An example of why I don't think DIPS will reflect pre-1960 pitchers accurately

[dougaiton]

Originally posted by Carlton
Big Train,

I believe it has to do with the fact that Rommel was the first "true" knuckleballer...which no simulation can acurately produce.

You're a smart man, Carlton. I totally forgot Rommel was a knuckleballer.

BigTrain, I'd like to see your results for Rommel in an OOTP sim. I've seen a few knuckleballers produce very realistic BABIPs
over time.

I think people think that BABIP in OOTP6 always hovers around the league average, but this isn't the case. There's as much variance built around it as we'd hope for, and I've seen a few guys who outperform their expected total. Markus said there would be a modifier, but it could just be luck. Either way, I'm happy

[Skipaway]

Originally posted by Big Train

I just intuitively and logically can't accept that the *only * difference between pitchers was control and defense, it's just too simplistic and illogical an argument.

Errr, the most illogical part of this is assuming it could not be simplistic. Any logical reason it can't be?

Intuition IS illogical.

[dougaiton]

Originally posted by Big Train
[BIt isn't the whole idea of DIPS that I have a problem with. Pitchers who strike people out a lot and don't walk many people and don't give up many homers and have a great defense playing behind them obviously have an advantage especially in a modern context. It's people who go to the extreme of saying that there is absolutely no ability (or a very insignificant ability) of the pitcher to prevent balls in play independant of the defense that I disagree with and I think Mr. Tippet has done a much better job than I could of proving that assertion false in the above article.

I'll concede that the new system with a BABIP rating for each pitcher would probably be more accurate than the old system but I'll settle for one or the other [/B]

I think that even Tippett would argue that while the ability may exist, it is virtually insignificant. This is an important point - while some may excel, the others may not.

The second point is excellent. DIPS does not promote a static formula of K+BB+HR=runs scored. It's just that those three stats are more indicative of ability to prevent runs scored than anything else.

What DIPS means for OOTP is a new way to record outs: work out whether its a BB/K, then calculate the chance of a hit. This is fantastically important in comparison with the old system. So what we are left with is a debate of how to discuss the hit calculation.

I don't think BABIP necessary answers the question any more than BA answers the question of 'what is the true talent level of the hitter'? However, it is a rough approximation of 'what happens', even if it gives us no idea of 'why'? More research may indicate - there are some of us that advocate defensive changes and just pure luck/physics, while there are some that think pitchers have more control. However, the central tenent is what's important - that hits, BBs, Ks and HRs are all interconnected, and its very hard for a pitcher to have consistent success relying on the first, while significantly easier if relying on the last 3.

[Big Train]

I also found the same type of walk/strikeout ratios with pitchers who to my knowledge weren't knuckleballers:

Tom Zachary 1920-1936 [season average] 225.3 IP 258H 65 BB 51K 3.73 ERA

Hugh McQuillan [9 years] 219.7 IP 239H 68BB 62K 3.83

Slim Hariss [8 years] 206IP 231H 74K 75BB 4.25

I don't feel like listing them all but nearly every pitcher from this year has a 2 rating for strikeout
[Betts,Vangilder, Dixie Davis and Rip Collins are some others you can look up]. It's not only knuckleballers though.

None of these were great pitchers but remember the 20s and 30s was a great hitters era so their numbers are probalbly a little inflated.

[Big Train]

Originally posted by dougaiton
.

I think people think that BABIP in OOTP6 always hovers around the league average, but this isn't the case. There's as much variance built around it as we'd hope for, and I've seen a few guys who outperform their expected total. Markus said there would be a modifier, but it could just be luck. Either way, I'm happy

The important thing for me is not variance but that the *right* pitchers for the most part have a lower BABIP. Since I try to produce reasonably accurate results, if Joe Schmo who was terrible in real life suddenly has a lights out BABIP and if this happens quite often...total randomness in who pitches well is what I really want to avoid.

[Skipaway]

Originally posted by Big Train
I also found the same type of walk/strikeout ratios with pitchers who to my knowledge weren't knuckleballers:

Tom Zachary 1920-1936 [season average] 225.3 IP 258H 65 BB 51K 3.73 ERA

Hugh McQuillan [9 years] 219.7 IP 239H 68BB 62K 3.83

Slim Hariss [8 years] 206IP 231H 74K 75BB 4.25

I don't feel like listing them all but nearly every pitcher from this year has a 2 rating for strikeout
[Betts,Vangilder, Dixie Davis and Rip Collins are some others you can look up]. It's not only knuckleballers though.

None of these were great pitchers but remember the 20s and 30s was a great hitters era so their numbers are probalbly a little inflated.

And what are you trying to say with these pithcers.

I did some quick check and found out that:

Tom Zachary: (H-HR)/(3IP+H-HR-SO) varied from .255~.313 in years with more than 100 IP.

The funny thing is, his ERA went from 3.12 to 4.49 to 2.75 from 1922~1924, and his crude BABIP went from .262 to .313 to .255 at the same time.

It's really hard for me to believe Tom Zachary got much control over his BABIP.

[Skipaway]

Originally posted by Big Train
The important thing for me is not variance but that the *right* pitchers for the most part have a lower BABIP. Since I try to produce reasonably accurate results, if Joe Schmo who was terrible in real life suddenly has a lights out BABIP and if this happens quite often...total randomness in who pitches well is what I really want to avoid.

Is it possible this Joe Schmo was terrible in real life simply because he was unlucky and got the wrong defense behind him?

Perhaps OOTP6/DIPS evaluates player talents better than you do?

[Big Train]

Skipaway, OOTP6 doesn't "evaluate talent" The only reason it knows how to rate Joe Schmo is because the database tells it his stats. If his stats over several years show that he was bad the odds are very low that he was just constantly unlucky.

But anyway there is luck in every facet of the game, hitting and fielding require luck as well as skill. aAfraction of an inch often determines whether a ball is a hit or a strike or an out. Fielders routinely just miss balls or happen to be positioned perfectly to make a play they couldn't otherwise make.

Your version of DIPS seems to take away a lot of the credit the pitcher rightfully deserves. A fielder catching a popup in the infield is showing great skill but a pitcher only shows his skill on the few times during the game where he strikes someone out. I doubt a major league infielder is proud of himself when he catches a pop-up. It's expected of him. I'm the first one to applaud a fielder who clearly takes away a hit from someone but on making a routine play he deserves little credit, he wouldn't even be in the majors if he couldn't make routine plays most of the time.

Hitters get lucky quite often too yet no one suggests that some
hitters aren't better than others. Hitters who produce stats near the top of the league every year must have some talent besides luck, most people would except that. So why is it so hard for people to accept that pitchers who don't give up many hits might have some talent besides luck which makes their pitches harder to hit? And sometimes they don't strike out the guy because he is able to make partial contact. [Before you make the argument about Pedro and Maddux's BABIP fluctuating I'd like to point out that *all* stats fluctuate. There are countless examples of hitters who hit .300 one year and .220 the next]

I think Tippet is right on and by the way he specifically said that some pitchers' ability to avoid hit on balls in play is *not* statistically insignificant.

Oh and Zachary et. al were just illustrations of more "decent" pitchers with very low strikeout totals.

[renn]

I'm a newbie, so be kind, but what does DIPS and BABIP stand for? Thnaks

[643gdp]

Sure. . .DIPS = Defense Inthe Pitcher Sense
and
BABIP = Balls and Bats In Play

Hope that helps!

[renn]

Thanks!

[Henry]

Originally posted by 643gdp
Sure. . .DIPS = Defense Inthe Pitcher Sense
and
BABIP = Balls and Bats In Play

Hope that helps!

LOL....

DIPS = Defensive Independent Pitching Stats

BABIP = Battings Average of Balls In Play

[oldtime]

Just to jump into the fray here, it seems like there are two parallel arguments going on here at once:

1) Are DIPS a valid IRL?
2) Does the DIPS engine produce valid historical stats in OOTP?

For question #1, let's go back to what DIPS stands for: "defense independent pitching stats". In other words these are the only stats not influenced by defense. That we can all agree on.

As a former pitcher through my college years, I have to believe that a good pitcher can keep the ball off the good part of the bat, and thus has influence over how hard ball are hit----giving his defense a better or worse chance of making a play---so I am skeptical of DIPS as a be all/end all tool to evaluate a pitcher. After a ball is hit, the pitcher has no control, and defense, luck, and the skill of the hitter take hold. But I do believe the pitcher has some influence over how hard it is hit. Otherwise you would have to conclude that flyball/ground ball % is just luck. If a pithcer can influence if the ball is in the air or on the ground, it only stands to reason that they influence the chances of it being a hit based on soid contact or not. How much and how to quantify that statistically, eliminating defense as a factor is the problem in a game with so many variables like baseball.

Question #2:

Does anyone have valid results they can share of with us, about early 1900 pitchers not being able to produce historically accurate numbers and success on a consistant basis??? (we understand the variables in the game allow for great pitchers to have bad years and average pitchers to have great years) Do some of the real life pitchers mentioned above, NEVER or RARELY produce accurate stats due to the reasons mentioned, or is this conjecture that they will not??? I am not accusing, just asking.

My point being that while DIPS has many holes in my eyes if used as anything more than "another tool" to evaluate pitching performance, it very well may produce very accurate recreated stats and results in OOTP. I don't need to know the entire formula that Markus uses, I just need to seet the results it produces. They are either good or bad, and there is no argument from that point on.

The one thing I like about the DIPS engine is that it seems to quantify defense very accurately in played games (at least to my liking) something that is tough to quanitify statistically.

Again, anyone out there seeing statistical results that are just not accurate for individual players---once you have the league totals set right so the league is performing accurately???

[Henry]

I think this thread (long) will answer the question about realistic results. In addition, I'm sure TigerFan would look up individual pitchers if asked.

Tiger Fan's Historical Tutorial and Feedback League

[oldtime]

Thanks Henry--

My point though is that while we can argue DIPS, I have yet to see much in the way of concrete proof that engine does not produce good results. Would you agree???

[Henry]

Yes, I would agree. From the limited testing I've done in this area, I see similar variances of OOTP results for these pitchers as in real life. If you total up their entire careers, you can see a "very small" error - but it's seems to be only limited to a handful (12 or so) pitchers throughout history.

[IatricSB]

BigTrain,

I agree with you 100%. The whole DIPS thing has left a bad taste for me and from what I've seen has not worked all that great with historical leagues. I find it odd that OOTP jumped on the DIPS bandwagon given that its such a relatively new idea (compared to other sabermatric stats).

I'm regretting the decision to move Boys of Summer to OOTP6. I would love to be able to use the OOTP5 engine in v6, but that's not possible. The conversion was done in a way to try to adjust a pitchers ratings to fit better into the DIPS system. So you end up with players rated based on one system while trying to play using another system (in addition to the fact that some fields like Avoiding Hits are no longer there).

I've read all the arguments both ways and tried to have an open mind to it. In fact, I enjoyed DIPS when I first used it. But the longer I used it, the less I liked the results in historical leagues.

[Henry]

Well... I actually like it the more I use it LOL - To each his own I guess. I will admit that editing historical ballplayers with the DIPS engine is tougher than the old ratings system, but the game results sure end up more realistic. It also plays a huge advantage to fictional leagues in the reality factor. I think if we can add some further interfacing for creating and modofying players, this will ultimately be a moot point.

Here's something I wrote back when the BIG discussion was going on - when somemone brought up the fact that Charlie Hough was a HUGE variance from the DIPS model. I think it's apropriate to repeat it here...
==========================

Analysis by Henry

Ok. Let's try and do a number of things here...

1. Try and define what "significant" means,
2. Convert Tom Tippett's latest article on DIPS to some meaningful statistics, and
3. Talk a bit about what it would mean to have to add this modifier to OOTP.

It is my opinion that there are various reasons why this "ability" is insignificant and why it doesn't belong in OOTP. I'm going to try to express those opinions here.

Let's start with Tippett's conclusions. He listed the top 10 pitchers that had the most control over BABIP. The table I chose to use is the BA point difference. Tippett remarked that (in Hough's case) a .026 point difference in BABIP was "significant".

Pitcher            IPAvg vs Lg  IPAvg vs Tm
-----------------  -----------  -----------

Charlie Hough         -.032       -.026
Don Wilson            -.015       -.023
Andy Messersmith      -.033       -.021
Ned Garver            -.008       -.020
Tim Wakefield         -.020       -.019
Catfish Hunter        -.028       -.017
Bud Black             -.020       -.017
Oral Hildebrand       -.015       -.017
Walter Johnson        -.021       -.016
Dave Stieb            -.022       -.016

I'm going to break down Hough's .026 into something more understandable.

Hough played 24 years (1970-1994). The first 9 years were as a reliever, the last 15 as a starter. During his reliever years, he gave up a BAA of .245 and during his starting years a BAA of .296. The only reason I point this out is because he was much more effective as a reliever which is probably one reason he leads the list above.

His career stats averaged to a 162 game season are...

IP	H	HR	BB	K
------- ------- ------- ------- -------
199.0	171	20	87	123

If we remove his HRs from his hits (171-20=151) that results in a .253 BABIP. If we then add Tippett's .026 we get a .279 BABIP. Each of these calculated over the course of a 162 game season results in a 15 hit difference (166 hits - 151 hits). Hough pitched in an average of 44 games per season (remember he was a reliever for 9 years). 15 hits divided by 44 games is about 1/3rd of a hit per game - or 1 hit per three games. These 15 hits would break down to 12 singles, 2 doubles, and maybe 1 triple (I used the 1979 1b-2b-3b ratios, although all years in that era were similar) over the course of a season.

Dave Stieb, the 10th player on the list calculates out to .31 hits per game. Note that his .016 figure compared to Hough's 0.26 figure didn't mean he had that much less effect.

What it DOES point out however, and what makes perfect logical sense, is that a list of all pitchers would range from Hough on the top at 0.26 and someone at the bottom with maybe -.026 ... in other words, the figure has to average to zero. So we have Charlie on top maybe getting 15 hits per season too much, and someone at the bottom of the list getting 15 hits per season to few.

What this means to us is we have a two edged sword. If we're going to adjust for the Charlie Hough's of the world, we also have to adjust for the "no names" of the world as well - to keep everything in balance.

Ok.... since the Lahman databse doesn't have anything we can use here directly, the only way we could add a modifier to correct for this would be to run each and every pitcher loaded from Lahman through the above calculation Tippett went through - or have access to his database. I'm not sure if either is possible (Trippett would have every right not to make his DB available, and Markus may or may not be able to come to the same conclusions himself.

In either case, this would be a major effort - which brings us back to a basic fact.

You've all heard this before, but OOTP was not designed to "replicate" past seasons. DMB does that much better. OOTP was designed to provide historical players with similar abilities to allow us to run "alternate" scenarios. Once a player is input into OOTP, other variables take over. Career curves are applied (not necessarily the same ones that played out historically). As the season progresses, talents and ratings change based on those curves, injuries, etc. Basically, there is no claim made that any historical player will follow the same or similar path that he did historically. He is only given the initial abilities when he is first loaded into OOTP.

To me, that means any attempt to correct for this "BABIP ability" is lost in the variances of the game.

Conclusion

Is there a plus/minus relationship in player ability to control BABIP? - Yes, but it is minor.

Is it worth investigating how to adjust for this ability? - No, since there are other game design variables that would drown out the effect.

Henry

[jbmagic]

Originally posted by IatricSB
BigTrain,

I agree with you 100%. The whole DIPS thing has left a bad taste for me and from what I've seen has not worked all that great with historical leagues. I find it odd that OOTP jumped on the DIPS bandwagon given that its such a relatively new idea (compared to other sabermatric stats).

I'm regretting the decision to move Boys of Summer to OOTP6. I would love to be able to use the OOTP5 engine in v6, but that's not possible. The conversion was done in a way to try to adjust a pitchers ratings to fit better into the DIPS system. So you end up with players rated based on one system while trying to play using another system (in addition to the fact that some fields like Avoiding Hits are no longer there).

I've read all the arguments both ways and tried to have an open mind to it. In fact, I enjoyed DIPS when I first used it. But the longer I used it, the less I liked the results in historical leagues.

you can change the engine in ootp 6 to traditional and it wont have dips..if that what you prefer...it will use the old engine...

[Henry]

Jbmagic.... please note

Originally posted by IatricSB
I would love to be able to use the OOTP5 engine in v6, but that's not possible. The conversion was done in a way to try to adjust a pitchers ratings to fit better into the DIPS system. So you end up with players rated based on one system while trying to play using another system (in addition to the fact that some fields like Avoiding Hits are no longer there).