Posted by pjfoster13 on 6/4/2017 2:33:00 PM (view original):

Posted by MikeT23 on 6/2/2017 3:55:00 PM (view original):
This is a plus/minus game. Are you implying that a failed defensive C in RF with an OPS of .950 is worse than a "normal" RF with a .710 OPS? Because, if you are, you're wrong.

Yes, I am implying that. WAR formula takes into account bat, defense, and baserunning. In your example, the .950 ops guy is probably +5 or +6 at the plate, but he's -4 in the field and probably also bad at baserunning too, so he ends up being a break-even or slightly above average guy. The .710 is +0 or -1 at the plate but +2 on defense and +2 baserunning. I dunno man I just think you're overestimating the bat and underestimating the overall negative influence. It's fine to say that it's "not so bad", because that's true. But to say that it's "best", that is definitely false. DHs play DH

I suspect you missed a couple of important points that are buried in the "C in RF" thread.

The whole exploit is based on the fact that the HBD game simply gives RFs the very least fielding chances of any position. The fact that the player just doesn't have to touch the ball that often makes the risk worth the reward. And further observation indicated that Arm Strength/Accuracy, not Glove or Range, seem to contribute most to successful RF play.

As my anecdotal observations indicate, playing a 1B in RF is almost as bad as playing a C in RF, and a 1B is supposed to actually have Range and Glove.

Just an aside, peripherally related to this, think about prospects you see when the system first generates them in the amateur draft (or as IFAs.)

RFs are generally players with weak gloves and big arms. 3Bs are generally players who do a little of everything but nothing well. Matter of fact, I can say that one thing I've never seen in an amateur draft is a truly great natural 3B. And in most of my drafts, unless there are elite hitters to be had at those positions, I adjust my settings to draft zero 3B, and half the time, zero RF.

The 40/48/44/43 guy is worse than 7/24/68/68.

And the 2nd guy is absolutely brutal at 1B.

Posted by MikeT23 on 6/4/2017 3:11:00 PM (view original):
Well, in HBD, we have all the numbers and we can crunch them. Ryu in RF was a plus.

You're still only using one specific metric which is +/-, and not whether there are additional bad ays which are registering as hits but not - plays. Sorta like how infield singles are neither hits nor minus plays, but are a function of sub-plus defense. I'm not the only person who posits this theory, and you've never done deeper analysis on it. It contradicts your worldview so you dismiss it instead of considering that you could be wrong.

I estimate that C or DH in RF would only have significant value (WAR) if the ops was closer to 1.100 rather than .950, because the bat has to go wayyy above and beyond its own intrinsic value to counterbalance the negatives. .950 is break-even. And if the 1.100 C/DH was in a secondary position you would also have to have a second C/DH that was already playing DH, because why would you play .950 with negative at DH and 1.100 with negative defense in the field, that's backwards

I guess it just depends whether you're trying to win every game 7-6 instead of 4-3?

Posted by pjfoster13 on 6/4/2017 2:27:00 PM (view original):

Posted by hockey1984 on 6/2/2017 11:17:00 PM (view original):

Posted by damag on 6/2/2017 4:28:00 PM (view original):
My C in RF in Riley World has 12 minus plays and 2 errors in 90 games. All the other RFs with minus plays over 5 are "natural" 1B. My guy's OPS is .929.

FWIW this exists in real life, one name for it is Schwarber.

As for a backup IF at C, try it sometime. The game will let you get away with it in the very short term, just like if your backup C has to play SS. Your pitchers won't immediately, like, give up ten HRs.

The Jays also put their C at third base a few days ago. No way I am doing that.

Pablo Sandoval did the C to 3b thing, a few guys have. The stupid thing about HBD is that range is absolute. The game considers a 10 range at C as also a 10 range in RF, as if the guy is still wearing all of his catcher gear. Very illogical, but from a coding perspective I wouldn't be interested in putting in the effort to fix that

Shut up. My catchers play every position in their gear. And come up to bat with it on. Hence why none of them steal bases or have good base running IQ.

Posted by pjfoster13 on 6/4/2017 5:11:00 PM (view original):

Posted by MikeT23 on 6/4/2017 3:11:00 PM (view original):
Well, in HBD, we have all the numbers and we can crunch them. Ryu in RF was a plus.

You're still only using one specific metric which is +/-, and not whether there are additional bad ays which are registering as hits but not - plays. Sorta like how infield singles are neither hits nor minus plays, but are a function of sub-plus defense. I'm not the only person who posits this theory, and you've never done deeper analysis on it. It contradicts your worldview so you dismiss it instead of considering that you could be wrong.

I estimate that C or DH in RF would only have significant value (WAR) if the ops was closer to 1.100 rather than .950, because the bat has to go wayyy above and beyond its own intrinsic value to counterbalance the negatives. .950 is break-even. And if the 1.100 C/DH was in a secondary position you would also have to have a second C/DH that was already playing DH, because why would you play .950 with negative at DH and 1.100 with negative defense in the field, that's backwards

I guess it just depends whether you're trying to win every game 7-6 instead of 4-3?

We have access to a little thing called range factor. We can easily figure out how many outs an average RF records. So, really, there is no guess work. You aren't quite sure how much analysis I've done, are you?

Posted by pjfoster13 on 6/4/2017 5:11:00 PM (view original):

Posted by MikeT23 on 6/4/2017 3:11:00 PM (view original):
Well, in HBD, we have all the numbers and we can crunch them. Ryu in RF was a plus.

You're still only using one specific metric which is +/-, and not whether there are additional bad ays which are registering as hits but not - plays. Sorta like how infield singles are neither hits nor minus plays, but are a function of sub-plus defense. I'm not the only person who posits this theory, and you've never done deeper analysis on it. It contradicts your worldview so you dismiss it instead of considering that you could be wrong.

I estimate that C or DH in RF would only have significant value (WAR) if the ops was closer to 1.100 rather than .950, because the bat has to go wayyy above and beyond its own intrinsic value to counterbalance the negatives. .950 is break-even. And if the 1.100 C/DH was in a secondary position you would also have to have a second C/DH that was already playing DH, because why would you play .950 with negative at DH and 1.100 with negative defense in the field, that's backwards

I guess it just depends whether you're trying to win every game 7-6 instead of 4-3?

I hate to pick a small issue out of a larger comment that I think may be partially valid (the part about additional bad plays), but you're wrong specifically about infield hits-- this is something I've actually studied.

I am quite certain, after looking at dozens of players, that HBD determines an "infield hit" after it determines "hit" vs. "out." Here's the basics:
1) Infield hits are very tightly correlated with speed.
2) I have a formula that predicts batting average from "blue" ratings decently, especially if you take care to note the park adjustment and use career stats to get a reasonable sample size. This formula, which does not use speed, predicts batting average equally well in players with very high speed vs very low speed, and equally well in players with an infield hit every 20 AB and those with no career infield hits.
3) Therefore, infield hits are not earned in place of outs, but in place of other hits.

I know this is ridiculous from a real baseball perspective. But there's a ton of other evidence that the sim determines very early in each plate appearance whether there is a hit or an out, and that several other decisions mostly add flavor. This is why, when power was de-emphasized in the engine about 30 seasons ago, batting averages stayed stable when homers dropped; homers are replacing singles in hitters with high power. This is why both my regression results and almost everybody's anecdotal observation suggest that velocity is essentially irrelevant to ERA, because Ks are replacing other outs, not balls in play. And it's why infield hits appear, as best as I can tell, to be unrelated to anything except player speed.

My RF is currently fifth in the league in range factor among qualified RFs, despite 31 minus plays. I'm not convinced range factor measures anything important.

You could be a flyball team. In part because 31 negative plays is HUGE and your RF is still #5.

lot to unpack here...

@dedelman- very good observations, speaks to HBD being fundamentally flawed in the sense that the decision tree for the sim is written completely out of order. "Hit" or "no hit" is a BABIP thing that needs to be decided AFTER "walk" vs "no walk" and "strike out" vs "no strikeout", but the decision tree places Pitching Split vs Hitting Split ("hit" vs "no hit") at the beginning of the decision tree and that's just completely not how the pitcher-batter interaction actually works in real baseball.

You sound very confident about the infield hit analysis, so I suppose we'll go with that. My problem is that there's incomplete data on the Team Stats page. I can go back and research how many infield hits each OFFENSE records but I cannot retrieve how many infield hits each team's DEFENSE concedes, so for me that's a problem. I'd love to be able to quantify whether the infield combo of Player Profile: Nick Holmes at 3B Player Profile: Lawrence Montgomery at 2B and Player Profile: Pat Suzuki at SS is denying infield singles without impacting +/- ratio. But we can't, so for me it's still sorta inconclusive, because we can't prove your hitters data is not heavily-skewed against teams with bad fielders as opposed to evenly-distributed against good-fielders and bad-fielders. Do you know what I mean? I can't reconcile your hitter data with corresponding fielding data. On a related note, Pitcher Stats includes Hits Allowed and HR Allowed but does not include 2B allowed and 3B allowed, so we can't derive the exact negative impact of Bad Outfield Range as it pertains to Total Bases.

I also disagree that K-rate has nothing to do with ERA, that is false. Baserunners don't advance on strikeouts (!!!) but they do on flyballs and they can on certain groundouts, so strikeouts have a different value of Total Bases (fewer) than other outs. Strikeouts have value, especially in hitters parks where Hit rate is higher... the more total bases you deny positively affects the number of runs you allow over large sample sizes. The problem with K-rate on HBD is that it's capped at approx 9.5 per 9, which is mis-calibrated compared to real life MLB http://www.rotowire.com/baseball/player_ex_stats.htm?pos=P HBD is hilariously low, considering Craig Kimbrel's K/9 is currently 17.05 lol and Dellin Betances 16.20 and Kenley Jansen 15.65... it's not just a few guys it's everybody, bullpen guys are all higher than starters but most of the starters consistently hover at 10-11 K/9. That's why real-life ERAs are at generational lows. Baserunners.

@saintonian- we recently talked about Range Factor in a separate post, and yes you are on the right track that range factor is not necessarily a measure of defensive ability. It is a volume statistic that measures total outs as opposed to fielding efficiency. As @Mike is alluding, your right fielder with horrendous efficiency is still getting huge out-volume. Your pitchers are conceding a ton of fielding CHANCES towards that region of the field. The lack of efficiency is a huge problem towards your expected-wins. Your 3 teams are all winning teams right now so that's cool I guess, but you'd be doing even better... ?

@pjfoster-- Only one quibble with your commentary, and that is that Bill James and others have shown that strikeouts, compared to other outs, have essentially no impact on run scoring. The reason is presumed to be simple-- when you strike out, you don't ground into a double play. Basically, DPs balance all the productive outs.

Now, in real baseball, that doesn't mean that striking out is OK. But in HBD, hit vs. out seems to be determined very early in the sim. Because of this, the comparison of Ks to other outs, not entirely relevant in MLB, is relevant in HBD. And strikeouts, compared to other outs, have essentially no impact on run scoring. Which is why velocity appears to be an irrelevant rating.

Except, of course, that while typing this I found a flaw in my analysis-- my regression has ERA as the outcome. Which means that unearned runs aren't relevant in my analysis. So I stand by my assertion that velocity has no relationship to ERA in HBD, but I'm now less certain that it has no relationship to offense.

The RF in question hit 354/464/653 this season, and I have a DH that's a perennial MVP candidate so this the perfect setup for this kind of experiment. My other teams have more conventional rightfielders. I know it's still a positive to play him there, I just haven't stumbled on a way to quantify the defensive impact.

this has turned into a very interesting thread. there is one assumption that needs to be changed through:
@dedelman- very good observations, speaks to HBD being fundamentally flawed in the sense that the decision tree for the sim is written completely out of order. "Hit" or "no hit" is a BABIP thing that needs to be decided AFTER "walk" vs "no walk" and "strike out" vs "no strikeout", but the decision tree places Pitching Split vs Hitting Split ("hit" vs "no hit") at the beginning of the decision tree and that's just completely not how the pitcher-batter interaction actually works in real baseball.

The sim does not place that decision at the beginning of the tree. The following is from SLB, but I would bet pretty heavily that they didn't completely re-imagine the event tree when they wrote HBD:
The tree starts with HBP vs no HBP. Then is it BB or AB. Then it is Out or Hit. If it is out, then it is K vs batted ball. Then normal out or error or (-) play. If it is a hit, it is HR or not, then hit vs (+) play, then type of hit.

Implications to this discussion-
1) K rate - Because Ks are a subset of outs, the more outs a pitcher produces, the higher the probability that he achieves more strikeouts. Also, because the out has already been determined, there should be a strong correlation between velocity and K rate. In my experience, there is, but only if that pitcher is good enough to generate outs without the velocity. The pitcher-batter interaction is skewed towards batter control, so a league with higher contact ratings amongst the starting batters will have an adverse impact on a pitcher's K rate.

2) infield hits - the description of the hit is added after the hit type is decided. For infield singles, the player already has the single, but based on something, likely speed, the sim assigns the infield hit designation instead of a regular hit. If a play doesn't have a (-), it wouldnt have been a hit regardless. The same is true of the inside the park HR. The HR is decided at the beginning, the type of HR is added later, likely based on the player's speed vs power ratings. You see a correlation between a player's speed an infield hits and attribute that to the infield hits being earned by that speed. In reality, the speed is simply causing the PBP to kick out "infield hit" rather than "singles in front of the left fielder".

Again, this is the tree as it was designed for SLB, so there could be alterations, but I think I have seen enough evidence to show that this is still pretty spot on.