Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...
Just doing some spreadsheet work on strikes/balls (in zone %) by pitch type, and came up with a back of the napkin method of assigning control and break to pitches, absent pitch fx data.
[1] Calculate a base control rating (BB9*.67+HR9*.33)
[2] Calculate a base break rating (k9*.67+H9*.33)
Essentially operating on the idea that better located pitches are harder to hit out, and better moving pitches are harder to contact in general.
When I looked at it in the spreadsheet the usual suspects were at the top for control (Hallday, Lee, etc) and same with break (Yes, Marmol is the devil).
[3] Multiply base control by a modifier for each pitch type (based on the frequency those pitches are in the zone per MLB average).
4SFB = CON * 1.03
2SFB = CON * .98
CUT = CON * .96
CURV = CON *.93
SLID = CON * .95
SPL/FK = CON * .81
CHANGE = CON *.84
KNUCK = CON * 1.12
To my surprise a greater % of knuckleballs are in the zone than fastballs, and changeups and splits are the least likely pitches to be "in the zone", but given their usage that could be expected (start in the zone, drop out). Nobody throws a "backdoor" splitter (and lives to tell about it).
[4] Pitchers with good movement, generally have good movement all the way around so the modifier is applied to pitches in terms of their usage (or value).
BEST PITCH = BRK * 1
2ND BEST = BRK * .95
3RD BEST = BRK * .9
4TH BEST = BRK *.85
5TH BEST = BRK * .8
Again, looking at it in the spreadsheet ... everything looks good. The question is how the values will work in the game, but again after 40 hours of entering rating and rating relationships, the values pass the sniff test. Nobody with a 67 k9 rating will have a pitch with a break of 92, so while playing the game, the tough pitchers will be the hardest to hit, and the average ones will still be average. There also won't be anything like Greinke's curveball on MVP05 (remember that 65mph huge bender?).
I'll use this for minor leaguers whose rating increased significantly from 2010 to 2011, or newly created players. One just needs to know what their best pitch is ...
Another source is OOTP ... where individual pitches have rating to along with a "stuff" rating, and control. The program essentially places a "pitch modifier" on the stuff and control ratings. Similar process.
Fangraphs and other sites can be used to get individual pitch values, that someone could work a formula out on, but it'd probably just be easier to use bahnzo's formulas while anually entering pitch fx data.
I'm just working on formulas that can give a realistic estimate given a pitchers repitoire and his general stat and rate lines.
============================
Edit: Marmol's slider in OSFMv2 is CON: 42, BRK: 98. With this formulas it's CON:55, BRK: 94. Close ... and you can't hit either version ... nor can you lay off of it as it goes of the corner with 2 strikes. *big grin*
Obviously, Lincecum, Strasberg, Kershaw, Champman, etc throw some devestating pitches ... Lincecum and Strasburg multiple devastating pitches.
Just doing some spreadsheet work on strikes/balls (in zone %) by pitch type, and came up with a back of the napkin method of assigning control and break to pitches, absent pitch fx data.
[1] Calculate a base control rating (BB9*.67+HR9*.33)
[2] Calculate a base break rating (k9*.67+H9*.33)
Essentially operating on the idea that better located pitches are harder to hit out, and better moving pitches are harder to contact in general.
When I looked at it in the spreadsheet the usual suspects were at the top for control (Hallday, Lee, etc) and same with break (Yes, Marmol is the devil).
[3] Multiply base control by a modifier for each pitch type (based on the frequency those pitches are in the zone per MLB average).
4SFB = CON * 1.03
2SFB = CON * .98
CUT = CON * .96
CURV = CON *.93
SLID = CON * .95
SPL/FK = CON * .81
CHANGE = CON *.84
KNUCK = CON * 1.12
To my surprise a greater % of knuckleballs are in the zone than fastballs, and changeups and splits are the least likely pitches to be "in the zone", but given their usage that could be expected (start in the zone, drop out). Nobody throws a "backdoor" splitter (and lives to tell about it).
[4] Pitchers with good movement, generally have good movement all the way around so the modifier is applied to pitches in terms of their usage (or value).
BEST PITCH = BRK * 1
2ND BEST = BRK * .95
3RD BEST = BRK * .9
4TH BEST = BRK *.85
5TH BEST = BRK * .8
Again, looking at it in the spreadsheet ... everything looks good. The question is how the values will work in the game, but again after 40 hours of entering rating and rating relationships, the values pass the sniff test. Nobody with a 67 k9 rating will have a pitch with a break of 92, so while playing the game, the tough pitchers will be the hardest to hit, and the average ones will still be average. There also won't be anything like Greinke's curveball on MVP05 (remember that 65mph huge bender?).
I'll use this for minor leaguers whose rating increased significantly from 2010 to 2011, or newly created players. One just needs to know what their best pitch is ...
Another source is OOTP ... where individual pitches have rating to along with a "stuff" rating, and control. The program essentially places a "pitch modifier" on the stuff and control ratings. Similar process.
Fangraphs and other sites can be used to get individual pitch values, that someone could work a formula out on, but it'd probably just be easier to use bahnzo's formulas while anually entering pitch fx data.
I'm just working on formulas that can give a realistic estimate given a pitchers repitoire and his general stat and rate lines.
============================
Edit: Marmol's slider in OSFMv2 is CON: 42, BRK: 98. With this formulas it's CON:55, BRK: 94. Close ... and you can't hit either version ... nor can you lay off of it as it goes of the corner with 2 strikes. *big grin*
Obviously, Lincecum, Strasberg, Kershaw, Champman, etc throw some devestating pitches ... Lincecum and Strasburg multiple devastating pitches.
Comment