Autocalc not respecting IBB target

[Joelman111]

I'm working on creating a fictional league with desired league total outputs. I've done this by editing the "era_stats.txt" file with my desired outputs for every year, then running autocalc every season. I set the IBB total to 645 (in 133k ABs), but I noticed that IBB modifier is always set too low and never reaches the target (roughly 150 IBB walks over the course of the year). I also noticed that the BB league total shown on the League Settings page means unintentional walks only because it takes the BB number in "era_stats.txt" and subtracts the IBB number to set that league total. I have attached a spreadsheet of my calculations, does anyone know what I'm doing wrong here?Era Stats Calculator.xlsx

[snepp]

I'm going to pull this one out of left field, because I've got nothing else.

Any chance this is somehow related to the managers and the team strategies just not favoring intentional walks? Somewhat similar to whatever was going on with steals in historicals when coaching was enabled?

[Joelman111]

Originally Posted by snepp

I'm going to pull this one out of left field, because I've got nothing else.

Any chance this is somehow related to the managers and the team strategies just not favoring intentional walks? Somewhat similar to whatever was going on with steals in historicals when coaching was enabled?

If that were the case, I would expect it to change when I run independent sims, which hasn't happened yet. Plus, I would hope OOTP would take that into account when autocalc determines what the modifier should be. It simulates 3 seasons, which I would expect to include the managerial decisions, and determines the modifiers based on league stat outcomes.

[snepp]

Originally Posted by Joelman111

If that were the case, I would expect it to change when I run independent sims, which hasn't happened yet.

Ok, wasn't sure how deep you had gone on that yet.

Plus, I would hope OOTP would take that into account when autocalc determines what the modifier should be. It simulates 3 seasons, which I would expect to include the managerial decisions, and determines the modifiers based on league stat outcomes.

Wasn't that part of the issue with the stealing, that coaching wasn't accounted for and threw things way off? I didn't pay too close attention to what was going on with that.

[Joelman111]

Originally Posted by snepp

Wasn't that part of the issue with the stealing, that coaching wasn't accounted for and threw things way off? I didn't pay too close attention to what was going on with that.

Not totally sure what you are referring to if you have a link to that thread. I'm not into historical sims, so I never dove into that.

[kq76]

I'm not sure, but there's a few things to note:

-OOTP only separated out iBB fairly recently, like OOTP26. I actually didn't even notice this until a few weeks ago. And when I copied over my era_stats.txt file from OOTP25 to OOTP26 I never added an iBB column like OOTP26 has, but the ratios I got for it (OOTP26) were still really good, even before tweaking the modifiers. So, if all else fails, maybe get rid of your iBB column and see if that helps. Crazy, I know, but maybe it'll help.

-iBB/G has varied quite a bit over the years, from .4 to under .1 in recent years. You seem to be targetting .166 which is reasonable. And you're right, 150 walks over 3888 games, which is .0385 iBB/G, is not reasonable.

-Your PA/AB ratio (1.111) is *maybe* a little low from it should be. It's not crazy low (it has been that low before, but historically it's about 1.123), but considering how small iBB is, I right away thought it could be a problem with PAs and ABs.

-Were you starting from a baseline year and tweaking it from there like I do or are you making it all up from scratch? I ask because it might help to know what are targets are and understand how plausible it likely is.

-Your AB formula is missing an "other" stat. I mean, yeah, you're subtracting AB+BB+HB+SF+SG from PA, but you're not subtracting interference and obstruction. Are those so small that they shouldn't make a difference? Maybe, but a number of small things could make a difference. I use .005 for my O/G.

-I find it odd that you're setting your ratios to /IP instead of to /G. I mean, yeah, I guess you could do that, but it seems to me that it's unnecessarily complicating matters. And yes, I understand that games aren't perfect 9 innings per, they're less than 9, and it looks like you are factoring that in, but I just don't think it's necessary to put them over IP rather than G. Which leads to...

-I'm really confused by how you're calculating PA. It seems to be (IP_Outs+(H+BB+HB+CI-RunnerOuts)*9IPGames+ROE)/9IPGames? I've never seen anything like that before. Your way might work perfectly fine, but I just think mine is so much simpler and I find, in general, the simpler you do things the less chances for mistakes. Maybe my way is too simple, but I first calculate AB = H + RBOE + RBOFC + non-sac Out (including SO) and then PA = AB + BB + HB + SF + SH + Other, whereas you obviously calculate PA first and then calculate AB from PA. I'm trying to make sense of your formula in my head and it's a struggle. It might be perfectly valid, but I can't help but wonder if something isn't quite right about it.

I don't know. What you might be doing might be perfectly fine and it's just something with the game, but you're calculating so many things and in such a complicated way that I don't really want to try harder to make sense of it all. Maybe take a step back and try to simplify it? When I think about league totals and modifiers in general, what I think is important is, "is everything plausible?" Meaning, PA can't be 2xAB, iBB can't be 1/G, etc. And no, I'm not saying you have done that (I didn't look at it all enough to judge whether you have or not), but I'm just saying, if I put my ideal totals line in the middle somewhere of rows of historical totals (or my ideal ratios among historical ratios), would my row appear plausible or is there just something about it that is just not plausible? And so I'm still wondering, is there something about your PA and AB #s that are squeezing out iBB? Maybe there's just simply not enough room for it given your other numbers.

[Joelman111]

Originally Posted by kq76

OOTP only separated out iBB fairly recently, like OOTP26. I actually didn't even notice this until a few weeks ago. And when I copied over my era_stats.txt file from OOTP25 to OOTP26 I never added an iBB column like OOTP26 has, but the ratios I got for it (OOTP26) were still really good, even before tweaking the modifiers. So, if all else fails, maybe get rid of your iBB column and see if that helps. Crazy, I know, but maybe it'll help.

I’ll give this a try and report back how the game responds to a missing IBB column.

Originally Posted by kq76

Your PA/AB ratio (1.111) is *maybe* a little low from it should be. It's not crazy low (it has been that low before, but historically it's about 1.123), but considering how small iBB is, I right away thought it could be a problem with PAs and ABs.

I'm really confused by how you're calculating PA. It seems to be (IP_Outs+(H+BB+HB+CI-RunnerOuts)*9IPGames+ROE)/9IPGames? I've never seen anything like that before. Your way might work perfectly fine, but I just think mine is so much simpler and I find, in general, the simpler you do things the less chances for mistakes. Maybe my way is too simple, but I first calculate AB = H + RBOE + RBOFC + non-sac Out (including SO) and then PA = AB + BB + HB + SF + SH + Other, whereas you obviously calculate PA first and then calculate AB from PA. I'm trying to make sense of your formula in my head and it's a struggle. It might be perfectly valid, but I can't help but wonder if something isn't quite right about it.

Yeah I was unsure of the best way to go about calculating PAs from the other given stats. The approach i decided on was to sum total outs and all ways to get on base, and then subtract all the ways the runners on base could get out (since that would be double counting as both an out and a batter that reached base). This can be challenging as the ways runners get out include caught stealing, pickoffs, and being thrown out in double plays or outfield assists. In practice, this led to too many subtractions and the PA total was too low, so I used real MLB data and made up a constant to subtract some amount less of the DPs. I am interested in considering another way, but I have no idea how you would get the data for RBOFC or Other?

Originally Posted by kq76

Were you starting from a baseline year and tweaking it from there like I do or are you making it all up from scratch? I ask because it might help to know what are targets are and understand how plausible it likely is.

I started by looking at the era_stats.txt included with OOTP27. I tried to reverse engineer how each stat was calculated (some of which i did not find an exact formula). I basically am trying to make a league that roughly mirrors the last 15 years of MLB data, so i used averages for much of them (excluding covid year). My desired league totals most closely match 2016 MLB, but i made some changes that made the league leaders more palatable to me.

Originally Posted by kq76

-Your AB formula is missing an "other" stat. I mean, yeah, you're subtracting AB+BB+HB+SF+SG from PA, but you're not subtracting interference and obstruction. Are those so small that they shouldn't make a difference? Maybe, but a number of small things could make a difference. I use .005 for my O/G.

I include CI as a small amount since that seems to be what it is in game (like 16-24 times per season in a 24 team league, but there’s no stat in the era_stats that controls this). I may have forgotten to subtract it for ABs but that’s pretty minuscule. I haven’t found stats for how often obstruction happens, but figured it was so rare to not really matter. Curious what are the events that would be classified as “Other” PAs? Outside of CI and obstruction, I can’t think of anything that isn’t already covered in another category.

Originally Posted by kq76

-I find it odd that you're setting your ratios to /IP instead of to /G. I mean, yeah, I guess you could do that, but it seems to me that it's unnecessarily complicating matters. And yes, I understand that games aren't perfect 9 innings per, they're less than 9, and it looks like you are factoring that in, but I just don't think it's necessary to put them over IP rather than G. Which leads to...

See I initially set this calculator up to use team games as the baseline, but then I considered that other leagues with different rule setups might lead to different amount of IP/G on average. For example, I play with modified extra innings off, but if someone were to turn that on, I suspect IP/G would go down due to shorter extra inning games. Considering this, I felt the most natural choice was to make everything scalable to a 9IP game, which notably is smaller than the team games by about 20. Either choice should likely still yield valid results tho, and it clearly works for all other stat categories- I’m only struggling with IBBs (and defense, but that’s a whole other can of worms).

I believe my numbers should be plausible, but it’s a good note about the PA/AB being small. I didn’t think to check that ratio and it’s certainly the most “hand-wavy” of all my calculations. I went to painstakingly lengths to make mostly accurate formulas for everything, but it’s clear there might still be work to do on this. I’ll give it another try on PAs and ABs and see if that helps reduce the issue.

[kq76]

Originally Posted by Joelman111

Yeah I was unsure of the best way to go about calculating PAs from the other given stats. The approach i decided on was to sum total outs and all ways to get on base, and then subtract all the ways the runners on base could get out (since that would be double counting as both an out and a batter that reached base). This can be challenging as the ways runners get out include caught stealing, pickoffs, and being thrown out in double plays or outfield assists. In practice, this led to too many subtractions and the PA total was too low, so I used real MLB data and made up a constant to subtract some amount less of the DPs. I am interested in considering another way, but I have no idea how you would get the data for RBOFC or Other?

Bold by me.

This might be the key. I started with historical numbers and tried to come up with numbers that look very much like they could be historical numbers themselves and then I just put as many of them as possible into OOTP. EDIT: I don't actually pay that much attention to historical totals themselves. I look at various historical ratios, tweak them, and then convert those tweaked ratios back into my own ideal totals.

So, for example, it's not a problem for me that OOTP might not have certain stats like RBOE or RBOI. Yeah, I'd very much like it if OOTP had those stats and if certain players were better at them than others (Ichiro and Ellsbury), but for league total purposes I just come up with plausible numbers (and for anyone who doesn't know, you can download the lahman database of historical baseball stats from SABR (particularly teams.csv) EDIT: or FanGraphs which has some data lahman doesn't (Batting, Pitching, Fielding)) and then I just input into OOTP what I can. So even though I'm not inputting stats like RBOE, RBOI, and whatever else, I'm still factoring them into my totals and ratios, they're just not inputted into OOTP is all. And so if someone doesn't factor in these other stats into their PAs, the game might not have the "fudge factor" it needs to make everything fit. If OOTP had all the stats as inputtable, it probably wouldn't be a problem, we could just input them all. But since we can't, if we don't leave the engine enough room to do everything it needs to do (IINM, the outcomes are there, the stats just aren't), then it likely needs to take away from something (in this case, iBB) to give something to these other things. And considering iBB was the most recent addition to the era_stats file, I'm not surprised if it's just taking from it instead of something else.

Or maybe I'm totally wrong. It's just a theory.

[Joelman111]

Originally Posted by kq76

Bold by me.

This might be the key. I started with historical numbers and tried to come up with numbers that look very much like they could be historical numbers themselves and then I just put as many of them as possible into OOTP. EDIT: I don't actually pay that much attention to historical totals themselves. I look at various historical ratios, tweak them, and then convert those tweaked ratios back into my own ideal totals.

So, for example, it's not a problem for me that OOTP might not have certain stats like RBOE or RBOI. Yeah, I'd very much like it if OOTP had those stats and if certain players were better at them than others (Ichiro and Ellsbury), but for league total purposes I just come up with plausible numbers (and for anyone who doesn't know, you can download the lahman database of historical baseball stats from SABR (particularly teams.csv) EDIT: or FanGraphs which has some data lahman doesn't (Batting, Pitching, Fielding)) and then I just input into OOTP what I can. So even though I'm not inputting stats like RBOE, RBOI, and whatever else, I'm still factoring them into my totals and ratios, they're just not inputted into OOTP is all. And so if someone doesn't factor in these other stats into their PAs, the game might not have the "fudge factor" it needs to make everything fit. If OOTP had all the stats as inputtable, it probably wouldn't be a problem, we could just input them all. But since we can't, if we don't leave the engine enough room to do everything it needs to do (IINM, the outcomes are there, the stats just aren't), then it likely needs to take away from something (in this case, iBB) to give something to these other things. And considering iBB was the most recent addition to the era_stats file, I'm not surprised if it's just taking from it instead of something else.

Or maybe I'm totally wrong. It's just a theory.

Yeah what you are saying totally makes sense- if there are not enough PAs, it probably defaults to reducing IBBs. But I'm unclear where you found the real stats for RBOE, RBOI, RBOFC, etc. Lahman DB doesn't appear to track this data from what I could tell and I couldn't find it on FanGraphs. Best I could find was searching on Baseball Savant and manually counting it up. I see what I could have missed in my PA formula- reaching on fielders' choice. This could lead to me undercounting PAs. I will try to find a way to estimate this total.

[kq76]

Sorry, I realize I'm not being clear. You don't at all need all these other stats, you just need a PA large enough to accommodate them. So, say you have the equation A=B+C+D and you know what A, B, and C equal, it's okay that you don't know what D is because you can still figure it out by swapping A and D's sides. And in this scenario, A is historical PA, B+C are historical stats like H+BB, which you also know, and D is this group of stats you don't know. But again, you don't need to know the exact values of all that goes into D, you just need the rest of the equation to figure out the total of D.

By the way, in era_stats.txt, PA isn't even there, but BF is. And, unless there's the odd scorekeeping quirk, PA=BF.

I don't have the time right now, but if you don't figure this out soon I'll try to dig into it more because I'm a bit concerned it's not this at all and I'll encounter the same problem when I switch over to 27.

Did you happen to try removing the iBB column? I wonder if it's simply a problem with it and I lucked into the fix by not adding it to my file.

[Joelman111]

Originally Posted by kq76

Sorry, I realize I'm not being clear. You don't at all need all these other stats, you just need a PA large enough to accommodate them. So, say you have the equation A=B+C+D and you know what A, B, and C equal, it's okay that you don't know what D is because you can still figure it out by swapping A and D's sides. And in this scenario, A is historical PA, B+C are historical stats like H+BB, which you also know, and D is this group of stats you don't know. But again, you don't need to know the exact values of all that goes into D, you just need the rest of the equation to figure out the total of D.

Right, I get that. But to adjust for my league settings I was hoping to find a way to estimate the components of D individually rather than just lumping them into an "Other" bucket and assuming OOTP will generate approximately the same amount. For example, ROE changes based on the fielding percentage set for the league, so if I want historically bad defense, it should drastically change the number of ROE I should account for. I am making some guesses for other things like CI, but those should be few enough that they don't really matter. ROFC is something I would need to use historical data and try to get an estimate of.

Originally Posted by kq76

By the way, in era_stats.txt, PA isn't even there, but BF is. And, unless there's the odd scorekeeping quirk, PA=BF.

Yes, I am aware of this. I checked historical MLB data and PA=BF every year.

Originally Posted by kq76

I don't have the time right now, but if you don't figure this out soon I'll try to dig into it more because I'm a bit concerned it's not this at all and I'll encounter the same problem when I switch over to 27.

Did you happen to try removing the iBB column? I wonder if it's simply a problem with it and I lucked into the fix by not adding it to my file.

Removing the IBB column just causes autocalc to default to 1.0 IBB modifier. This value is better, but still not exactly what I'm looking for. For context, my 645 IBB has been causing the modifier to be in the .5-.7 range, which is far too low. I've tried adding in some "Other" PAs and rose the PA/AB ratio well above the typical 1.123, but I'm still seeing the same issue. My guess is this is a genuine bug with how the IBB column in era_stats.txt is implemented. The only thing I can find to raise the IBB modifier is entering an inflated IBB number (~2.2x what I actually want gets me close). If you wouldn't mind taking a look at some point to sanity check my results, that would be incredibly helpful! Also, if any devs are lurking, would you be able to help me tackle this issue and check how the IBB column is interpreted?

[Matt Arnold]

FYI, we more or less ignore the PA columns, I wouldn't worry about that. Especially around the edges, stuff like catcher's interference is a very small rounding error in the grand scheme.

Otherwise, some modifiers are easier to match to the totals, some are harder. IBB is a bit tougher than some, just because there's a number of cases where regardless of where you want your totals to be, you should be intentionally walking the player (eg if Judge is up with 2 outs, man on 3rd, the pitcher is on deck, in a close game). So a pure modifier has trouble sometimes around those cases.

IBB is also trickier, since we don't really have any specific era modifiers for them, but over the course of history, they have varied wildly. We're currently in one of the lowest points in history for them, on a per-AB basis. Although I do think it should still be mildly close - if your target was 645 IBB in 133k AB, how many did you actually get in your season when it was played out?

We can potentially do some tweaks to some of the logic and values, maybe help it along.

[Joelman111]

Originally Posted by Matt Arnold

FYI, we more or less ignore the PA columns, I wouldn't worry about that. Especially around the edges, stuff like catcher's interference is a very small rounding error in the grand scheme.

Otherwise, some modifiers are easier to match to the totals, some are harder. IBB is a bit tougher than some, just because there's a number of cases where regardless of where you want your totals to be, you should be intentionally walking the player (eg if Judge is up with 2 outs, man on 3rd, the pitcher is on deck, in a close game). So a pure modifier has trouble sometimes around those cases.

IBB is also trickier, since we don't really have any specific era modifiers for them, but over the course of history, they have varied wildly. We're currently in one of the lowest points in history for them, on a per-AB basis. Although I do think it should still be mildly close - if your target was 645 IBB in 133k AB, how many did you actually get in your season when it was played out?

We can potentially do some tweaks to some of the logic and values, maybe help it along.

Thank you so much for the response! I believe it is an issue with auto-calc just giving me the wrong multiplier. When I run auto-calc, it gives me a modifier in the range of ~.5-.7. This has led to IBBs in the range of 150-250, which is quite far from the desired total. I have found that a modifier of ~1.2 reliably gives me what I’m looking for, and the only way I’ve found that auto-calc will give me that modifier is if I boost my IBBs column way up (around 2.1x what I actually want, so 1355 will give me an actual total around 645.). This happens independent of the sim, so it appears the game engine does have enough control of IBBs to get the total where I want it to be (I.e. it’s not just a problem of game situations not aligning properly to warrant that many IBBs); it’s just a problem with auto-calc giving me the wrong multiplier. Let me know if I should provide league files to show an example of the problem!

Finally, I have one other quick question. It looks like the BB column in era_stats.txt is the total number of uBB + iBB. But when I’m loading in the era_stats.txt, the walk league total on the settings page shows only uBB (I.e. BB column - iBB). Does this mean when we control the BB league total on the league settings page, we are only messing with the uBB league total? When I check the league stats on the team batting statistics page, does the BB total shown include the IBB column or are those separated there as well? I find it a bit confusing to know when a BB total is referring to uBB vs. total walks (uBB+iBB).

[Joelman111]

Tested this in the newest public beta patch and this has been resolved! Really appreciate the quick turnaround time to fix this!