Roster Generation Part 2!

[alexfall862]

Hi! 

Trying something a little bit different this week - I made a video walkthrough of the dev diary where I dive into the source code and talk through the mechanics of the features outlined below! I've not done one since the OG google sheets recruiting walkthrough - check it out here:

https://www.youtube.com/watch?v=0V_d4HyeEF4

Last dev diary covering player generation we talked about the attributes tied to integers, which leaves out a couple of other aspects of player evaluation - so today is about diving into that realm!

L/R Handedness

~20% of players pitch, throw, and bat left-handed, while ~60% do so right handed. The remaining group are mixed between position players listing their throwing due to positional requirements (infielders typically throw right handed) or are genuinely mixed between being lefty, righty, or switch. 

To capture this, at least roughly, players will be generated with a 20/60/5/5/5/5 split with all left, all right, l/r, r/l, s/l, and s/r respectively.

Switch hitters historically have stabilized around 10-15% of batters in the post-2010 era, and while 10% is on the low side, I think the slight competitive advantage of not facing platoon split issues will likely have more switch hitters on MLB rosters (and someone in the sim will probably make a team of entirely switch hitter or something)

I’m not going to add switch pitching to the sim at the moment as I’m not entirely sure how to handle the case where a switch pitcher and switch hitter face each other and don’t want to add recursive loops that cause my computer to melt instead of simming games.

I’ll probably create some kind of position change check on throwing hands down the road, but not in the scope of the MVP for season 1.  

Batter Height/Weight

Height impacts a couple of things in baseball:

Your strike zone expands (particularly on high strikes)
Your reach expands (particularly on outside balls)

Heavier/taller batters hit for more power. (weirdly the study on this was published in a medical journal?!) 

The way we’ll work this, instead of having height/weight become a deterministic statistic is that higher power players will be more likely to have a taller height and heavier weight assigned to them. 

Being taller will benefit a player slightly in the size of their ‘reachable’ zone when swinging outside the strike zone, but will also hurt them as pitchers will find it slightly easier to throw strikes, particularly with breaking pitches.

That doesn’t mean other players are short changed though - as the strike zone shrinks as the batter shrinks. Shorter players will have an easier time getting balls called and will be slightly more likely for breaking pitches to slip out of the zone.

Height will also play a very small malus/boost in the field as well, particularly as a boost when basemen are catching (Altuve at first is a much smaller target than Morneau). 

Interestingly, for pitchers the common wisdom is that a taller pitcher has a faster apparent speed on their pitches due to a long step when pitching - but from a performance standpoint the outcome doesn’t change very much in any domain based on height, so pitchers are immune from this effect for now.

The way that this will work under the hood is that players will be assigned a height by a normal distribution of heights from baseball players (mean height is ~6’1”), with the higher power potential players pulling from a higher mean value, and the lower power potential players pulling from a lower mean value. The player will then get a BMI assigned to them to generate a weight as well.

This way the rosters and players retain some ‘real life’ logic in their construction (no 4” LBers), while retaining the capacity for smaller players to demonstrate power (Yogi Berra) and for ‘bigger’ players to show relatively low pop (Billy Butler, Pablo Sandoval)

How Specific Pitches are Generated:

Pitches fall into 3 broad categories, with the specific pitches showing slight : 

• Fastball - slightly more likely to generate harder contact, but also slightly more likely to generate bad contact and misses.
• OffSpeed - slightly more likely to generate quality contact, but also more likely to generate misses
• Movement - much more likely to generate harder contact, but much more likely to generate misses and bad contact
• #4 - Can be any pitch in the above 3 categories
• #5 - Can be any pitch in the above 3 categories

Players may not get 5 pitches developed to the point of being consistently effective (or effective at all), so users will select the pitches and desired approach of each pitcher (e.g. 60% pitch 1, 20% pitch 2, 20% pitch 3, 0% pitch 4, 0% pitch 5)

The pitches in the sim right now are as below:

Spoiler

Fastball:
Four-Seam Fastball
Two-Seam Fastball
Cutter
Sinker 

Offspeed:
Change Up
Circle Change
Splitter

Movement:
Curveball
Screwball
Slider
Slurve 
Knuckle Curve

Other: 
Knuckleball
Eephus

Arm Slots

Pitchers have a slight advantage in getting called strikes when throwing from deeper angles. The work on this isn’t sure the cause is from something like an intrinsic benefit as far as arm angle, or whether umpires/hitters are just weirded out by the novelty of a pitcher mowing the grass around the mound with his off hand. 

That said, the effect is real and while pitchers generally vary their arm slot pitch to pitch, most show a general tendency around 1 delivery type. 

Overhead, 3/4, Sidearm, and Submarine will each have slight effects on pitcher’s performance, with 3/4 and Overhead being most common and some sidearm and submariners being generated as well.

Archetypes

Another area of player characteristics/value that will carry with them through their career is an ‘archetype’. 

This is a way to get a quick summary of a player’s profile without having to dig through their entire ratings portfolio. It will be regenerated each offseason to capture how the player plays right now with a slight weighting towards their growth potential in those stats as well. 

Different than the simFBA archetypes, these are purely descriptive and do not play a role in player generation, but are instead a way to quickly process rosters and players. The names of these are still being compiled, but think ‘speedster’ for players with only speed as a ++ skill or ‘patient hitter’ for hitters with solid eye and discipline.

Clutch-ness

Baseball nerds have consistently pointed out that leverage or batting average with RISP isn’t something that predicts future performance in similar scenarios - in fact the batting average overall is more predictive (which makes sense if you figure that the ‘clutch’ scenarios are a much smaller sample size).  The performance of 1 player or another that far outclasses their peers do not replicate themselves year to year.

So tentatively, I’m not going to include a clutch rating into the sim at this moment. 

Pinch-Hitting

Similarly, while the concept of a player being more adept at pinch hitting is a long standing baseball trope, there is only a slight to null effect on a batter’s ability to successfully pinch hit compared to their starting line. 

This effect is also somewhat handled by the ‘n times through the order’ effect on pitchers, where pinch hitters will not have faced this particular pitcher in the game (by virtue of being a pinch hitter) and thus has a comparatively lower ability to succeed than batters that haven’t. 

Last Week's Poll

No particularly overwhelming preference from folks for or against a particular concept, although one suggestion that came up a fair bit were bunts and sac hits. I think that's a great idea and popped it on the list for future seasons, but will probably not get a 'logic' for when to do that sorted out for the first season.

 

This Week's Poll

Since we covered player generation already, there's not a ton to cover, but a brief poll to help collect feedback!

https://forms.gle/Bo7Rqw3M4WRk9wcr7

Thanks,
alexfall862