Avoid Ks and chance to advance runners

Timofmars · 02-26-2018, 06:26 AM

I have been in the process of making a detailed spreadsheet of sabermetric-type calculations for the game. I want to include a player's ability to avoid Ks in the calculations. I've played with the editor to find approximately how many strikeouts the avoid ks rating translates to, but I don't know how often balls in play that are outs actually causes runners to advance a base.

Does anyone have data on this, or even just guesses? I haven't watched a lot of baseball, so I don't have a good feel for the odds here.

My guess would be like maybe 20% of balls in play with a runner on 1st (where either the batter or the runner get out, i.e. no hit) will lead to the runner getting to 2nd.

I don't know the odds of a double play in the situation either. Anybody know?

I'd guess also that whenever a runner is not able to be forced out (runner on 2nd, 3rd, etc. with no runner on 1st) then those runner have a much higher chance of advancing them. Maybe 50%?

And of course, with 2 outs, none of this matters since getting the batter out ends the inning.

Drstrangelove · 02-26-2018, 02:38 PM

This information is at BBRef, although I don't doubt it could be at Fangraphs, Retrosheet, etc. or in digitized book form.
https://www.baseball-reference.com/

Look for league by year, e.g., AL 2012, and then within that page, drill down to SPLITS for batting or pitching stats. It goes back some number of decades, so you'll be able to find a lot of data. It varies in that the farther back you go, the less you get, but that makes sense.

NoOne · 02-26-2018, 04:03 PM

avoid k's will be easy.

if you don't have 100% accurate ratings, it will be less clear. stick to mlb players and hopefully a good scout and budget if you don't want to use 100% accuracy.

just about everyhting is a ~normal distribution, so we cna make some assumptions.

for example, ~1/2 scale should be ~baseline average SO per player. (normal dist, so mean=median or so close it won't matter)

the max scale players will be some specific SO less based on LTM and LT of the league for SO. 3 s.d. out for max? good guess. then the >max even further... but it will be tied to these metrics mentioned... and it wil be consistent and repeatable under the same settings/context.

this particular endeavor requires absolute ratings, not relative ratings based on nonsense for nonsense reasons clouding the picture... on purpose i might add!

having relative ratings for the ai's function would be a good reason for them... making ratings more difficult to read accurately and precisely from a human's standpoint is not a good thing. context would dictate absolute for this type of data mining.

advancing runners and such.. lots of info to google... scoring % per out and base the runner is on -- just read an article in last week about that. rl will be different from your league.. but it can show you how to evaluate it for your leageu and learn how it differs.

if terminolgy is a problem, look at some lists of abbreviations to help your google search for relevant data.

Timofmars · 02-26-2018, 09:49 PM

Quote:

Originally Posted by Drstrangelove

This information is at BBRef, although I don't doubt it could be at Fangraphs, Retrosheet, etc. or in digitized book form.
https://www.baseball-reference.com/

Look for league by year, e.g., AL 2012, and then within that page, drill down to SPLITS for batting or pitching stats. It goes back some number of decades, so you'll be able to find a lot of data. It varies in that the farther back you go, the less you get, but that makes sense.

I looked around at baseball reference. I found the GIDP data I was looking for. But I couldn't find anything about baserunners advancing on an out. I could just find how far players advanced on singles and doubles only. Maybe this is not a tracked statistic?

NoOne · 02-27-2018, 12:44 PM

you won't find somethign like that at ESPN etc.. well, not on the team's stats pages, that is.

maybe a blog or some math phd candidate's thesis you cna find through google etc. a more 'eccentric' person or website will have it.

if they don't explain methodology, don't put much weight behind their conclusions - even ignore it if it can't be verified elsewhere... if they do, critique it for possible flaws and why their conclusions 'might' be off, etc. or maybe just to know it's a ballpark concept, and not precise.

Timofmars · 02-28-2018, 07:01 AM

TLDR: I found some data and used it for calculations, and I found that every 32 Ks you avoid (but the batter still gets out) will result in 1 extra run gained. But the data is incomplete, and the game may not even reflect real life in this area.

I'd be happy with a ballpark figure, because right now I'm not giving my players any value for avoiding Ks, so I'm likely undervaluing them.

Using the GIDP info for one season, I found roughly that men are on (any base) 43.3% of the time, and double plays happen 4.5% of that time (of course, about 1/3 of the time there are 2 outs, making double plays impossible). And with men on first base (other bases may or may not be occupied) with less than 2 outs, GIPD happens almost 11% of the time.

So that gives me the negative outcomes of better Avoid Ks ability.

Actually, I just now found that before 1940, all times that a runner advanced when the batter hit into an out were counted at sacrifice hits, not just on bunts. And in some years, sacrifice flies were added in as well.

I took data for 1930, when SH included non-bunts and SF even if they only advanced a runner but did not score them. There were 2606
SH in 97262 plate appearances. So I took PA, subtracted hits, Ks, BB, HBP, which should give outs on balls-in-play. Then multiplied by 2/3 since that's the average time there should be less than 2 outs. Then multiplied by 0.433 to find times that men are on base (2012 data). That gives 16070 instances where runners are on base with less than 2 outs.

So divide SH by 16070 to show a 16.2% chance of advancing a runner on a ball-in-play that gets the batter out. If that data is still roughly reflective of baseball today, it probably overestimates the SH chance slightly since you're probably more likely to have runners on the closer you get to 3 outs. On the other hand, there can be more than 1 runner on base, so you can advance 2 or 3 runners in some of that 16.2% chance. So if I use average # of baserunners (0.622 is some rough data I have on that from 2017) and divide by the 0.433 of times runner are on, it should tell me that 1.436 runners are on base on average when there's at least 1 runner on. I assume the SH data only credits a maximum of 1 SH each PA, even if more than 1 runner advances. Sometimes only 1 of the baserunners will advance, but for simplicity, I'll assume they all advance. So 16.2% chance of getting a SH on a ball-in-play out, that moves an average of 1.436 runner gives 0.233 bases gained on balls-in-play with runners on where the batter doesn't get a hit.

Based on my data, each base gained (but not scoring) on the SH is about an extra 0.10 or 0.11 runs, compared to just a strikeout, while a runner scoring from 3rd is a 0.42 or 0.76 run gain (depending on whether you are left with 1 or 2 outs). Runner are on 3rd about 16% of the time that any runners are on, but that's where your biggest gains are from. If I multiply these values out by the expected chance of baserunners on each base, I get 0.404 runs gained per SH compared to a K. And multiply that by the chance of that ball in play being a SH instead of an out (16.2%) give you an average 0.0654 runs on each ball in play with less than 2 outs that is not a hit (nor a double play), compared to just a strikeout.

Doing a similar calculation for GIDP, there's a 0.139 chance of a GIDP on balls-in-play with runners on that are not hits. Compared to just a normal out, this takes away an additional out and a baserunner. Comparing this to a strikeout, assuming the baserunner was on 1st, taking him off the base results in a loss of 0.229 expected runs (if there was 0 outs before) or 0.111 expected runs (if there was 1 out before). The penalty from the additional out lowers the expected runs of other baserunners by about 0.2 each (though they may advance a base to offset penalty), and maybe like a 0.15 loss of expected runs on the batter that doesn't get to bat now (or slightly less, since he surely has at least 1 runner fewer to have a chance to hit in, and I already calculated the loss from that). Multiplying these numbers by the expected baserunners (just as I did with SH before), I get a 0.41 loss of expect runs per DP, but then assuming any runners on 2nd and 3rd advance at that time, it reduces the penalty of a DP down to 0.247. Multiply by the chance of a GIDP for a ball-in-play with runners on that are not hits, with less than 2 outs, and get an average of 0.034 runs lost. Subtract that from runs gained (due to SH), and you are left with 0.031 runs gained for hitting a ball in play for an out instead of striking out. This all works out to give you an extra run every 32 Ks you avoid.

If I didn't make any mistakes in the math, there's a few other issues. The GIDP data only includes standard double plays, so every DP gets the batter and 1st base runner out. So there should be more DPs that what is shown in the data, though they are much more rare I think. Also, I have no idea how accurately OOTP models these stats in particular. They may do a good job on making hits and HRs appear in realistic amounts, but who knows what they used to figure out how often baserunners advance on an out.

NoOne · 02-28-2018, 02:57 PM

there's more stongly correlated things to advancing a runner than just avoid k's

i'd look at big picture of all factors and apply the same type of research.

1 example:
babip -- % AND broken into fb/gb/ line drive ratios - drastically affects ability to move a runner up and drastically different from player to player... i can't think of one factor that would be more stongly correlated??? open ended not trying.

that throws a wrench in to the usefulness of the ~32avoid k per 1 additional run you calculated above. you'd have to know baselines and implications of being above or below the baselines for all factors for that player that is relative to advancing a runner in order to apply your 1run/32 avoid k's to any individual player or 2+ players for comparison.

if it's part of a bigger picture idea, that's great, nevermind. one pice of the puzzle mapped, on to the next!

one really important edit: If you plan to apply this or any similar research, it must be in a ~similar statistical environemtn -- not just the problem of converting odd sf/sh stats of the past.

e.g. data from 1940 doesn't really apply to 2018 in RL... it may on some facets.. it is still the same game essentially, but it'll be imprecise relative to better samples of data that can be used.

also, 10years isn't a very big sample accurately include ebb and flow of talent over time in its proper proportions. so, picking data from an era (even if longer than 10 years) is inherently off a bit no matter what.

also, ootp isn't real life, even though it models it closely... use ootp simulations to test your theories and it will apply to teh game of OotP more accurately and precisley... not only that you can control the statistical environment. you set boundaries and keep rules the same -- unlike RL. e.g. i use 100+ year sims to hammer out specific LTM. i allow player talent changes to cause change to league results, not a "button" or changin a value for a league total.

Timofmars · 02-28-2018, 11:32 PM

Yeah, I was just getting this info to use in the game, not to apply to real baseball. It appears the game doesn't have any kind of tendency differences for players to hit in the air or on the ground, other than the pitchers' %GB ratings, so I don't see a reason to figure out fb/gb/line drive ratios. And babip is already its own rating in the player editor, and it works with avoid Ks (and Power) to determine contact. So it seems the Avoid Ks rating is mostly useless to look at, since it is already reflected in contact.

I don't have a lot of confidence that the game has similar levels of runners advancing bases on an out as what I calculated, since it was pretty difficult to gather the data because it is not tracked specifically in baseball statistics, so I have doubts the developers did much more than making their own estimates that seemed realistic to them. Though double plays are likely to be more accurate since there is data on them available.

It would be nice to test in-game to see how often baserunners advance on outs, but I would think you'd need to set up some kind of search algorithm that looks through individual game logs to see where there were groundouts/flyouts/fielder's choice when runners were on base and see how often baserunners advanced in those instances. Or how often double plays happen too, since that is the negative side of avoiding Ks.

NoOne · 03-01-2018, 02:05 AM

they use real data... if somethign isn't "there" specifically it's incorporated into another overall % success/failure, or what have you.

another thread just showed that speed doesn't affect singles vs doubles, ceterus paribus. Which sounds bad onthe surface, but the over %s and ratios are and can be realistic -- from league-wide or individual viewpoint.

If it had everything mapped out and appropriately weighted etc etc.. it would be a mult-imillion dollar modelling program used by 30 MLB teams right now. oh, and don't forget about Vegas.

Alas, it's just a video game that is quite sophisticated from that perspective. So, don't get too down about it. Some things are inevitably glossed over, but i promise it's not whimsical in nature. maybe a bit simplified, but still tied to reality overall.

i complain about it on occasion.. had a nerd rage a month or two ago... i don't even recall what it was, now, but it was serious then, LoL. (not drawing a comparison to this thread onthe nerd rage vein... it's just understandable to lose proper perspective on it wen you pull the curtain back and not like what you see on a very particular factor of baseball)

Timofmars · 03-01-2018, 11:03 AM

Oh, I just now noticed now that there's a rating on the player profile page showing if they tend to hit flyballs, ground balls, line drives, or "normal". There doesn't seem to be a way to change it in the editor, and changing ratings didn't do anything either. There does seem to be a tendency for high gap ability players to be more likely to hit line-drives, and for high power players to be more fly ball hitters, and lower power to be groundballers.

How do people create their own new custom players to add to the game? Do they get to specify this hitter type rating then?

If we assume this rating doesn't change the number of hits, doubles, triples, or HRs, etc. of a player, then I think its main effects must be on what it means for the other runners. Or maybe also on the effectiveness of different defensive formations, such as infield shift having less effect on flyball batters. Maybe groundball hitters benefit more from hit and runs.

As far as the runners, what might the different effects be, if they exist in the game? Fly ball hitters might have fewer GIDPs and less chance of advancing a runner on an out, unless they are batting lefty with a good amount of pull to make sac flies to right field more likely. Line-drives sound good in theory, but if the result of the play for the batter is already determined by his ratings, not this line-drive tendency, then it would just change what kind of outs he gets, not make his hits any better. And line-drive outs seem just like fly outs, except a bit more dangerous for the risk of getting doubled up if the runner can't get back to the base in time.

It's interesting that there are 4 settings (ground, fly, line, normal). That implies that if there's are hidden values behind this rating, it's not on a single scale from ground to fly, since both normal and line are in there too.

I don't know what other value the hitter types (pull and fly/ground) could have to make it a useful piece of information, other than for the defensive and offensive strategies.

NoOne · 03-01-2018, 03:53 PM

i assume the same things, from your first paragraph.. hr hitters are flyball guys.. all about power and gap ratings. defintieyl is the trend.

high gap rating can actually reduce HR, btw. test with 2 players all the same excapt pow/gap... make one max and 1/2 and 1/2 and max and watch the difference. not saying it's better at an extreme, just using for a more profound effect to be seen.

create a player:

all sorts of ways... you can create fictional and then the game will offer options like quality - legend, all-star, average*[sic]. etc... then you can edit further...

or you can start from scratch and jsut type it in the boxes as you wish. (editor with comissioner mode on)... use an existing player or create one and edit.

if you know how the editor's predicted stats translates to your league, you can type in stats and click "create ratings from stats" button. (if oyu know hr are higher/lower in your league, incorperate that into the # you use for hr before clicking)

e.g. if you see less power in league, make sure to bump up homeruns a bit above what the editor would predict... 34= 40 even though it will result in ~34 in your league.

(the stats are not callibrated to your league in the editor's predicitons... i assume it's ~near default ootp values for modern day mlb - it has a note in the editor relating to this)

power definiteyl affects hits. it is part of how Contact is cacluated (babip, avoidk and power = Contact). gap is limited to how many doubles/3b they hit, as you saw above in your findings.

the line drive, flyball thing.. not sure how in depth that's applied. a gap skewed player may be more likekly to maintian a higher average given same Contact (and same 3 proportions of contact - babip/avoid k / power). that would tell you if it's simply feedback or has some causal nature to it.

i've had 60/100 gap and 80+/100 pow guys hit for just as high of averages in my experience, i wouldn't bet on it meaning much. at least not at the high end of players.

hitter type as far as pull, extreme pull most definitely has an effect. look at infor form real life and it will almost assuredly follow this... extreme pull hitter in correct park will hit a ton of HR (looking at you lefties / ny stadium - you can subtact ~10hr from there season average when they leave, a la the 2b in seattle - age too for his example).

a gap hitter is probabyl best utilized with a spray hitter attribute -- thats a guess, btw. i prefer hr hitters to be normal or spray. i prefer consitency over volatility in some cases, even if consistency proves a few less runs in 162g. it can result in more wins due to increased clumping with the alternative.

that's all a guess in last paragraph.

you're mxing the 2 -- flyball hitter, line drive hitter (is there a neutral? there is for pitchers) then 2nd category: pull has - extreme, pull, normal, spray.

flyball guys probably are better at hitting sac flies. i'd start seeing if you see a trend between them and similar traits.. one guy consitently hits higher average and maybe you are surprised his hr total is not meeting a typical expectation etc... you could set up experiemetns to flesh this all out too... don't flood leageue with too many test players or you taint the environement. but, you can use more than 1 test player per season to incerase rate of research.

i'd say whatever fundamental theories oyu can find through google will be very close to how ootp handles pull-attribute and hitter-types (fb vs ld). (i don't know the labels from game by rote, but remember there 2 different things there)

Drstrangelove · 03-03-2018, 11:19 PM

Quote:

Originally Posted by Timofmars

I looked around at baseball reference. I found the GIDP data I was looking for. But I couldn't find anything about baserunners advancing on an out. I could just find how far players advanced on singles and doubles only. Maybe this is not a tracked statistic?

It sounds like you are looking for what are called "productive outs."

Drstrangelove · 03-03-2018, 11:28 PM

Quote:

Originally Posted by Timofmars

[Also, I have no idea how accurately OOTP models these stats in particular. They may do a good job on making hits and HRs appear in realistic amounts, but who knows what they used to figure out how often baserunners advance on an out.

I have tested the OOTP18 scoring model using the 2002 Runs Created formula (see wiki https://en.wikipedia.org/wiki/Runs_created.) OOTP18 came within 1.3% of the 2002 RC formula, so I would infer that it is modeling the correct behavior. (Keep in mind there is a separate caculation done elsewhere in-game that shows the leagues actual versus historical calculations as well.) The RC model is itself simply a theoretical model of real baseball scoring using all the statistics that are available to statisticians.

A difference within 1.3% means the OOTP model is somehow taking into account the general effect of the hits, stolen bases, advances on outs, etc. in a manner that leads to a similar result that "real" baseball generates. This doesn't prove it's working as real baseball, but it's a hurdle it has to cross in order to make that claim. It was successful. Without actually seeing the code, I would not be surprised if the modeling engine is working realistically. Again, this isn't proof, but it's a strong indicator. Mishandling the effect of outs in run generation would be relatively obvious to pick up in a test like this since outs are the most numerous event in baseball.

NoOne · 03-03-2018, 11:46 PM

that's a league-wide thing... that won't tell you if a particular % is incorporated into some other portion of what OotP tracks/enforces however it should be stated.

it won't have "all" %'s properly hashed out for all the minutiae that baseball provides.

so, it's easy to make the league-wide stuff match up well. even if some stuff is a "conglemerate" of multiple forces, it will still add up when you look at league-wide stats.

the code could be much simpler or much more complex than it is now, and still be ~1.3% in your test above. "too simple" would be a problem eventually. probably more volatility or extremely static year to year. 1 of the extremes, likely.

to simplify - the difference between all players batting baseline and having differentiation.. the league-wid stats would come out in a fairly similar fashion regardless of players being different from each other (a point could be found where it works out this way, if not baseline). applies to various forces that are enforced through % success/fail too. 1 might include many and jsut applied as an average to to all contexts in same way... which isn't how it would work for many things outside of physics (gravity etc) or similar contexts. some aspects of baseball are modeled this way in ootp, or it'd be perfect. which we know isn't possible.

Timofmars · 03-04-2018, 10:19 AM

Quote:

Originally Posted by Drstrangelove

It sounds like you are looking for what are called "productive outs."

Hmm, well that could be useful, though it seems to be a bit odd. It only counts advancing a runner with 0 outs to begin the AB, or scoring a run with 1 out to begin, or a pitcher getting a Sac with 1 out to begin.

So advancing a runner who doesn't score, when beginning the AB with 1 out, doesn't count and actually is counted as a failure instead. So the fact they are negatively counting a portion of something that I'm trying to add up together as a whole really throws a wrench in the usefulness of that stat, especially since they provide no totals for how many times they do that.

Drstrangelove · 03-04-2018, 02:30 PM

I've followed saber metrics since long before it was called that. I only mention that because the questions you raise aren't really new, nor do people not understand them. People have for decades continuously refined models to reflect how baseball actually works. The reason the stats you are seeking seemingly don't exist in the format you want is likely because it's unavailable or not considered useful. (I can't tell you which.)

Even "productive outs", much like "clutch hitting" are not used in many models because they are considered reflections of random luck, not skill. People doing saber metrics want to extract skill not luck from the data. Yes, Joe Smith might have moved runners from 1st to second with 1 out more than everyone else in 1997, but what did he do in 1996, or 1998? Is it a skill or luck? Did his team do that too, or were others worse? And did those performances change the next year?

Were the outs "productive", was the hitting "clutch", because the SS played out of position, because the ball hit second base, because a gust of wind blew dust into the third baseman's eye? Who knows? If it's skill, shouldn't it be repeatable? If it's random, why track it? Yes, a person might score because the manager stupidly did back to back sacrifices with a runner on 1st and no out, and then the pitcher threw a wild pitch. That's not skill or knowledge or forethought on the part of a manager or player, but random, dumb luck. What everyone wants to know is whether or not that is a good strategy (i.e., increases scoring) in baseball (it doesn't), not whether it once happened and a team scored a run.

In one of Ted Williams' books (My Turn at Bat?) he related a story of how he was hitting against a lefty and hit a line drive deep to left field, near the line, where it was caught by Yogi Berra. Williams was simply flabbergasted. To paraphrase, "You can't be there. You can't play a lefty pull hitter (like him), to hit a liner against a left-handed pitcher, to go down the left field line. Can't be done." Berra was either brilliant or out of position. The reality is more likely he was out of position and got lucky.

The models people have built predict actual scoring within 1%, league after league, year after year, using the metrics I posted. They use large volumes of data, not anecdotal Yogi Berra -Ted Williams stories to model. People of course are welcome to debate whether Berra was in the right spot, much like whether Jeter was in the right spot for the errant relay throw, whether Mays got lucky in 1954, why Larsen threw a perfect game, or Vander Meer threw back to back no-hitters. Anecdotal or one-off plays are imo a poor source for scoring models.

The conclusion people draw from that is not that every perfect stat has been found, but that enough stats have been found to model actual scoring with the remainder tending to nullify itself out. One can of course argue it's not proven. Some of Einstein's equations are still being unproven... One was "more" proved recently because we now have the equipment to measure light and planetary motions more exactly.

If anyone can add insight into how baseball data can be made more accurate or useful, people should publish it. As for OOTP, it's a simulation of baseball, not baseball, so all one should ask, imo is that, to the extent it can, it's statistics model closely to what a real baseball league might look like within the software's constraints. I give credit that it does that although I can't see under the hood. If people are saying that it's statistics don't do that, then of course, that is another discussion.

Brad K · 03-04-2018, 04:41 PM

It looks to me like if a guy can hit home runs, tell him to not worry about strikeouts and try to hit more.

NoOne · 03-04-2018, 06:32 PM

someone like JD Martinez, sure... Someone like ozzie smith, not so much.

launch angle is great for high power guys, it seems. inevitable creates a hole in your zone and high SO-rate. So, as long as they have enough power to overcome the drawbacks that occur, it's a good idea.

i can tell you for sure in OotP a slugger that walks less is 'better'. they will put up more rbi+r if they walk less. ~100+BB and they get ~10 less hr, and at least 20 fewer RBI on a ~1000run team. the loss of rbi outpaces the loss of rs @4th in the lineup for sure.

i won't go so far to say zero eye is good, but ~70-80+ kinda sucks at a slot in order you expect many rbi from.

obviously talent trumps that theory easily.. e.g. a superstar, future-hof is a superstar and even with 50 extra walks he will be better than the next tier down in quality.

Timofmars · 03-04-2018, 10:39 PM

Quote:

Originally Posted by NoOne

i can tell you for sure in OotP a slugger that walks less is 'better'. they will put up more rbi+r if they walk less. ~100+BB and they get ~10 less hr, and at least 20 fewer RBI on a ~1000run team. the loss of rbi outpaces the loss of rs @4th in the lineup for sure.

i won't go so far to say zero eye is good, but ~70-80+ kinda sucks at a slot in order you expect many rbi from.

Yeah, that's why I adjust all averages (batting, HRs) to Plate appearances. The adjusted figures show all that power on a guy is used slightly less as walks increase. But I wouldn't say that lower eye is preferred on these talented sluggers, but rather that eye on them offers less return on them than on other players.

Brad K · 03-05-2018, 12:25 AM

Don't know how we got from strikeouts to walks. Perhaps because traditional statistics compare strike outs with walks I don't see the relevancy of the comparison.

Anyway we have a break even rate of one solo home run per 32 strikeouts. With an average runner load its about one home run per 50 strikeouts.

Real life, if a guy can hit HRs and it appears he could hit more, ask him to try. As Harry Walker did with Roberto Clemente prior to 1966. (Clemente's HRs went up by 19, strike outs up by 31, BA down by 012.)

Course, it didn't work with Charlie Finley and Manny Jiminez. <G>

In ootp, I doubt anyone will be faced with a choice between players where strike out rate will be significant a factor. Except for people designing their own players.

02-26-2018, 06:26 AM	#1
Timofmars Minors (Triple A) Join Date: Feb 2018 Posts: 251	Avoid Ks and chance to advance runners I have been in the process of making a detailed spreadsheet of sabermetric-type calculations for the game. I want to include a player's ability to avoid Ks in the calculations. I've played with the editor to find approximately how many strikeouts the avoid ks rating translates to, but I don't know how often balls in play that are outs actually causes runners to advance a base. Does anyone have data on this, or even just guesses? I haven't watched a lot of baseball, so I don't have a good feel for the odds here. My guess would be like maybe 20% of balls in play with a runner on 1st (where either the batter or the runner get out, i.e. no hit) will lead to the runner getting to 2nd. I don't know the odds of a double play in the situation either. Anybody know? I'd guess also that whenever a runner is not able to be forced out (runner on 2nd, 3rd, etc. with no runner on 1st) then those runner have a much higher chance of advancing them. Maybe 50%? And of course, with 2 outs, none of this matters since getting the batter out ends the inning. Last edited by Timofmars; 02-26-2018 at 06:27 AM.

02-26-2018, 02:38 PM	#2
Drstrangelove Major Leagues Join Date: Jun 2014 Posts: 346	This information is at BBRef, although I don't doubt it could be at Fangraphs, Retrosheet, etc. or in digitized book form. https://www.baseball-reference.com/ Look for league by year, e.g., AL 2012, and then within that page, drill down to SPLITS for batting or pitching stats. It goes back some number of decades, so you'll be able to find a lot of data. It varies in that the farther back you go, the less you get, but that makes sense. Last edited by Drstrangelove; 02-26-2018 at 02:41 PM.

02-26-2018, 04:03 PM	#3
NoOne Banned Join Date: Apr 2015 Posts: 7,273 Infractions: 0/1 (3)	avoid k's will be easy. if you don't have 100% accurate ratings, it will be less clear. stick to mlb players and hopefully a good scout and budget if you don't want to use 100% accuracy. just about everyhting is a ~normal distribution, so we cna make some assumptions. for example, ~1/2 scale should be ~baseline average SO per player. (normal dist, so mean=median or so close it won't matter) the max scale players will be some specific SO less based on LTM and LT of the league for SO. 3 s.d. out for max? good guess. then the >max even further... but it will be tied to these metrics mentioned... and it wil be consistent and repeatable under the same settings/context. this particular endeavor requires absolute ratings, not relative ratings based on nonsense for nonsense reasons clouding the picture... on purpose i might add! having relative ratings for the ai's function would be a good reason for them... making ratings more difficult to read accurately and precisely from a human's standpoint is not a good thing. context would dictate absolute for this type of data mining. advancing runners and such.. lots of info to google... scoring % per out and base the runner is on -- just read an article in last week about that. rl will be different from your league.. but it can show you how to evaluate it for your leageu and learn how it differs. if terminolgy is a problem, look at some lists of abbreviations to help your google search for relevant data. Last edited by NoOne; 02-26-2018 at 04:06 PM.

02-28-2018, 07:01 AM	#6
Timofmars Minors (Triple A) Join Date: Feb 2018 Posts: 251	TLDR: I found some data and used it for calculations, and I found that every 32 Ks you avoid (but the batter still gets out) will result in 1 extra run gained. But the data is incomplete, and the game may not even reflect real life in this area. I'd be happy with a ballpark figure, because right now I'm not giving my players any value for avoiding Ks, so I'm likely undervaluing them. Using the GIDP info for one season, I found roughly that men are on (any base) 43.3% of the time, and double plays happen 4.5% of that time (of course, about 1/3 of the time there are 2 outs, making double plays impossible). And with men on first base (other bases may or may not be occupied) with less than 2 outs, GIPD happens almost 11% of the time. So that gives me the negative outcomes of better Avoid Ks ability. Actually, I just now found that before 1940, all times that a runner advanced when the batter hit into an out were counted at sacrifice hits, not just on bunts. And in some years, sacrifice flies were added in as well. I took data for 1930, when SH included non-bunts and SF even if they only advanced a runner but did not score them. There were 2606 SH in 97262 plate appearances. So I took PA, subtracted hits, Ks, BB, HBP, which should give outs on balls-in-play. Then multiplied by 2/3 since that's the average time there should be less than 2 outs. Then multiplied by 0.433 to find times that men are on base (2012 data). That gives 16070 instances where runners are on base with less than 2 outs. So divide SH by 16070 to show a 16.2% chance of advancing a runner on a ball-in-play that gets the batter out. If that data is still roughly reflective of baseball today, it probably overestimates the SH chance slightly since you're probably more likely to have runners on the closer you get to 3 outs. On the other hand, there can be more than 1 runner on base, so you can advance 2 or 3 runners in some of that 16.2% chance. So if I use average # of baserunners (0.622 is some rough data I have on that from 2017) and divide by the 0.433 of times runner are on, it should tell me that 1.436 runners are on base on average when there's at least 1 runner on. I assume the SH data only credits a maximum of 1 SH each PA, even if more than 1 runner advances. Sometimes only 1 of the baserunners will advance, but for simplicity, I'll assume they all advance. So 16.2% chance of getting a SH on a ball-in-play out, that moves an average of 1.436 runner gives 0.233 bases gained on balls-in-play with runners on where the batter doesn't get a hit. Based on my data, each base gained (but not scoring) on the SH is about an extra 0.10 or 0.11 runs, compared to just a strikeout, while a runner scoring from 3rd is a 0.42 or 0.76 run gain (depending on whether you are left with 1 or 2 outs). Runner are on 3rd about 16% of the time that any runners are on, but that's where your biggest gains are from. If I multiply these values out by the expected chance of baserunners on each base, I get 0.404 runs gained per SH compared to a K. And multiply that by the chance of that ball in play being a SH instead of an out (16.2%) give you an average 0.0654 runs on each ball in play with less than 2 outs that is not a hit (nor a double play), compared to just a strikeout. Doing a similar calculation for GIDP, there's a 0.139 chance of a GIDP on balls-in-play with runners on that are not hits. Compared to just a normal out, this takes away an additional out and a baserunner. Comparing this to a strikeout, assuming the baserunner was on 1st, taking him off the base results in a loss of 0.229 expected runs (if there was 0 outs before) or 0.111 expected runs (if there was 1 out before). The penalty from the additional out lowers the expected runs of other baserunners by about 0.2 each (though they may advance a base to offset penalty), and maybe like a 0.15 loss of expected runs on the batter that doesn't get to bat now (or slightly less, since he surely has at least 1 runner fewer to have a chance to hit in, and I already calculated the loss from that). Multiplying these numbers by the expected baserunners (just as I did with SH before), I get a 0.41 loss of expect runs per DP, but then assuming any runners on 2nd and 3rd advance at that time, it reduces the penalty of a DP down to 0.247. Multiply by the chance of a GIDP for a ball-in-play with runners on that are not hits, with less than 2 outs, and get an average of 0.034 runs lost. Subtract that from runs gained (due to SH), and you are left with 0.031 runs gained for hitting a ball in play for an out instead of striking out. This all works out to give you an extra run every 32 Ks you avoid. If I didn't make any mistakes in the math, there's a few other issues. The GIDP data only includes standard double plays, so every DP gets the batter and 1st base runner out. So there should be more DPs that what is shown in the data, though they are much more rare I think. Also, I have no idea how accurately OOTP models these stats in particular. They may do a good job on making hits and HRs appear in realistic amounts, but who knows what they used to figure out how often baserunners advance on an out. Last edited by Timofmars; 02-28-2018 at 09:20 AM. Reason: Add conclusion in TLDR

02-28-2018, 02:57 PM	#7
NoOne Banned Join Date: Apr 2015 Posts: 7,273 Infractions: 0/1 (3)	there's more stongly correlated things to advancing a runner than just avoid k's i'd look at big picture of all factors and apply the same type of research. 1 example: babip -- % AND broken into fb/gb/ line drive ratios - drastically affects ability to move a runner up and drastically different from player to player... i can't think of one factor that would be more stongly correlated??? open ended not trying. that throws a wrench in to the usefulness of the ~32avoid k per 1 additional run you calculated above. you'd have to know baselines and implications of being above or below the baselines for all factors for that player that is relative to advancing a runner in order to apply your 1run/32 avoid k's to any individual player or 2+ players for comparison. if it's part of a bigger picture idea, that's great, nevermind. one pice of the puzzle mapped, on to the next! one really important edit: If you plan to apply this or any similar research, it must be in a ~similar statistical environemtn -- not just the problem of converting odd sf/sh stats of the past. e.g. data from 1940 doesn't really apply to 2018 in RL... it may on some facets.. it is still the same game essentially, but it'll be imprecise relative to better samples of data that can be used. also, 10years isn't a very big sample accurately include ebb and flow of talent over time in its proper proportions. so, picking data from an era (even if longer than 10 years) is inherently off a bit no matter what. also, ootp isn't real life, even though it models it closely... use ootp simulations to test your theories and it will apply to teh game of OotP more accurately and precisley... not only that you can control the statistical environment. you set boundaries and keep rules the same -- unlike RL. e.g. i use 100+ year sims to hammer out specific LTM. i allow player talent changes to cause change to league results, not a "button" or changin a value for a league total. Last edited by NoOne; 02-28-2018 at 03:03 PM.

02-27-2018, 12:44 PM	#5
NoOne Banned Join Date: Apr 2015 Posts: 7,273 Infractions: 0/1 (3)	you won't find somethign like that at ESPN etc.. well, not on the team's stats pages, that is. maybe a blog or some math phd candidate's thesis you cna find through google etc. a more 'eccentric' person or website will have it. if they don't explain methodology, don't put much weight behind their conclusions - even ignore it if it can't be verified elsewhere... if they do, critique it for possible flaws and why their conclusions 'might' be off, etc. or maybe just to know it's a ballpark concept, and not precise.

02-28-2018, 11:32 PM	#8
Timofmars Minors (Triple A) Join Date: Feb 2018 Posts: 251	Yeah, I was just getting this info to use in the game, not to apply to real baseball. It appears the game doesn't have any kind of tendency differences for players to hit in the air or on the ground, other than the pitchers' %GB ratings, so I don't see a reason to figure out fb/gb/line drive ratios. And babip is already its own rating in the player editor, and it works with avoid Ks (and Power) to determine contact. So it seems the Avoid Ks rating is mostly useless to look at, since it is already reflected in contact. I don't have a lot of confidence that the game has similar levels of runners advancing bases on an out as what I calculated, since it was pretty difficult to gather the data because it is not tracked specifically in baseball statistics, so I have doubts the developers did much more than making their own estimates that seemed realistic to them. Though double plays are likely to be more accurate since there is data on them available. It would be nice to test in-game to see how often baserunners advance on outs, but I would think you'd need to set up some kind of search algorithm that looks through individual game logs to see where there were groundouts/flyouts/fielder's choice when runners were on base and see how often baserunners advanced in those instances. Or how often double plays happen too, since that is the negative side of avoiding Ks.

03-01-2018, 02:05 AM	#9
NoOne Banned Join Date: Apr 2015 Posts: 7,273 Infractions: 0/1 (3)	they use real data... if somethign isn't "there" specifically it's incorporated into another overall % success/failure, or what have you. another thread just showed that speed doesn't affect singles vs doubles, ceterus paribus. Which sounds bad onthe surface, but the over %s and ratios are and can be realistic -- from league-wide or individual viewpoint. If it had everything mapped out and appropriately weighted etc etc.. it would be a mult-imillion dollar modelling program used by 30 MLB teams right now. oh, and don't forget about Vegas. Alas, it's just a video game that is quite sophisticated from that perspective. So, don't get too down about it. Some things are inevitably glossed over, but i promise it's not whimsical in nature. maybe a bit simplified, but still tied to reality overall. i complain about it on occasion.. had a nerd rage a month or two ago... i don't even recall what it was, now, but it was serious then, LoL. (not drawing a comparison to this thread onthe nerd rage vein... it's just understandable to lose proper perspective on it wen you pull the curtain back and not like what you see on a very particular factor of baseball) Last edited by NoOne; 03-01-2018 at 02:10 AM.

03-01-2018, 11:03 AM	#10
Timofmars Minors (Triple A) Join Date: Feb 2018 Posts: 251	Oh, I just now noticed now that there's a rating on the player profile page showing if they tend to hit flyballs, ground balls, line drives, or "normal". There doesn't seem to be a way to change it in the editor, and changing ratings didn't do anything either. There does seem to be a tendency for high gap ability players to be more likely to hit line-drives, and for high power players to be more fly ball hitters, and lower power to be groundballers. How do people create their own new custom players to add to the game? Do they get to specify this hitter type rating then? If we assume this rating doesn't change the number of hits, doubles, triples, or HRs, etc. of a player, then I think its main effects must be on what it means for the other runners. Or maybe also on the effectiveness of different defensive formations, such as infield shift having less effect on flyball batters. Maybe groundball hitters benefit more from hit and runs. As far as the runners, what might the different effects be, if they exist in the game? Fly ball hitters might have fewer GIDPs and less chance of advancing a runner on an out, unless they are batting lefty with a good amount of pull to make sac flies to right field more likely. Line-drives sound good in theory, but if the result of the play for the batter is already determined by his ratings, not this line-drive tendency, then it would just change what kind of outs he gets, not make his hits any better. And line-drive outs seem just like fly outs, except a bit more dangerous for the risk of getting doubled up if the runner can't get back to the base in time. It's interesting that there are 4 settings (ground, fly, line, normal). That implies that if there's are hidden values behind this rating, it's not on a single scale from ground to fly, since both normal and line are in there too. I don't know what other value the hitter types (pull and fly/ground) could have to make it a useful piece of information, other than for the defensive and offensive strategies.

03-01-2018, 03:53 PM	#11
NoOne Banned Join Date: Apr 2015 Posts: 7,273 Infractions: 0/1 (3)	i assume the same things, from your first paragraph.. hr hitters are flyball guys.. all about power and gap ratings. defintieyl is the trend. high gap rating can actually reduce HR, btw. test with 2 players all the same excapt pow/gap... make one max and 1/2 and 1/2 and max and watch the difference. not saying it's better at an extreme, just using for a more profound effect to be seen. create a player: all sorts of ways... you can create fictional and then the game will offer options like quality - legend, all-star, average*[sic]. etc... then you can edit further... or you can start from scratch and jsut type it in the boxes as you wish. (editor with comissioner mode on)... use an existing player or create one and edit. if you know how the editor's predicted stats translates to your league, you can type in stats and click "create ratings from stats" button. (if oyu know hr are higher/lower in your league, incorperate that into the # you use for hr before clicking) e.g. if you see less power in league, make sure to bump up homeruns a bit above what the editor would predict... 34= 40 even though it will result in ~34 in your league. (the stats are not callibrated to your league in the editor's predicitons... i assume it's ~near default ootp values for modern day mlb - it has a note in the editor relating to this) power definiteyl affects hits. it is part of how Contact is cacluated (babip, avoidk and power = Contact). gap is limited to how many doubles/3b they hit, as you saw above in your findings. the line drive, flyball thing.. not sure how in depth that's applied. a gap skewed player may be more likekly to maintian a higher average given same Contact (and same 3 proportions of contact - babip/avoid k / power). that would tell you if it's simply feedback or has some causal nature to it. i've had 60/100 gap and 80+/100 pow guys hit for just as high of averages in my experience, i wouldn't bet on it meaning much. at least not at the high end of players. hitter type as far as pull, extreme pull most definitely has an effect. look at infor form real life and it will almost assuredly follow this... extreme pull hitter in correct park will hit a ton of HR (looking at you lefties / ny stadium - you can subtact ~10hr from there season average when they leave, a la the 2b in seattle - age too for his example). a gap hitter is probabyl best utilized with a spray hitter attribute -- thats a guess, btw. i prefer hr hitters to be normal or spray. i prefer consitency over volatility in some cases, even if consistency proves a few less runs in 162g. it can result in more wins due to increased clumping with the alternative. that's all a guess in last paragraph. you're mxing the 2 -- flyball hitter, line drive hitter (is there a neutral? there is for pitchers) then 2nd category: pull has - extreme, pull, normal, spray. flyball guys probably are better at hitting sac flies. i'd start seeing if you see a trend between them and similar traits.. one guy consitently hits higher average and maybe you are surprised his hr total is not meeting a typical expectation etc... you could set up experiemetns to flesh this all out too... don't flood leageue with too many test players or you taint the environement. but, you can use more than 1 test player per season to incerase rate of research. i'd say whatever fundamental theories oyu can find through google will be very close to how ootp handles pull-attribute and hitter-types (fb vs ld). (i don't know the labels from game by rote, but remember there 2 different things there)

03-03-2018, 11:46 PM	#14
NoOne Banned Join Date: Apr 2015 Posts: 7,273 Infractions: 0/1 (3)	that's a league-wide thing... that won't tell you if a particular % is incorporated into some other portion of what OotP tracks/enforces however it should be stated. it won't have "all" %'s properly hashed out for all the minutiae that baseball provides. so, it's easy to make the league-wide stuff match up well. even if some stuff is a "conglemerate" of multiple forces, it will still add up when you look at league-wide stats. the code could be much simpler or much more complex than it is now, and still be ~1.3% in your test above. "too simple" would be a problem eventually. probably more volatility or extremely static year to year. 1 of the extremes, likely. to simplify - the difference between all players batting baseline and having differentiation.. the league-wid stats would come out in a fairly similar fashion regardless of players being different from each other (a point could be found where it works out this way, if not baseline). applies to various forces that are enforced through % success/fail too. 1 might include many and jsut applied as an average to to all contexts in same way... which isn't how it would work for many things outside of physics (gravity etc) or similar contexts. some aspects of baseball are modeled this way in ootp, or it'd be perfect. which we know isn't possible. Last edited by NoOne; 03-03-2018 at 11:52 PM.

03-04-2018, 02:30 PM	#16
Drstrangelove Major Leagues Join Date: Jun 2014 Posts: 346	I've followed saber metrics since long before it was called that. I only mention that because the questions you raise aren't really new, nor do people not understand them. People have for decades continuously refined models to reflect how baseball actually works. The reason the stats you are seeking seemingly don't exist in the format you want is likely because it's unavailable or not considered useful. (I can't tell you which.) Even "productive outs", much like "clutch hitting" are not used in many models because they are considered reflections of random luck, not skill. People doing saber metrics want to extract skill not luck from the data. Yes, Joe Smith might have moved runners from 1st to second with 1 out more than everyone else in 1997, but what did he do in 1996, or 1998? Is it a skill or luck? Did his team do that too, or were others worse? And did those performances change the next year? Were the outs "productive", was the hitting "clutch", because the SS played out of position, because the ball hit second base, because a gust of wind blew dust into the third baseman's eye? Who knows? If it's skill, shouldn't it be repeatable? If it's random, why track it? Yes, a person might score because the manager stupidly did back to back sacrifices with a runner on 1st and no out, and then the pitcher threw a wild pitch. That's not skill or knowledge or forethought on the part of a manager or player, but random, dumb luck. What everyone wants to know is whether or not that is a good strategy (i.e., increases scoring) in baseball (it doesn't), not whether it once happened and a team scored a run. In one of Ted Williams' books (My Turn at Bat?) he related a story of how he was hitting against a lefty and hit a line drive deep to left field, near the line, where it was caught by Yogi Berra. Williams was simply flabbergasted. To paraphrase, "You can't be there. You can't play a lefty pull hitter (like him), to hit a liner against a left-handed pitcher, to go down the left field line. Can't be done." Berra was either brilliant or out of position. The reality is more likely he was out of position and got lucky. The models people have built predict actual scoring within 1%, league after league, year after year, using the metrics I posted. They use large volumes of data, not anecdotal Yogi Berra -Ted Williams stories to model. People of course are welcome to debate whether Berra was in the right spot, much like whether Jeter was in the right spot for the errant relay throw, whether Mays got lucky in 1954, why Larsen threw a perfect game, or Vander Meer threw back to back no-hitters. Anecdotal or one-off plays are imo a poor source for scoring models. The conclusion people draw from that is not that every perfect stat has been found, but that enough stats have been found to model actual scoring with the remainder tending to nullify itself out. One can of course argue it's not proven. Some of Einstein's equations are still being unproven... One was "more" proved recently because we now have the equipment to measure light and planetary motions more exactly. If anyone can add insight into how baseball data can be made more accurate or useful, people should publish it. As for OOTP, it's a simulation of baseball, not baseball, so all one should ask, imo is that, to the extent it can, it's statistics model closely to what a real baseball league might look like within the software's constraints. I give credit that it does that although I can't see under the hood. If people are saying that it's statistics don't do that, then of course, that is another discussion. Last edited by Drstrangelove; 03-04-2018 at 08:25 PM.

03-04-2018, 04:41 PM	#17
Brad K Banned Join Date: May 2016 Location: St Petersburg Florida USA Posts: 6,693 Infractions: 0/2 (4)	It looks to me like if a guy can hit home runs, tell him to not worry about strikeouts and try to hit more.

03-04-2018, 06:32 PM	#18
NoOne Banned Join Date: Apr 2015 Posts: 7,273 Infractions: 0/1 (3)	someone like JD Martinez, sure... Someone like ozzie smith, not so much. launch angle is great for high power guys, it seems. inevitable creates a hole in your zone and high SO-rate. So, as long as they have enough power to overcome the drawbacks that occur, it's a good idea. i can tell you for sure in OotP a slugger that walks less is 'better'. they will put up more rbi+r if they walk less. ~100+BB and they get ~10 less hr, and at least 20 fewer RBI on a ~1000run team. the loss of rbi outpaces the loss of rs @4th in the lineup for sure. i won't go so far to say zero eye is good, but ~70-80+ kinda sucks at a slot in order you expect many rbi from. obviously talent trumps that theory easily.. e.g. a superstar, future-hof is a superstar and even with 50 extra walks he will be better than the next tier down in quality.

03-05-2018, 12:25 AM	#20
Brad K Banned Join Date: May 2016 Location: St Petersburg Florida USA Posts: 6,693 Infractions: 0/2 (4)	Don't know how we got from strikeouts to walks. Perhaps because traditional statistics compare strike outs with walks I don't see the relevancy of the comparison. Anyway we have a break even rate of one solo home run per 32 strikeouts. With an average runner load its about one home run per 50 strikeouts. Real life, if a guy can hit HRs and it appears he could hit more, ask him to try. As Harry Walker did with Roberto Clemente prior to 1966. (Clemente's HRs went up by 19, strike outs up by 31, BA down by 012.) Course, it didn't work with Charlie Finley and Manny Jiminez. <G> In ootp, I doubt anyone will be faced with a choice between players where strike out rate will be significant a factor. Except for people designing their own players.