but the centre-field defence means the role fit matters. House’s framing is the one defensive caution that slightly tempers an otherwise strong profile.

The liabilities are Freeman, Schoop, Cabrera and Renick, with Hunter and Culberson fitting as glove-first reserves rather than regular answers. Freeman is the most significant problem because first base demands more offence than he is providing, and the glove is not saving him. Schoop is simply not hitting enough to justify regular at-bats. Cabrera is thin in both offensive upside and defensive impact. Renick still has some raw damage in the bat, but right now both the offensive process and the third-base defence are below standard.

The decision-oriented takeaway is to build the everyday core around Bancroft, Heavy Johnson, House and Smith, while continuing to invest in Jackson as a long-term pillar. Hernández, Cain, Alou and Baker are useful supporting pieces if deployed correctly. The pressure points in the roster are first base and the lower-end infield depth, where the current offensive output is not good enough.


Starting Pitchers

Don Sutton is the unquestioned staff anchor and the clearest core arm in the organisation. In 256.1 IP he has delivered 8.8 WAR and 9.2 rWAR with a 154 ERA+, 3.18 FIP, 64 FIP-, 3.48 SIERA and 2.52 xERA. That is ace-level performance by both results and estimators. What makes the line especially convincing is that there is no obvious fluke carrying it. He misses enough bats with a 20.8% K rate, barely walks anyone at 4.5%, and therefore posts a strong 16.4% K-BB%. He also suppresses damage extremely well, allowing just a 6.8% HR/FB, a 4.9% barrel rate and 22.6% hard-hit rate. The batted-ball shape is excellent for a frontline starter, with a 53.7% ground-ball rate against only 25.0% fly balls. He is not a pure overpowering strikeout monster, but he combines command, weak contact and ground-ball shape at a very high level. This is a genuine No. 1 starter profile with no real qualification needed.

Frank Tanana is the second core starter and, if anything, the more electric arm from a bat-missing standpoint. In 151.0 IP he has produced 4.0 WAR and 5.1 rWAR with a 158 ERA+, 3.77 FIP, 76 FIP-, 3.46 SIERA and 2.68 xERA. The strikeout profile is the standout feature: a 23.4% K rate, 15.9% K-BB% and 37.4% whiff rate, all of which are the best marks in this rotation outside of Sutton’s overall completeness. The 92.8 fastball velocity also gives him the best pure velocity base among the starters. The walk rate at 7.5% is acceptable rather than elite, and the 70.3% Z-Contact allowed tells you hitters are not comfortable against him even when they swing in the zone. The contact profile is also strong, with 82.5 EV, 5.3% barrel rate and 21.6% hard-hit rate. The main thing keeping him just behind Sutton is a bit more volatility in command and a slightly higher .311 BABIP, but the overall read is clear: this is a frontline starter whose stuff-driven profile is fully supported by the estimators.

Carl Hubbell looks like a dependable mid-rotation starter whose true talent may actually be a little better than the surface line first suggests. Over 199.2 IP he has logged 2.1 WAR and 3.2 rWAR with a 100 ERA+, 5.16 FIP, 105 FIP-, 4.22 SIERA and a much better 3.33 xERA. That spread is the key to the whole profile. The overall strikeout-walk line is fine rather than dominant, with a 16.5% K rate, 5.8% BB rate and 10.7% K-BB%, but he keeps himself afloat by limiting damage better than the ERA+ line alone would imply. The 83.3 EV, 6.7% barrel rate and 24.4% hard-hit rate are all respectable, and the 47.5% ground-ball rate gives him a useful shape foundation. The obvious concern is the 34 home runs allowed and 16.9% HR/FB, which look high relative to the rest of the contact profile. That is why the xERA is so much kinder than the FIP. The front-office view is that Hubbell is a trusted rotation piece, probably more stable than his top-line run prevention suggests, but still a tier below the two true impact starters.

Bill Holland profiles as a back-end innings arm whose baseline skills are weaker than his usage load would ideally call for. In 146.2 IP he has posted 0.6 WAR and 0.3 rWAR with an 85 ERA+, 5.74 FIP, 117 FIP-, 4.52 SIERA and 4.25 xERA. There is no hidden ace case here. The strikeout line is light at 15.1%, and while the 6.6% walk rate is solid enough, the resulting 8.5% K-BB% is too modest for a starter who is also giving up 29 home runs. His damage suppression is middling at best, with 84.6 EV, 8.6% barrel rate and a 23.6% hard-hit rate that is not terrible in isolation, but the 18.9% HR/FB turns those mistakes into too many runs. The ground-ball rate at 42.6% is playable, but not strong enough to offset the home-run problem. This is a usable fifth-starter or swingman profile, but not a pitcher you want overexposed against stronger clubs.

Milt Pappas is the hardest starter to place because the run prevention and expected results pull in different directions. In 64.1 IP he has a 113 ERA+, 4.82 FIP, 98 FIP-, 5.13 SIERA and 3.29 xERA with 0.9 WAR and 1.5 rWAR. The reason the profile feels unstable is that he does very little to miss bats, carrying just a 13.4% K rate and a weak 3.6% K-BB%, yet the contact allowed is fairly soft at 82.1 EV, 5.7% barrel rate and 22.5% hard-hit rate. He also keeps the ball on the ground enough at 49.8% to avoid too many disaster innings, and the 7.5% HR/FB is supportive. The issue is that the 9.7% walk rate leaves him very little margin for error given the lack of strikeouts. This looks like a soft-contact starter whose current line may be a bit better than the true process, but whose xERA still gives some reason to think there is more here than the K-BB line would suggest. I would view him as a fragile depth starter rather than a secure long-term piece.

Relievers

Huston Street is the clear bullpen anchor by role, but he is not as dominant as a 30-save closer label might imply. In 54.0 IP he has recorded 30 saves with a 134 ERA+, 4.74 FIP, 96 FIP-, 4.00 SIERA and 3.20 xERA, alongside 0.4 WAR and 1.2 rWAR. The overall shape is solid. His 18.7% strikeout rate and 6.7% walk rate give him a healthy 12.0% K-BB%, and the .242 BABIP, 83.3 EV and 23.4% hard-hit rate indicate decent contact suppression. The issue is that the home-run damage is a little high for a closer, with 9 HR allowed and a 14.5% HR/FB. His fly-ball rate is also 37.1%, which adds some ninth-inning volatility. So while Street is clearly the most established end-game arm in the file, the underlying profile is more “good closer” than “elite shutdown reliever.” He can hold the role, but he is not untouchable.

Aaron Fultz is probably the steadiest support reliever in the bullpen and one of the better all-around non-closing arms. Across 57.0 IP he has a 120 ERA+, 4.37 FIP, 89 FIP-, 3.98 SIERA and 3.29 xERA, with 0.5 WAR and 0.8 rWAR. He misses enough bats to matter at a 19.9% K rate and 32.0% whiff rate, and although the 8.0% walk rate is not pristine, the 11.9% K-BB% is clearly playable. The contact profile is also stable: 85.6 EV, 8.0% barrel rate, 23.5% hard-hit rate. There is no one overwhelming carrying trait, but there is also no glaring weakness beyond ordinary reliever volatility. He looks like a legitimate seventh- or eighth-inning support arm rather than just bullpen filler.

Will Harris fits alongside Fultz as a useful support reliever, though with a slightly weaker top-line profile. In 72.1 IP he has a 104 ERA+, 5.08 FIP, 103 FIP-, 4.10 SIERA and 3.29 xERA with 0.1 WAR and 0.2 rWAR. The interesting thing here is that his expected indicators are a bit better than the actual run prevention line. He has a respectable 19.2% K rate, 8.1% walk rate and 11.0% K-BB%, and he suppresses contact better than most of the bullpen at 82.7 EV with just a 6.3% barrel rate. The concern is that the ball still leaves the yard too often, with 12 home runs allowed and a 16.9% HR/FB, despite a solid 47.7% ground-ball rate. That suggests some underperformance on the damage side. This is not a premium leverage arm, but it does look like a credible middle-relief piece whose line may be a bit worse than his actual skill level.

Al Hrabosky is a volatility arm with real upside and equally real risk. In 38.2 IP he has a superb 211 ERA+, but the estimators are much more cautious at 4.16 FIP, 84 FIP-, 4.40 SIERA and 3.85 xERA. The good part is clear. He gets whiffs at a 36.0% rate, keeps the ball in the park with only 3 HR allowed and a 7.0% HR/FB, and allows only 83.0 EV. The problem is that the process is not especially clean underneath the results. His strikeout rate is 19.9%, but the 11.5% walk rate drags the K-BB% down to just 8.3%, and the 10.3% barrel rate says hitters can still do damage when they square him up. He is one of the more interesting bullpen arms because the raw ingredients for real value are there, but I would still treat him as a development bet rather than a locked-in high-leverage weapon.

Justin Duchscherer is one of the biggest problem cases in the bullpen because the role usage looks better than the performance quality. In 56.2 IP he has 23 holds, which implies trust, but the line underneath is poor: 79 ERA+, 6.32 FIP, 128 FIP-, 4.44 SIERA and 4.51 xERA, along with -0.8 WAR and -1.1 rWAR. The strikeout-walk profile is not disastrous, 16.0% K and 7.0% BB for an 8.9% K-BB%, but it is not strong enough for a reliever who is also allowing 13 HR and a 19.4% HR/FB. His 10.1% barrel rate and 35.4% fly-ball rate add to the concern. The .280 BABIP makes the top-line numbers look slightly less ugly than they could have. This is not a hidden-value reliever. It is a leverage-exposed arm whose skills are not supporting the role.

Tony Castillo also sits in the liability bucket, though he is a little less extreme than Duchscherer. In 56.2 IP he has a 75 ERA+, 5.21 FIP, 106 FIP-, 4.75 SIERA and 4.93 xERA. The issue is that almost every line item is merely ordinary or slightly worse. He strikes out only 13.8% of hitters, walks 6.9%, and therefore manages just a 6.9% K-BB%. He also gives up too much damaging contact for a low-strikeout reliever: 86.4 EV, 9.4% barrel rate, 26.6% hard-hit rate and 8 homers. There is no obvious carrying tool here. He is not wild enough to be a meltdown specialist, but not good enough in any area to justify meaningful innings. That is a replaceable bullpen profile.

Guy Bush is another weak relief option whose profile fails in too many places at once. In 38.0 IP he has a 69 ERA+, 5.67 FIP, 115 FIP-, 4.86 SIERA and 5.77 xERA, plus negative WAR and rWAR. The 14.0% strikeout rate is too low, the 8.1% walk rate is not low enough to compensate, and the 5.8% K-BB% leaves him with no real weapon. The contact profile is poor as well: 87.3 EV, 11.9% barrel rate and 29.9% hard-hit rate are among the worst marks in the bullpen. He is not a ground-ball specialist either, sitting at 44.8% GB. This is a replacement-level relief arm with no clear case for improvement.

Andrew Miller is the clearest pure stuff-versus-results disappointment in the bullpen. On paper some of the raw ingredients are attractive: 96.3 fastball velocity, 34.6% whiff rate and a strong 49.7% ground-ball rate. But the actual profile is a mess. In 49.2 IP he has a 71 ERA+, 6.72 FIP, 137 FIP-, 5.48 SIERA and 5.51 xERA, with -0.8 WAR and -0.8 rWAR. The killer is control. He walks 15.7% of hitters, strikes out only 15.3%, and therefore has a negative -0.4% K-BB%, which is almost impossible to survive with. Add in 9 homers, a 22.0% HR/FB and a 31.1% hard-hit rate, and the profile becomes untenable. This is an arm with live stuff but currently non-functional execution. Unless the command changes materially, he is not a trustworthy roster piece.

Patrick Sandoval is another below-standard relief arm, though slightly more salvageable than Miller because the baseline skill set is not completely broken. In 92.2 IP he has an 83 ERA+, 5.81 FIP, 118 FIP-, 4.64 SIERA and 4.46 xERA. The strikeout-walk line is ordinary at best, 16.5% K and 9.6% BB for a 6.9% K-BB%, and the 17 home runs allowed are a major problem. He does some things acceptably, such as 84.9 EV allowed and a 45.6% ground-ball rate, but the 9.5% barrel rate and 18.6% HR/FB show that his mistakes are being punished too often. He looks like a low-leverage innings absorber rather than a real bullpen weapon.

Porter Moss is a fringe depth arm with too little bat-missing ability to project forward confidently. In only 13.2 IP he has a 93 ERA+, 6.00 FIP, 122 FIP-, 6.02 SIERA and 6.21 xERA, so even the small sample gives no reason for optimism. His 7.9% strikeout rate and 9.5% walk rate are both unacceptable for a reliever, producing a -1.6% K-BB%. He does not offset that with special contact management, either, despite only moderate damage indicators. This is a non-roster quality performance line until proven otherwise.

Swingman / Depth

Patrick Sandoval also deserves a small role note because of the 7 starts and 92.2 IP. He can soak up mixed usage, but the skill profile still points to low-end depth rather than a meaningful rotation candidate. The home-run issue and mediocre K-BB line would make him a risky starter exposure, and the lack of a relief weapon keeps him from standing out in the bullpen. He is the definition of coverage depth.

Front-office summary

There are two clear core rotation assets: Sutton and Tanana. Sutton is the full-package ace, driven by command, ground-ball shape and elite damage suppression. Tanana is the more electric No. 2, with the bat-missing profile of a true frontline arm and only modest command risk keeping him just behind Sutton.

Hubbell is the clear next-tier starter and a trustworthy mid-rotation piece. Holland and Pappas are depth-grade starters, with Holland fitting better as a fifth starter or innings arm and Pappas as a more fragile soft-contact option whose weak K-BB line limits confidence.

In the bullpen, Street is the established closer, but more as a good closer than a dominant one. Fultz is the best support reliever, Harris is a credible middle-relief stabiliser, and Hrabosky is the main upside bet with real volatility. Those are the only non-closing bullpen arms who presently look like legitimate positive-value pieces.

The liabilities are Duchscherer, Castillo, Bush, Miller, Sandoval and Moss, with Miller the most glaring problem because the raw stuff is not translating into functional strike-throwing. Duchscherer is particularly dangerous because the hold total may make the role look stronger than the skill. The decision-oriented takeaway is to build around Sutton and Tanana, keep Hubbell in the trusted tier, use Fultz and Street as the main bullpen anchors, and aggressively treat the lower-end relief mix as an area needing upgrades.

[luckymann]

The exact wording I used for this was as follows:

To help me better understand how well calibrated these simulated advanced stats are, can you please compare the earlier reports you did with these newer ones and tell me how the various players panned out compared with the preliminary analysis.

Here's what it fired back:

Calibration - hitters


At a high level, the earlier August 3 read was directionally strong. The end-of-season report did not overturn many of the original judgments. What it mostly did was sort the players into three buckets: the early signal held, the early signal softened, or the early signal was clarified by defence. The biggest lesson is that the simulated advanced stats seem reasonably well calibrated on skill shape. They were generally better at identifying who had real underlying quality, who was skating on thin ice, and who was structurally limited than at forecasting exact magnitude.

Frank House is a good example of the model holding up well. In August he looked like a real offensive asset at catcher, with a 115 wRC+, .463 SLG, .353 xwOBA and .524 xSLG, and the report’s conclusion was that the bat was sustainable. By season’s end he had come back a touch to a 110 wRC+, .462 SLG and .350 xwOBA, but the offensive shape was essentially unchanged, and the newer report still called him a net asset. The big new wrinkle was defensive framing, which slightly reduced the ceiling of the total valuation. So the early offensive read was good; the later defensive layer just refined it.

Freddie Freeman is another case where the early signal proved accurate. The August report called him a real problem because the profile lacked impact for first base: 74 wRC+, .370 SLG, .289 xwOBA, .383 xSLG, low barrel rate, too many ground balls. By year’s end, he had improved from catastrophic to merely bad, finishing at 88 wRC+ and .397 SLG, but the underlying conclusion did not change. The bat still was not good enough for the position, and the defensive data made the overall outlook worse, not better. That is a good example of the stats being properly skeptical early, even though the exact endpoint was a little less severe than the first snapshot implied.

Jolbert Cabrera is one of the clearest wins for the earlier report. In August he looked like a batting-average mirage: 111 wRC+, .343 AVG, .410 BABIP, but only .254 xwOBA and poor contact quality. By the end of the season that profile basically collapsed, down to 81 wRC+, .312 OBP and .307 wOBA, with the newer report explicitly treating him as depth rather than a player to rely on. That is exactly the sort of player advanced stats should catch early, and here they did.

Enrique Hernández landed almost exactly where the first report suggested. The August view was that he was a stable, discipline-driven support bat at 102 wRC+ and 1.8 WAR. By year’s end he had drifted to a 95 wRC+ hitter with 2.2 WAR, and the new report still framed him as a useful supporting piece whose value came from competence, discipline and now defensive versatility. That is a strong calibration outcome: not flashy, but correctly identified as a steady roster stabiliser rather than a breakout or collapse candidate.

Travis Jackson also came through in line with the original read. The first report identified him as the highest-upside infield bat because the power indicators were already real — 19 HR, .494 xSLG, .599 xSLGCON — even though the total offensive line was still incomplete at a 97 wRC+. By the end of the season he had not become a finished hitter, but the later report still treated him as the main developmental upside bat, and the defensive data at third base made the total player even more valuable than the early offensive-only report could show. So the underlying metrics were right about the power foundation, and the later file added that the glove makes the overall asset even stronger than first thought.

Rick Renick is a partial win for the early report, but a good example of where raw damage indicators can tempt you a bit too much. In August, the read was that there was more bat here than the surface line showed because the quality of contact was not awful, but that the profile was trapped by chase, strikeouts and ground balls. By the end of the year, that basically proved true in the negative sense: the bat never unlocked, and once the defensive file was added he looked even worse overall because the third-base defence graded poorly as well. So the early report was right that the profile was fringy and flawed; the only thing it may have slightly overstated was how much hidden offensive upside was realistically accessible.

Dave Bancroft is where the earlier report probably undersold the player a bit. In August he was already viewed as a very playable offensive shortstop with 112 wRC+, good discipline and enough power, but “not a star bat.” By season’s end he had reached 122 wRC+ and 5.7 WAR, and the addition of strong defensive value at short turned him into the standout all-around position player in the organisation. So the offensive signal was correct — the process was real — but the total player ended up stronger than the early offensive-only framing suggested. This is not really a miss by the stats so much as a reminder that once you add premium-defensive value, the total outcome can look much bigger.

Billy Hunter is another case where the broad judgment held. The August report saw a non-viable bat despite some better-looking expected stats. At season’s end he was still a terrible hitter, but the defensive information clarified that he could survive as a glove-leaning reserve. So offensively the early read was right, but the total player was slightly more useful than the offensive report alone suggested.

Dusty Baker is hard to score because the early report explicitly said the sample was too small to treat as actionable. That caution turned out to be appropriate. By year’s end he had regressed from the absurd 174 wRC+ micro-sample to a still-useful 120 wRC+ over 169 PA, and the later report settled on “promising young corner bat” rather than star signal. That is exactly how the early stats should have been handled, and they were.

Al Smith is probably the best example of an early report getting the shape exactly right but not the scale. In August he looked like a real offensive driver who was probably outperforming the underlying contact quality somewhat: 168 wRC+, .444 wOBA, but much lower expected marks. By season’s end he had come back to a still-excellent but more believable 136 wRC+ and .393 wOBA, and the later report framed him as a real asset with some regression risk rather than an illusion. That is almost textbook good calibration. The advanced stats did not say “fake”; they said “real hitter, probably not this hot,” and that is basically what happened.

Lorenzo Cain is another strong hit for the earlier report, though the end-of-season defensive data elevated him. In August he was seen as a below-average hitter whose offensive value depended on defence and baserunning carrying the rest. By season’s end the bat was still weak, 79 wRC+, but the centre-field defence turned out to be strong enough that he became a clearly playable overall regular. So the early offensive diagnosis was right; the full-season fielding data simply completed the picture.

Leon Culberson also tracked the early report well. He looked like a weak-contact, regression-risk bat in August, and by the end of the year he was exactly that offensively, at 64 wRC+ with poor expected numbers. The only real update was that his centre-field defence gave him a legitimate reserve floor. Again, the advanced offensive stats correctly caught the weak bat early.

Felipe Alou is a mixed case. The August report saw a stable, serviceable bat with a little hidden upside if the plate discipline improved. The end-of-season report was less optimistic on the bat, with just a 79 wRC+, but once the right-field defence came in, he still graded as more valuable than the offensive line alone suggested. So if you isolate the hitting forecast, the early report was a bit too generous. If you look at total player value, though, the broad idea that he was playable rather than dead weight still held.

Heavy Johnson is the other big success story for the earlier model. In August he already looked like a genuine middle-order power bat, but one with volatility because of strikeouts, chase and low Z-Contact. By year’s end he had become exactly that: an offensive cornerstone at 144 wRC+ and .610 SLG, but with enough overperformance against expected stats and enough defensive drag that the later report still warned the bat was carrying everything. That is a very good calibration result. The early report correctly identified both the impact and the volatility.

The one big structural difference between the two reports is not that the earlier one got many hitters wrong. It is that the end-of-season version had fielding, which materially changed total value for Bancroft, Jackson, Cain, Alou, Culberson and Hunter, and added important caution on House and Heavy Johnson. That means the offensive advanced stats seem to have been pretty well calibrated for identifying hitter quality, but total player valuation improved a lot once defence was layered in.

So my overall answer is this: the simulated advanced stats look pretty solid. They were especially good at flagging false-positive bats like Cabrera, weak-contact problems like Freeman and Culberson, and real underlying assets like House, Jackson, Smith and Heavy Johnson. Where they were less complete was in telling you total player value without defence. In other words, the offensive metrics seem reasonably trustworthy for skill diagnosis; the biggest source of revision from August to season-end was not that the stats were wrong, but that the defensive information changed how good some of the same players actually were.


Calibration - pitchers


The short version is that the earlier report was again stronger on identifying underlying type than on nailing exact end-state magnitude. It was especially good at separating true core arms from merely decent surface performers, and at flagging bullpen arms whose early ERA lines were more attractive than the actual skill base. The biggest difference from the hitter exercise is that there were fewer defensive-style hidden variables here, so this one is a purer test of whether the pitching indicators held up. In most cases, they did.

Don Sutton is the cleanest confirmation. Earlier, he was identified as the clear staff anchor, a genuine No. 1/No. 2 type, driven by command, ground-ball shape and homer suppression rather than overpowering strikeout volume. That held completely. He finished as the unquestioned ace, and if anything the later report was even more emphatic: 8.8 WAR, 154 ERA+, 64 FIP-, 3.18 FIP, 3.48 SIERA, 2.52 xERA. So the early report did not just get him right directionally, it got the exact shape right. He stayed the stable ace rather than morphing into some other kind of top arm.

Frank Tanana is another major hit for the earlier report. The first read was that he looked like a real front-line starter with the best bat-missing upside in the group, and that if the walks tightened even a bit, the profile could get stronger. That held. In the new report he remained the second core starter, still driven by strikeouts, whiff rate and strong expected run prevention. The exact numbers shifted over the larger sample, but the identity did not. That is very good calibration: the advanced stats correctly flagged him early as more than just a hot short run.

Carl Hubbell also came through very close to the original read. Earlier, he was described as a playable mid-rotation arm whose true talent looked a bit better than the superficial run-prevention line because the contact profile was cleaner than the homer totals suggested. The later report said essentially the same thing. He still was not dominant, still looked more stable than spectacular, and still had that tension between respectable contact quality and too many home runs. So again the early indicators held. If anything, Hubbell is one of the best examples of the model correctly resisting overreaction to the surface line.

Bill Holland tracked almost exactly as projected too. The first report called him a back-end innings arm who could survive on shape and ground balls but lacked the strikeout dominance to be trusted broadly, especially given the homer issue. The new report said essentially the same thing, just with more innings confirming it. He stayed a fifth-starter or innings-eater type rather than breaking out or collapsing completely. That is another sign the simulation’s pitching indicators are doing a good job of identifying real role level.

Milt Pappas is the one case where the earlier report could not really be tested, because at that point there were no innings to evaluate. In the later report he emerged as a fragile soft-contact depth starter with a weak K-BB foundation. That is not really a calibration win or loss so much as a previously missing data point becoming available.

Huston Street was read very well. Earlier, the report said he was a usable closer but not an elite one, with the command keeping him afloat more than true shutdown dominance, and with home-run risk making him more of a hold than an untouchable stopper. The new report said essentially the same thing. He remained the closer, but again more as a good closer than a dominant one. That is exactly the kind of distinction a good model should make, and it held.

Aaron Fultz also held up well. The first report identified him as one of the steadiest support arms in the pen, not overpowering but trustworthy. The later report still had him as the best non-closing support reliever. There was no real shift in role interpretation. That is another good calibration outcome.

Will Harris is similar. The earlier report saw him as a functional relief arm, stable rather than dangerous, and not someone to push too far up the leverage ladder. The later report again landed on credible middle-relief stabiliser, with some evidence that his actual line may even have been a bit worse than his underlying skill. So the early report again got the type right.

Al Hrabosky is a strong example of the early report correctly identifying a volatile upside arm rather than buying the shiny ERA. The first report saw him as a development bet with real whiff and real risk. The later report did the same. He never graduated into a trusted bullpen weapon, but he still looked more interesting than the lower-end relievers because the raw ingredients were there. That is good calibration, especially for a reliever in a small sample.

Justin Duchscherer is one of the clearest wins for the earlier report. At the earlier stage, he had a 127 ERA+ and looked superficially useful, but the report explicitly said the headline line was flattering him and that he was more average than dominant. By the later report, the mask had come off completely: 79 ERA+, ugly estimators, and a full liability read. That is exactly the kind of false-positive reliever advanced indicators are supposed to catch early, and this one did.

Andrew Miller is another very strong calibration success. Earlier, he was already flagged as the clearest problem arm in the bullpen because the stuff was there but the command failure made the profile non-functional. The later report did not soften that at all. He remained one of the most glaring liabilities, still defined by live stuff and unusable strike-throwing. That is a textbook example of the model correctly choosing process over raw stuff.

Patrick Sandoval also tracked cleanly. Earlier, he looked like a deservedly below-average bullpen arm with little hidden upside. Later, that view held. The updated report gave him a little more nuance because of the mixed starter-reliever usage, but the underlying conclusion did not change: he is depth, not a meaningful weapon. Again, a good calibration result.

If I had to isolate where the earlier report may have been a touch off, it would mostly be in degree, not direction. Duchscherer deteriorated more sharply than the first report implied. Fultz and Harris were perhaps a bit more secure as support arms than the early version suggested. And Tanana and Sutton solidified their core-arm status across bigger workloads rather than merely sustaining promising early shapes. But those are refinements. They are not reversals.

So the overall calibration read on the pitchers is very positive. The earlier report got the hierarchy right: Sutton and Tanana as the real pillars, Hubbell as the dependable next starter, Holland as the back-end innings arm, Fultz as the steadiest support reliever, Street as a closer but not an elite stopper, Hrabosky as the upside volatility arm, and Miller and Sandoval as liabilities. The biggest single success was probably Duchscherer, where the advanced indicators correctly warned that the early results were stronger than the skill base.

The biggest general takeaway is that these simulated pitching stats seem to be doing a good job of identifying pitcher type and underlying sustainability, especially when FIP-, SIERA, xERA, K-BB% and contact-quality indicators are all pointing in the same direction.

[luckymann]

There are so many things at play here, but let's be clear, this is AI analysing "questionable" advanced stats generated in a computer simulation.

They couldn't possible be more derivative.

Plus, the agent is comparing the same "questionable" stats in each case, so them tracking really tells us nothing, because they are like for like.

Are they accurate? Useful? Valid? These are the $64k questions I think we'd all like answered to some degree.

Like I said, without having had time to full pore over all of this, I can nevertheless see almost immediately that the pictures these "questionable" advanced stats are painting are really very similar to the sort of thoughts you'd formulate on these players looking only at their trad and modern stats. I see no real anomalies.

But that's just an eyeball test. I'm no analytics guru, let's make sure everyone understands that if it wasn't already painfully obvious. This for me is a fun exercise and nothing more. I know the limitations at play here.

All the same, I hope we can figure out some way of measuring just how valid, if at all, these advanced stats are in an OOTP context.

Again, I'll attach the source csv files for all who want to dive deeper, and am happy to provide more info in the quest for truth in this regard.

G

[luckymann]

Originally Posted by crashsports1

Does anyone, who did any of these aforementioned posts, have any sort of saved templates or steps they took that they can share for someone who wants to get started on trying this quickly but had a whale of a time trying to get their own detailed one launched through chatGPT pro lol

Here are two prompts generated by Chat for you to try, one for the hitters report (which should include defence) and one for pitchers. Hope that gets you rolling.

G



Hitters

Read the attached hitter CSV and produce a Front Office Report - Hitters. Start by stating the row count, column count, and headers. Then group the report by defensive subset, using catchers, infielders, outfielders, and utility players if applicable. Analyse every player individually using the actual stat figures throughout, and explain what those figures indicate rather than listing them mechanically. Judge each player relative to his position. If defensive data is included, incorporate it where it materially affects player value. Write in the tone of a head of analytics briefing a GM, and finish with a concise front-office summary grouping the roster into core contributors, stable support pieces, upside or development bets, regression risks, and liabilities.


Pitchers

Read the attached pitcher CSV and produce a Front Office Report - Pitchers. Start by stating the row count, column count, and headers. Then group the report by role, using starters, relievers, and closer if appropriate. Analyse every pitcher individually using the actual stat figures throughout, and explain what those figures indicate rather than listing them mechanically. Judge each pitcher relative to his role. Focus on run prevention, estimators, strikeout-walk profile, contact quality allowed, batted-ball shape, and role context. Write in the tone of a head of analytics briefing a GM, and finish with a concise front-office summary grouping the staff into core arms, stable support pieces, upside plays, regression risks, and liabilities.

[luckymann]

I suspect most of you are also following this thread on the same topic, but for those who aren't I thought you'd be interested in my calibration report for this from the agent.

Here's my post, check the original thread for context if you feel the urge.

From ChatGPT...

I kept the full file intact, including the two-way players, and then calculated the league environment three ways: with no PA cutoff at all, with a 50 PA minimum for a cleaner player pool, and with a 100 PA minimum as an even more stable reference point. That gives you both the pure league environment and the cleaner signal once the smallest samples are stripped out.

League environment

With no cutoff, the file contains 503 players and 152,819 total PA. The PA-weighted league-average line comes out to .3363 wOBA against .3135 xwOBA, .3426 OBP, .2745 AVG, .4360 SLG, .3035 BABIP, 100.4 wRC+, 101.4 OPS+, 8.83% BB, 15.63% K, 28.80% chase rate, 77.82% Z-Contact, 29.24% whiff rate, 83.78 EV, 114.60 max EV, 10.85 launch angle, 7.43% barrel rate, 23.71% hard-hit rate, and 11.78% HR/FB.

At 50 PA and above, the pool drops to 375 players and 150,769 PA, which means you are barely losing any plate appearances and mostly just removing noise. The weighted league line becomes .3370 wOBA against .3140 xwOBA, .3432 OBP, .2750 AVG, .4370 SLG, .3040 BABIP, 100.8 wRC+, 101.8 OPS+, 8.86% BB, 15.63% K, 28.77% chase, 77.81% Z-Contact, 29.24% whiff, 83.82 EV, 114.71 max EV, 10.85 launch angle, 7.47% barrel rate, 23.79% hard-hit rate, and 11.85% HR/FB.

At 100 PA and above, the numbers barely move again. The weighted line is .3377 wOBA against .3146 xwOBA, .3438 OBP, .2754 AVG, .4383 SLG, .3042 BABIP, 101.3 wRC+, 102.3 OPS+, 8.88% BB, 15.60% K, 28.76% chase, 77.84% Z-Contact, 29.22% whiff, 83.87 EV, 114.78 max EV, 10.89 launch angle, 7.52% barrel rate, 23.90% hard-hit rate, and 11.92% HR/FB.

The important point is that the environment is very stable once you remove the tiniest samples. That is a good sign in itself.

Actual versus expected tracking

This is where the file gets more interesting. Leaguewide, actual production is running ahead of expected production.

With no cutoff, league wOBA is .3363 versus .3135 xwOBA, so the weighted gap is +.0228. At 50 PA and above, the gap is still +.0230. At 100 PA and above, it is +.0231. That consistency matters. This is not a tiny-sample artifact. The sim appears to be producing a league where actual wOBA sits about twenty-three points above xwOBA.

By contrast, slugging is almost perfectly aligned. League SLG is .4360 versus .4309 xSLG with no cutoff, a gap of only +.0050. At 50 PA and above the gap is +.0053, and at 100 PA and above it is +.0053 again. That is very tight.

Batting average also runs a little hot, but not wildly so. AVG is .2745 versus .2672 xBA with no cutoff, a gap of +.0074. At 50 PA and above it is +.0076, and at 100 PA and above it is +.0080.

So if you are looking for the headline takeaway, it is this: the sim’s expected slugging and expected batting average are tracking fairly tightly to the realised league environment, but xwOBA is running meaningfully below actual wOBA at the league level.

Player-level spread

Using the 50 PA pool, the average player-level wOBA minus xwOBA gap is +.0203, and the median is +.0230. The 10th percentile is about -.019 and the 90th percentile is about +.0596. More importantly, 54.4% of players are more than +.020 above xwOBA, while only 9.6% are more than -.020 below it. That is a very lopsided distribution. If the expected model were fully centred, you would normally expect something more balanced around zero.

For SLG minus xSLG in the 50 PA pool, the average gap is only +.0022 and the median is +.0060. The 10th percentile is -.066 and the 90th percentile is +.0706. That looks much more naturally centred. There is real variance, but not an obvious directional bias.

For AVG minus xBA, the average gap is +.0023 and the median is +.0030. The 10th percentile is -.0396 and the 90th percentile is +.0410. Again, that looks much healthier than the xwOBA distribution.

Relationship tracking

This is the other major realism check. I focused on the 50 PA pool here, because it gives you cleaner player-to-player signal.

wOBA versus xwOBA has a correlation of 0.663. SLG versus xSLG is 0.689. AVG versus xBA is 0.549. EV versus xwOBA is 0.673. Barrel% versus SLG is 0.574. BB% versus OBP is 0.597. K% versus Z-Contact is -0.868. Chase rate versus BB% is -0.477.

Those are actually pretty encouraging. The stats are behaving in the right directions, and in most cases with healthy strength. The strongest relationship in the set is K% versus Z-Contact, which is exactly what you would want to see. EV and barrel rate are also mapping sensibly to damage production. So the internal logic of the metrics looks broadly baseball-like.

What this suggests

My read is that the simulation’s modern stats do not look random or decorative. They are interacting in coherent ways. Contact quality is mapping to power. Zone contact is strongly tied to strikeout rate. Chase is negatively tied to walks. Expected slugging is close to realised slugging. Those are all good signs.

The main red flag is xwOBA. On a leaguewide weighted basis it is consistently about twenty-three points lower than actual wOBA, and the player-level deltas are skewed heavily toward overperformance. That suggests either the xwOBA model is too conservative, or the sim is generating enough non-contact or sequencing value that wOBA is systematically outrunning the expected version. Either way, that is the one metric in this file that looks least well calibrated.

The practical verdict

If I were grading this file purely on realism behaviour, I would say the contact-quality and shape metrics look broadly believable, xSLG looks pretty well calibrated, xBA looks acceptable, and xwOBA is the main area that appears biased low relative to realised outcomes.

So the best single sentence summary is this: the stat ecosystem looks internally coherent, but the xwOBA model appears to be understating leaguewide offensive value.

[Guthrien]

Originally Posted by WhiskyTango

I think AI is taking OOTP to the cusp of a new era -- where the stats will be brought to life in ways unthought of. Am surprised no one has as of yet incorporated AI into generating entire scripts of play-by-play options. May even be possible to get AI to write the xml necessary for incorporating new pbp. For example - give me 100 pbp examples of a 6 4 3 double play. Or - give me 25 example of Vin Scully calling a 6 4 3 double play, etc. Could create an entire library easily.

This is where my mind was going as well. It's so much fun having it build a story out of a game log, etc. why not use those powers to freshen and greatly expand the tired language and descriptions in OOTP? I love them, but doubling or tripling some of the color of the game would be amazing. There are certain areas I've seen people do handmade mods for, I imagine there are tips there for someone to try.

[pstrickert]

I could be wrong, but I'd imagine Vin, like most announcers, used a limited number of expressions to describe a 6-4-3 double play. Something out of the ordinary would have to happen for him to introduce a bit of flair. Routine plays, though, generally do not lend themselves to colorful descriptions. Listen to different broadcasters and you'll notice a large amount of overlap when describing routine plays. But who knows? As I said, I could be wrong.

[WavesOfActivity]

Not exactly what is being used here, but in terms of the cool things AI can do -- I have a site tracking my team, and I wanted to create a season summary so I could easily go back and recall what happened years in to the save. It's a very directed prompt, and I export a large number of stats to JSON for it to use, but it is pretty cool.

Sample prompt:

Prompt:
You are a baseball historian summarizing a single season.
Using ONLY the JSON data below, write a concise season recap in a narrative style. The recap should:
1. Read as an article without section headers.
2. Include overall team performance, playoff outcome, top offensive players, top pitchers (starters, relievers, closers), major awards (only mention the month for weekly or monthly awards, not All-Star games, Gold Gloves, Cy Youngs, or Silver Sluggers), notable injuries (only for players who played with the team this season), draft picks, trades, position battles, major roster changes or transactions (include dollar amounts for contracts and the month), and one or two memorable storylines.
3. List trades in bullet points, including the month/date of the transaction and players acquired/lost.
4. Mention draft picks in a short paragraph.
5. Present position battles as part of the narrative.
6. Avoid hyperlinks for player names.
7. Maintain a factual but engaging tone.
8. Target length: 12 paragraphs.
9. Include HTML code to separate paragraph blocks.
Narrative Structure Guidelines:
• Paragraph 1: Regular season overview (team record, division finish, team batting, pitching, and fielding stats).
• Paragraph 2: Playoff outcome (series results, wins/losses).
• Paragraph 3: Top offensive players (4–5 hitters, batting averages, HR, RBI, OBP/SLG, WAR).
• Paragraph 4: Starting pitchers (wins, ERA, strikeouts, WAR).
• Paragraph 5: Relievers and closers (saves, ERA, key stats).
• Paragraph 6: Position battles (who split time, performance).
• Paragraph 7: Awards (monthly/weekly only, integrate into narrative).
• Paragraph 8: Injuries (players who played this season, expected time missed).
• Paragraph 9: Draft picks (include round and pick number).
• Paragraphs 10–11: Trades and major roster moves
○ Pre-April trades in one paragraph.
○ July trades reference trade deadline.
○ Post-October trades as offseason moves.
○ Use bullet points with html code to list trades with month/date and players acquired/lost.
○ Include contract signings with month and dollar amount.
• Paragraph 12: Memorable storylines (rookie emergence, standout performances, key trades).
Style Notes:
• Maintain a narrative flow like a magazine article.
• Paragraphs should be short (3–6 sentences).
• Only trades may use bullet points; all other content flows in narrative text.
• Do not add information not in the JSON.
• Output should be 12 paragraphs, but may extend slightly to include all key events.

2028 season:

The 2028 Atlanta Braves delivered a strong regular season, finishing 93–69 (.574) and narrowly missing the division crown by just one game, settling for second place. The club combined solid all-around performance, hitting .246 as a team while posting an excellent 3.38 ERA and a .988 fielding percentage. Anchored by standout seasons from Michael Harris II and Spencer Strider, Atlanta remained one of the most balanced teams in the league. Their consistency on both sides of the ball kept them firmly in contention from start to finish.

In the postseason, the Braves made a deep run, winning eight games across three rounds before falling just short of a World Series appearance. After advancing through the early rounds, Atlanta pushed the Cincinnati Reds to a full seven-game Championship Series. Ultimately, the Braves were eliminated in a hard-fought 4–3 series loss. The narrow defeat underscored both their resilience and how close the club was to reaching the final stage.

Offensively, Michael Harris II led the charge with a stellar .317 average, 25 home runs, 95 RBIs, and a .887 OPS, producing a team-best 6.6 WAR. Ronald Acuna Jr. added significant power with 33 home runs and 104 RBIs while posting a .279 average and .864 OPS. Drake Baldwin provided value behind the plate with 18 home runs and a 4.1 WAR season, while Austin Riley contributed 21 home runs and 73 RBIs. Steven Kwan rounded out the group with a contact-oriented approach, hitting .254 and providing steady outfield play.

The starting rotation proved to be a major strength, led by Spencer Strider’s 14–8 record, 2.65 ERA, and 219 strikeouts across 183.2 innings. Kyle Harrison was equally dominant, going 14–6 with a 2.44 ERA and an elite 0.95 WHIP. Tarik Skubal added another high-end arm with a 13–7 record, 2.94 ERA, and 231 strikeouts. Spencer Schwellenbach handled a heavy workload despite a higher ERA, while Riley Quick showed flashes in 20 starts before his season was cut short.

The bullpen provided steady support throughout the season, with Ben Joyce emerging as the primary closer and recording 17 saves with a 2.94 ERA and dominant underlying metrics. Joe Jimenez added reliability in high-leverage situations, posting a 2.72 ERA across 61 appearances. AJ Smith-Shawver contributed both as a starter and reliever, while Sam Hentges and Anthony Maldonado offered additional depth, with Maldonado particularly effective with a 1.90 ERA over 62 outings.

One of the key platoons came at shortstop, where Jacob Wilson and Aeverson Arteaga split time due to Wilson missing significant time. Wilson ultimately handled the majority of starts, hitting .274 with a .691 OPS and providing slightly below-average defensive metrics. Arteaga, while stronger defensively, struggled offensively with a .181 average and .552 OPS. The contrast between Wilson’s offensive edge and Arteaga’s defensive upside defined the position throughout the season.

Individual performances were recognized throughout the year, with Michael Harris II earning Player of the Week honors in April and again in August. Kyle Harrison’s dominance was highlighted by a Pitcher of the Month award in July, while Tarik Skubal followed with his own Pitcher of the Month recognition in August. Ronald Acuna Jr. capped the regular season with a strong October, earning Batter of the Month honors and reinforcing his role as a middle-of-the-order force.

Injuries played a notable role during the season, particularly in the second half. Jacob Wilson suffered a torn groin in July that sidelined him for three months, impacting the stability of the shortstop position. Riley Quick was also lost in late July due to forearm inflammation, missing three to four months and reducing the rotation’s depth heading into the stretch run.

The Braves added to their future through the 2028 draft, selecting Johnny Guzman in the first round (Pick 30) and Roger Crocker in Supplemental Round 1 (Pick 4, 36th overall). Both players represented key additions to the organization’s pipeline, addressing positional depth and long-term upside. The draft strategy reflected a continued focus on building sustainable talent.

Atlanta was active early in the year with a mix of trades, signings, and international additions that reshaped the roster. Key moves included:

January 16: Acquired Paxton Kling, Manuel Beltre, and Tucker Toman from Toronto; traded Ryan Ritter.
January 16: Signed Devin Williams to a 1-year, $6,200,000 contract.
January 16: Signed A.J. Minter to a 1-year, $2,400,000 contract.
January 22: Signed Lazaro Sierra (Panama) with a $1,100,000 bonus.
January 24: Signed Haruyoshi Dai (Japan) with a $2,200,000 bonus.
April 19: Acquired Addison Barger from Chicago Cubs (50% retained); traded Austin Hays and Gregory Barrios.
At midseason and beyond, the Braves continued to adjust their roster while also securing key players for the future. Notable moves included:

June 30: Acquired Sam Hentges from San Diego; traded Sean Keys, Dylan Lee, and Brett Sears.
July 16: Signed international amateur Rupert Houghton with a $440,000 bonus.
August 17: Extended Ben Joyce (3 years, $16,100,000).
August 28: Extended Ronald Acuna Jr. (6 years, $247,200,000).
October 10: Extended Jacob Wilson (1 year, $4,600,000).
November 7: Extended Addison Barger (1 year, $3,640,000), Anthony Maldonado (1 year, $1,460,000), AJ Smith-Shawver (1 year, $2,580,000), and Nathan Lavender (1 year, $1,150,000).
November 16: Extended Demetrio Crisantes (1 year, $940,000).
December 19: Signed Ryan Walker (1 year, $1,200,000).
December 23: Lost Landen Roupp in the Rule 5 Draft to Kansas City.
The 2028 season will be remembered as one of near-breakthrough, defined by elite pitching, a breakout campaign from Michael Harris II, and a deep postseason run that came within one game of the World Series. The aggressive midseason acquisition of Sam Hentges and the blockbuster extension for Ronald Acuna Jr. signaled a franchise committed to both present contention and long-term stability. While the Championship Series loss left a sense of unfinished business, the foundation established during the season positioned Atlanta as a formidable contender heading into the future.

[luckymann]

Nice work WoA, that's one of the less openly AI reports of that nature I have seen. The direct prompt is definitely the way to go.

I understand that with limited manpower and resources, a full pivot to AI is beyond OOTP, and I don't even know that I want it introduced directly. But I feel the devs should really give serious consideration to creating a series of interfaces within the game for referencing the data to an external AI agent.

The Almanac is simply too unwieldy, but if there was a page within OOTP on which preset reports could be accessed and perhaps / hopefully custom ones created for this express purpose, then that would be a superb compromise that I think would be hugely popular with users, and increasingly so as time goes on. This method also allows those players who have no interest in using AI to remain unaffected.

Anyhoo, thanks for the input.

G

[WhiskyTango]

Originally Posted by pstrickert

I could be wrong, but I'd imagine Vin, like most announcers, used a limited number of expressions to describe a 6-4-3 double play. Something out of the ordinary would have to happen for him to introduce a bit of flair. Routine plays, though, generally do not lend themselves to colorful descriptions. Listen to different broadcasters and you'll notice a large amount of overlap when describing routine plays. But who knows? As I said, I could be wrong.

I used inflated numbers to sell the idea.

[WhiskyTango]

btw - chatgpt just announced free and go versions will contain ads.