Special Report - Kenny Powers: Pitching Analysis by the Numbers

As the season gets under way, here is a special report put together by cosborn detailing his methods for analyzing Starting Pitchers.  Thanks to cosborn for his contribution to the blog!

Kenny Powers: Pitching Analysis by the Numbers
by: cosborn

Last season, I posted a few comments in the league chat about pitchers, pitcher ratings and my method or system for evaluating pitchers. I’ve also (as is usual each season, worked to refine the formula a bit more, and the following is based upon a couple slight tweaks to the system.)

When I started trying to figure out the correlation between pitcher ratings and pitcher performance, I wanted to find pitchers I could use for my study. (I’ve run this exact method in several leagues, and the results are nearly exactly the same, BTW.) It turns out that I had two pitchers for the last four seasons - BC Escobar and Casey O’Neill.

Escobar has a steady OVR of 77 and O’Neill an OVR of 70. If OVR works, Escobar ought to be a better pitcher than O’Neill. The actual performance of the pitchers says, “No.” In fact, O’Neill might be a little bit better. Why? What ratings actually serve as lead indicators of performance?

Some ratings appear to be obvious such as control and splits. But what is the effect of pitch quality? How many good pitches does a pitcher need? Are there any interrelationships between and among the ratings? What about velocity and ground ball to fly ball ratio?

Here are the cumulative numbers for O’Neill and Escobar for seasons 19-23:

	Sea	Innings	Wins	Losses	Opp BA	Opp Slg	Opp OB	WHIP	ERA
Escobar	20	203.33	10	14	0.254	0.338	0.396	1.43	4.51
	21	211.33	9	13	0.260	0.331	0.422	1.34	4.39
	22	218.33	12	11	0.241	0.304	0.372	1.22	3.96
	23	210	12	9	0.253	0.324	0.398	1.34	3.69
O’Neill	20	206.67	12	14	0.253	0.331	0.427	1.41	3.92
	21	222.67	11	13	0.261	0.325	0.414	1.34	3.96
	22	217	15	7	0.253	0.337	0.377	1.42	3.77
	23	219	15	9	0.233	0.317	0.377	1.31	3.62

As you can see, neither of these starting pitchers are dominant #1 types. Both, however are useful and reasonably effective. But the differences between the pitchers, to the extent differences exist at all, certainly do not reflect the fact that over this entire stretch of four seasons Escobar’s OVR was 77 and O’Neill’s 70.

Without getting really deep into the weeds, here are the ratings that make their way into my system: durability, stamina, control, splits, velocity, GB/FB ratio, and each pitch rating from 1 through 5. Ratings that are excluded have little to no impact I could find on pitcher effectiveness. Some of these ratings - unless “extreme” - affect pitcher performance. Extreme is defined for my purposes is a category rating of 90 or higher.

The three big ratings are exactly as you’d expect, and they apply in this order: control, RH split and LH split. As it turns out, unless a pitcher has AT LEAST a control rating of 50, it’s virtually impossible for the pitcher to realize consistent success, even with ALL other ratings being A+ to extreme. For the splits a COMBINED total rating of a MINIMUM of 120 is required PLUS the minimum RH split rating MUST be 60 or higher. So, a pitcher with an 80 LH split, but a 55 RH split would not make a good ML caliber starting pitcher. (Please note, I am ONLY talking about SP. Relievers have a much different system . . . .)

There are rating levels in these areas that appear to return a premium in terms of effectiveness as measured by WHIP, App BA, SLG and OB. For control, the premiums appear at 70, 80 and 90. AND the premium is cumulative. In my system, a rating of 70 adds a full point to a pitcher’s final SP rating. At both 80 and 90, the pitcher gets an additive of .5 at each level. So a pitcher with a control of 91 has a total control value of 2 added to his total SP rating. At ANYTHING below 50, the pitcher rating is -3. A hit like that makes it nearly impossible to top 2.0 on the final rating.

For splits, I start with a blended rating by simply adding the LH and RH split together. Next I add 15% of the RH split IF it is 60 or above to the total. The first premium worth a full point is at a combined total splits of 140. Then another premium is earned at 180 or above. Like control, the premiums added are cumulative. So if a pitcher has an adjusted combined split of 185, he earns 2 full points added to his SP rating.

Okay - so let’s look at these three rating within the system I use and compare Escobar to O’Neill.

	OVR	Con	LH	RH	Con Bonus	Split Rating	Split Bonus	Rating So Far
Escobar	77	70	78	65	1	1	0	2
O’Neill	70	51	70	73	0	1	1	2

If we looked exclusively at OVR, Escobar ought to be a better pitcher - by a fair margin. But when we look at these ratings, these two pitchers are virtually indistinguishable. Escobar has much better control (we’ll get into more on that in a moment), but O’Neill has a better set of splits ratings - even thought the sum of their respective splits is identical at 143. Since both at at 140 or better, they get a split bonus. O’Neill gets an additive bonus due to his 70+ versus righthanded hitters. NOTE: the ratio of the bonus EXCEEDS the ratio of RH hitters in the league. However, the data says this is likely a fair bonus.

Now let’s look at the results, and from there we’ll get into the remaining ratings that matter - pitch quality.

Escobar	IP	W	L	OBA	OB%	SL%	WHIP	ERA
20	203.33	10	14	0.254	0.338	0.396	1.43	4.51
21	211.33	9	13	0.260	0.331	0.422	1.34	4.39
22	218.33	12	11	0.241	0.304	0.372	1.22	3.96
23	210	12	9	0.253	0.324	0.398	1.34	3.69
O’Neill	IP	W	L	OBA	OB%	SL%	WHIP	ERA
20	206.67	12	14	0.253	0.331	0.427	1.41	3.92
21	222.67	11	13	0.261	0.325	0.414	1.34	3.96
22	217	15	7	0.253	0.337	0.377	1.42	3.77
23	219	15	9	0.233	0.317	0.377	1.31	3.62

Over this 4-season sample, Escobar is 43-47, and O’Neill is 53-43. Their average records are Escobar 10.75 - 11.75 (.477 winning %), and O’Neill 13.25 - 10.75 (.552 winning %). Over this same 4-year span, the team has a cumulative winning % of .508. There are many ways to look at the value of a pitcher, but one number that is worth looking at is the winning percentage of the pitcher compared to the team. You’d like a top pitcher to make your team better, or expressed another way, good pitchers put their teams in positions to win more often. This type of analysis also bakes out the expected differences between pitchers on excellent teams compared to god pitchers on poor teams. That’s why wins and losses standing alone are NOT good measures of pitcher quality. In my view, it’s how the pitcher’s performance improves his team that is a better perspective. (For example, many experts think Steve Carlton’s 1972 season with the Phillies might be the best single season performance in MLB history.)

Now - I KNOW the sabermatrics of this are wildly complex, and I am radically simplifying things. Sorry. I simply don’t have the type of data needed to calculate WAR for these pitchers. So, I am looking at a very simple comparison of the pitcher’s winning % to the team winning %. I thin this is fair for the following reasons: 800+ innings pitched each, for the same team, same defense, same park and same run support. The sample size is large enough to bake out any anomalies that would occur over one or two season.

But - WHY did O’Neill do so much better? The control and splits appear to even things out between the two pitchers. O’Neill pitched 32.1 more innings over the 4-season span. This is the product of O’Neill’s excellent durability and stamina ratings of 31/91 respectively. Escobar’s ratings are 27/82 - very good, but not as strong as O’Neill’s. While O’Neill’s mediocre control rating of 51 means he has to work harder for outs, the workload of the two pitchers is close with O’Neill holding an 8 inning per season edge. So - I don’t think we can attribute the difference in results to these ratings.

Let’s go to pitch quality, which is where the distinctions become more apparent.

There are 5 possible pitches, but I cannot find ANY correlation at all to the existence or quality of a 5th pitch on predicting starting pitcher performance. Zero. But, there is a difference when SP have a 4th quality pitch. Here’s how the pitch quality works.

SUM of Pitches 1 and 2: if below 140, pitcher has a 2.0 point penalty assed. Yes - 2 full points. These first two pitches are very important! At 140-149 - neutral, 150-169 = .5 increase, and 170+ earns another .5.

Examples: Pitch 1 - 90, pitch 2 - 50 for a sum of 140. NEUTRAL - no penalty and no increase. Next: pitch 1 - 82, pitch 2 - 70 for a sum of 152 earns the pitcher .5 increase.

The magic number for pitch quality is 70. That’s more or less the threshold for ML quality pitch quality.

Next, I go to the combination of pitches 1-3 and 1-4, and YES. this is a cumulative set of additions and penalties. In RL (and the programmers appear to have this pretty well correct) pitchers with 3 quality pitches are generally much better than those with 2, and SP with 4 highend pitches are REALLY good. The pitcher simply has more ways to keep hitters off-balance and guessing, and it appears that this reality is accounted for in the pitching algorithms used in the game engine.

Here’s how the next pitch combinations work:

Sum of 1-3: Below 210 = .5 penalty; 220+ = .5 bonus, AND 240 + = .5 bonus

Sum of 1-4: Below 240 = .5 penalty (note the AVERAGE of the 4 pitches needs to be 60 per pitch. Pitchers with 3 excellent pitches can still be effective starters, but it’s more difficult for them.) 250+ = .5 bonus and 280+ (a 4-pitch average of 70 per pitch) = another .5 bonus.

Here’s how Escobar and O’Neill compare: Escober at 5.0 O’Neill at 5.5. Therefore, based upon my system, O’Neill at OVR 70 is a better pitcher than Escobar at 77 OVR. Do the qualitative numbers support this statement or rating method?

Escobar	AVG	OB	SLG	WHIP	ERA
20	102.36%	96.45%	104.80%	95.10%	94.46%
21	99.23%	98.19%	97.16%	101.49%	96.36%
22	107.88%	96.45%	111.56%	111.48%	107.58%
23	103.16%	100.93%	104.77%	102.24%	115.72%
Cumulative Difference	12.64%	-7.99%	18.29%	10.31%	14.11%

This data represents the comparative difference between Escobar’s actual performance to the National League’s average in each category. Escobar did well compared to the league is 4 of the 5 areas, but was about 8% worse than the league in ob base percentage allowed. The differences in slugging and WHIP are likely attributable to the ball park. (This is why I wish we could view WAR for pitchers in the extended stats!)

O’Neill	AVG	OB	SLG	WHIP	ERA
20	102.77%	98.49%	97.19%	96.45%	108.67%
21	98.85%	100.00%	99.03%	101.49%	106.82%
22	101.58%	95.55%	106.63%	94.37%	108.22%
23	112.02%	103.15%	110.61%	104.58%	117.96%
Cumulative Difference	15.22%	-2.81%	13.46%	-3.11%	41.67%

This table show O’Neill’s performance compared in exactly the same way as Escobar. O’Neill is better than the league over these four seasons in three of four categories and a LOT better in ERA. The below league performance in two categories (on-base and WHIP) are pretty small. Overall, ERA was the biggest driver of the comparative difference in the two pitchers. Otherwise, they end up pretty close, as the 5.0 to 5.5 final rating number would indicate.

That said, two ratings which appear to ONLY affect performance at the VERY highest extreme are velocity and GB/FB. O’Neill’s velocity rating went from 25 to 23 as he aged, and Escobar’s went from 81 to 78. Given the wide gulf in these ratings, IF this rating was a true indicator of results, we’d see Escobar emerging as a better pitcher given the overall closeness of the other ratings. Neither pitcher is a great GB pitcher (Escobar 56 and O’Neill 48). However, in Busch Stadium you don’t HAVE to be a GB pitcher to be effective.

Other Pitchers

Last season there was some discussion about the comparative quality of these two pitchers to pitchers on a playoff opponent.

Boomer Lange rates out in my updated system at 4.9. He has an OVR of 88. The reason he ranks out below both Escobar (only by .1) and O’Neill is pitch quality.

Allan Blackley rates out at 6.9 - a true monster! His OVR is 84. Why so much higher than Escobar, O’Neill and specifically Lange - pitch quality. The advantages Lange enjoys in control and splits is offset by relatively average pitch quality.

Alex Rosa rates out at 4.4 with an OVR at 81. Again - it’s pitch quality.

Would I take ANY of these pitchers? Yes. In a heartbeat. They are all very good. They also each pitched for the best team in the NL last season and arguably the best team over the last 3 seasons. So their rather gaudy win loss record has to be viewed in that context. Each of them throw hard and are GB types, which doesn’t HURT performance, but doesn’t provide a big lift either. But all things being equal, I’d be thrilled to have any of these three pitchers in my rotation!

By comparison, my team ran into three REALLY beastly pitchers in our NLCS win en route to a WS loss: 

Ralph Moseley - OVR 81, 6.4 in my system - monster with very good splits and excellent pitch quality.
Norberto Morales - 83 OVR and 7.3 my system - EXTREME control plus excellent pitch quality.
Glen Stokes - 81 OVR and 5.9 in my system - solid splits and outstanding pitch quality.

Conclusion

Control and splits form the basis of deciding if a pitcher will be effective. Once a pitcher passes the threshold for inclusion on an ML roster, the next critical ratings are pitch quality. I’ve seen pitchers with great pitch quality that could not pitch in the ML, and conversely I’ve had and used a few pitchers with GREAT splits, pitches and velocity but poor control that did not pan out at all. The reason O’Neill ends up delivering a bit better results is linked to his superior overall pitch quality.