Author: Jeffrey Brover, Sport Management and Statistics at Rice University

Note: Statistics for KCL through July 4; Statistics for MLB through June 30

0. Introduction

At the end of every MLB season, thirty members of the Baseball Writers’ Association of America, one for each team, place a vote for their top five ranked candidates for the Cy Young Award, given annually to the season’s best pitcher in baseball. Each writer uses their own method of organizing the pitchers and sometimes it results in unanimous decisions, such as Shane Bieber in 2020, while other times, the vote of one writer can make a difference, such as in 2012 when David Price won the AL nod over Justin Verlander by 4 points.

With the KCL season winding down and postseason award deliberation already underway, it is time to start thinking about some ways the pitchers of the KCL can be ranked. While there is a wide range of ways to go about ranking pitchers, including ESPN’s Cy Young Predictor formula, one-by-one comparison of statistical categories against each other, or even the infamous eye test, to name a few, this post aims to look at there potential methods that involve three different levels of statistics: fantasy points, MLB and KCL player sim scores, and pitcher WAR(pWAR).

1. Fantasy Points

1.1. Fantasy Points Explained

Perhaps the most simple method of ranking players is by calculating their value from fantasy points. ESPN Standard Fantasy Points for Head-to-Head leagues are scored as follows:

  • Win(W): +5
  • Save(SV): +5
  • Out Pitched(IP/3): +1
  • Strikeout(K): +1
  • Hit Allowed(H): -1
  • Walk Issued(BB): -1
  • Earned Runs(ER): -2
  • Losses(L): -5
  • Save(SV): +5

In other words, fantasy points use eight of the most common and easy-to-measure pitching statistics, assign a weight to each of them, and then calculate the amount of value they provide to a hypothetical fantasy team.

Through GameChanger, each player who has pitched at least one game for their team can be calculated. One slight exception to this is Isaiha Immke, who has pitched for two different teams. As such, his stats have been merged to represent one individual.

1.2. Fantasy Points Overview

Boxplot 1.2

This initial boxplot gives a sense of scale for the magnitude of fantasy points pitchers have earned in the KCL through six weeks. While the median value of points is 11, the 75th percentile is 25 (only 25% of pitchers have a score above that mark), with one pitcher currently in the outlier range, Austin Baker (Ground Sloths) at 63 (21.1 IP, 4-1, 23 H, 8 ER, 9 BB, 32 K). Tied at second is Griffin McCluskey of the BlueCaps (20.0 IP, 1-1, 17 H, 6 ER, 5 BB, 29 K) and Will Jackson of the Bobcats 55 (26.2 IP, 2-1, 23 H, 13 ER, 12 BB, 31 K), followed closely behind by Keegan Gagliardo (Sloths) at 53 (15.2 IP, 2-1, 15 H, 3 ER, 3 BB, 26 K). Baker’s best start of the season by fantasy points came on June 26, going 4 IP with 3 H, 2 ER, 2 BB, 10 K, and the win, good for 24 fantasy points, which is almost the 75th percentile for the KCL totals.

1.3 Split by Starter/Reliever

One potential flaw with Boxplot 1.2 is that it treats starting pitchers and relief pitchers the same. Relief pitchers who tend to throw fewer innings have fewer average fantasy points, as outs pitched and wins tend to be a large bulk of starting pitcher points. Meanwhile, relievers, who usually do not go more than one inning, tend to rely on strikeouts, limiting baserunners and saves for their points.

Boxplot 1.3

Boxplot 1.3 creates two separate boxplots, one for pitchers who are primarily starters in the KCL, and one for those who are primarily relievers. This distinction is based on whether the pitcher has started more games or relieved more games, with ties going to which they have done most recently.

When split up by position, Austin Baker is no longer considered an outlier, as the relief pitchers, who have a max score of 46 are no longer present in the starting pitcher graph. The median fantasy value of a starting pitcher is around 25, while for relief pitchers it is around 12.5, with many more (approx. 25% earning a negative value for a hypothetical fantasy team. Logan Wombles (14.2 IP, 1 SV, 11 H, 1 ER, 7 BB, 17 K) leads all relievers in fantasy points with 46, which is sixth most overall in the league and 13 more than the nearest reliever. Logan Wombles has been the Sloths’ go-to option whenever they need some clutch close-game relief work which has resulted in his team going 6-2 when he takes the mound.

1.4 Average Fantasy Points

If pitching well, more starts are likely to lead to more points, but for a variety of reasons (injuries, high school playoffs, travel, etc.), starting rotations are not nearly as fixed as in an MLB environment. This means that pitchers putting out stellar numbers may not be near the top in this case, simply as a result of not seeing enough action.

Boxplot 1.4

Therefore, a second potential way to view the fantasy points data is to view it by average fantasy points, i.e. fantasy points per game played. When displayed this way two players emerge as potential highlights. First, Jake Swartz(5 IP, 3 H, 0 ER, 7 K), an outlier, who truly is an outlier, as he “averages” over 20 points, but has only played in one game on June 9. A better option from this graph is Mitchell Sampson (12 IP, 5 H, 5 ER, 24 K, 5 BB), who despite only playing 3 games has accumulated 40 points. Mitchell Sampson joined the KCL in Week 4 and has pitched three of the past four weeks following Heartland CC’s playoff exit. Although Sampson joined the KCL later than most, it is not as much about the start as the finish. With a couple more potential starts remaining, he has a chance to continue to skyrocket more into the conversation.

1.5 Split by Team

Boxplot 1.5

One final way to view the data is to split it up by team, seeing which players emerged with the highest total this way, compared to the rest of the team. McCluskey and Baker still appear here as their team leaders, with an interesting dichotomy in their team boxplots. Although only separated by eight total points, McCluskey is an outlier for the BlueCaps with a fantasy points score nearly 30 points higher than his next closest teammate, while the Sloths’ 75th percentile is higher than the BlueCaps’ second-highest pitcher. The dominance of the Sloths’ pitching staff, both to start and relieve games, is made even more clear in the pitching WAR section. Two pitchers who have yet to be discussed for fantasy points before this graph are the two leading the Merchants as outliers: Cade Sharp (26 IP, 1-2, 19 H, 7 ER, 16 BB, 22 K) and Nathan Garard (25.2 IP, 1-1, 30 H, 15 ER, 2 BB, 31 K). The Merchants as a whole have been consistently near the middle of the pact leaning towards the bottom for pitching, as evident, by their boxplot having the shortest whiskers of any club, as well as the lowest median score. Despite this, Garard and Sharp still impress with stellar noteworthy statistics, such as Sharp’s league-leading 1.89 ERA or Garard’s 31-2 K/BB ratio.

1.6 Candidates via Fantasy Points Summary

  • BlueCaps Griffin McCluskey (2nd Most Total, Leads BlueCaps)
  • Bobcats Mitchell Sampson (Highest Average, Min. 2 Games)
  • Bobcats Will Jackson (Leads Bobcats)
  • Ground Sloths Austin Baker (Most Total)
  • Ground Sloths Keegan Gagliardo (4th Most Total, 2nd Highest Average)
  • Ground Sloths Logan Wombles (Most Among Relievers)
  • Merchants Cade Sharp (Leads Merchants-Tie)
  • Merchants Nathan Garard (Leads Merchants-Tie)

2. Similarity Scores

2.1 Similarity Scores Explained

In Section 1, a baseline was established for what eight of the best pitchers in the KCL by statistics look like. With that in mind, these top players can be compared to each player to see which player has the most similar aggregate statistics to them, through the use of pitcher similarity scores.

The father of Sabermetrics, Bill James, originally invented the concept of similarity scores to compare across time which players each player were most similar to. The maximum value a player can have for a similarity score is 1000, meaning each player would therefore have a similarity score of 1000 with themselves. Similarity scores can be used to see the best players in the KCL by comparing which players have the highest associated sim scores with the projected best pitchers in the league, but also by scaling their stats and comparing them against MLB-level data for the best pitchers in the MLB.

Simscore is calculated by starting at 1000 and subtracting as follows:

  • One point for each difference of 1 win.
  • One point for each difference of 2 losses.
  • One point for each difference of .002 in winning percentage (max 100 points).
  • One point for each difference of .02 in ERA (max 100 points).
  • One point for each difference of 10 games pitched.
  • One point for each difference of 20 starts.
  • One point for each difference of 20 complete games. (1 in KCL so far, omitted)
  • One point for each difference of 50 innings pitched.
  • One point for each difference of 50 hits allowed.
  • One point for each difference of 30 strikeouts.
  • One point for each difference of 10 walks.
  • One point for each difference of 5 shutouts. (1 in KCL so far, omitted)
  • One point for each difference of 3 saves.

Additionally

  • If they throw with a different hand and are starters subtract 10, relievers 25.
  • For relievers halves the winning percentage penalty.
  • Although each individual statistic is typically rounded to a whole number for a whole number similarity score, due to the lack of innings pitched overall, they are kept as decimal values.

2.2 KCL-to-KCL Similarity Scores

Although intended to be looked at throughout an entire player’s career, they are still reasonably accurate for season-by-season data. Before moving on to MLB data, the top eight players from the previous section can be utilized to see which players are most similar to them and therefore deserve attention.

Tables 2.2.1-2.2.3
Tables 2.2.4-2.2.6

From these eight tables, first, notice that Griffin McCluskey’s most similar pitcher is Mitchell Sampson and vice versa as two of the top pitchers in the league in the two top left tables. Next, in the two bottom right tables, notice that Nathan Garard and Cade Sharp are also most similar to these two pitches, sharing similar statistical lines across the board. Meanwhile, the two bottoms left and two top right tables add three new potential candidates to the mix.

Firstly, both Keegan Gagliardo and Will Jackson’s most similar player is Nate Righi of the Ground Sloths with 16.0 IP, 2-1, 13 H, 5 ER, 8 BB, and 28 K, including a KCL-season high 13 strikeout performance. Although only two starts and four appearances total hampered him from appearing in fantasy points, he shows up here behind his impressive ERA, and ability to limit contact. Meanwhile, the top reliever of fantasy points is unsurprisingly most similar to another top reliever, leading to a second reliever option in Logan Antrim of the Bobcats with 10.1 IP, 0-1, 3 ER, 6 BB, 12 K. Finally, Austin Baker’s most similar plater is Wes Hunt of the Merchants (16 IP, 2-1, 13 H, 10 ER, 11 BB, 11 K), albeit very far enough to compared to some of the other comparisons.

2.3 MLB-to-KCL Similarity Scores

Similarity scores do not only have to be limited to KCL players. KCL and MLB pitcher datasets can be merged to compare 2022 MLB pitchers to KCL pitchers and vice versa. To make it more accurate, however, it does require scaling the KCL pitcher counting statistics (W, L, K, etc.), while leaving rate statistics (ERA, Wpct, etc.) the same, by multiplying counting statistics by the ratio of average games pitched in the MLB divided by average games pitched in the KCL.

The Neyer/James Guide To Pitchers – co-authored by Bill James and ESPN.com’s Rob Neyer presents a method, based on past results, to predict Cy Young balloting, called the Cy Young Predictor (CYP) (https://www.espn.com/mlb/features/cyyoung). From it, one can use similarity scores to see which KCL players are most similar to the top three candidates in each league through June.

Tables 2.3.1-2.3.3
Tables 2.3.4-2.3.6

This selection continues to showcase some frontrunners of the KCL(Justin Verlander/Tyler Anderson -> Austin Baker and Shane McClanahan -> Keegan Gagliardo), but once again, three new names emerge for their similar statistics to MLB players. The player comparison of the best reliever in the Americal League by CYP, Clay Holmes of the Yankees, is Thomas Harper (7.2 IP, 1 SV, 1 ER, 2 BB, 8 K). Secondly, The player comparison of the current NL front runner Tony Gonsolin of the Dodgers is Ryan Borberg (8.2 IP, 1-0, 4 ER, 7 BB, 8 K). Finally, the current NL runner-up, Sandy Alcantara of the Marlins, matches up with Mason Telford (9.0 IP, 2-1, 9 H, 4 ER, 5 BB, 8 K). While none of these three pitchers have over ten innings, which will likely be a major concern for Cy Young voters, if they can accumulate more innings in the late stretches of the season, while continuing to match the rates they put up, they might be able to continue putting themselves in the conversation for the Cy. At the very least, they should be placed into consideration for reliever of the year.

2.4 Additional Candidates via Similarity Scores

  • BlueCaps Mason Telford (Sandy Alcantara Similarity Score)
  • BlueCaps Ryan Borberg (Tony Gonsolin)
  • Bobcats Thomas Harper (Clay Holmes)
  • Bobcats Logan Antrim (Logan Wombles)
  • Sloths Nate Righi (Keegan Gagliardo)
  • Merchants Wes Hunt (Austin Baker)

3. Pitcher War(pWAR)

3.1 pWAR Explained

Finally, an advanced statistic that can be calculated from the KCL to compare pitchers (or at the very least closely estimated) is their WAR.

For those interested in the calculation read on. Otherwise, feel free to skip to section 3.2 for the results

According to FanGraphs, power can be calculated as follows:

pWAR = [[([(“League FIP” – “FIP”) / Pitcher_Specific_Runs_Per_Win] + Replacement_Level) * (IP/9)] * Leverage_Multiplier_for_Relievers] + League_Correction

With the data available, pWAR cannot be calculated exactly, but a relatively close estimate of the number can be calculated.

Formulas for each individual calculation of pWAR are below:

  • Lg_FIP = (((13HR)+(3(BB+HBP))-(2(K+IFFB)))/IP + lgERA – (((13lgHR)+(3*(lgBB+lgHBP)) – (2*(lgK+lgIFFB)))/lgIP) + lgRA9 – lgERA/(PF/100)
    • HR = Home Runs
    • BB = Walks
    • HBP = Hit-by-pitch
    • K = Strikeouts
    • IFFB = Infield Fly Balls (Counted as strikeouts in the formula, but not including in KCL calculation)
    • IP = Innings Pitched
    • lgERA = League Earned Runs Average
    • lgHR = League Home Run Total
    • lgBB = League Walks Total
    • lgHBP = League Hit-by-Pitch Total
    • lgK = League Strikeout Total
    • lgIFFB = League Infield Fly Balls (Counted as strikeouts in the formula, but not including in KCL calculation)
    • lgIP = League Inning Pitch Total
    • lgRA7 = (League Runs)/(League Innings Pitched) x 7
    • PF = Park Factor (Equal to 100 for all KCL situations, since all games are played in same stadium)
    • Note: For lgFIP for statistics that do not start with lg, use league stats for lgFIP, while for FIP, use individual player statistics
  • Pitcher_Specific_Runs_Per_Win = (lgFIPR7 – FIPR7) / ([([(14 – IP/G)(lgFIPR7)] + [(IP/G)FIPR7]) / 14] + 2)*1.5
    • lgFIPR7 = Fielding Independent Pitching adjusted for park and scaled to RA7 for the league (calculated above)
    • FIPR7 = Fielding Independent Pitching adjusted for park and scaled to RA7 for the player (calculated above)
    • G = Games played
  • Replacement _Level = 0.03(1 – GS/G) + 0.12(GS/G)
    • GS = Games Started
  • Leverage_Multipliers_for_Relievers = (1 + gmLI) / 2
    • gmLI = a pitcher’s average leverage when he enters the game, unable to be calculated in the KCL without Win Probability Added(WPA), so the stat is assumed to be equal to 1 for all, making it equal to 1 overall. This should not have a large overall impact on pWAR
  • League_Correction = WARIP * IP
    • WARIP = Uniform adjustment per IP for WAR, unable to be calculated and omitted

3.2 KCL pWAR Leaders

Scatterplot 3.2

Of the top ten performers in pitcher WAR, eight of them have already been mentioned, but two yet-to-be-named pitchers are Keegan Buksa at .447, and Ryan Behling at .847. Four of the top six pitchers by pWAR in their league are on the Ground Sloths, including Ryan Behling, with the only exceptions being Griffin McCluskey and Nathan Garard. Austin Baker and Keegan Gagliardo are also the only two to have a WAR above 1. The Sloths has been the dominant team when it comes to starting pitcher this season and it is hard to pick just one out of the bunch, with each having strong cases.

3.3 KCL Over/Undervalued pWAR

Scatterplot 3.3

Without the leverage index being included, pitcher WAR for relievers tends to be drastically undervalued, while those that have gone very long into games but maybe have not pitched the best, tend to be overvalued. Three relief pitchers who deserve recognition for their performances based on expected pitcher WAR are Logan Wombles and two Bobcats’ relievers, Jack Bach and Ryne Willard, who has 5 saves and is the only pitcher with more than 1. Both have done stellar work in the Bobcats pen and have helped them maintain leads when they have them.

3.4 Fantasy Points vs. pWAR

In this day of modern baseball, it feels like sabermetrics and advanced statistics are being given more and more weight and attention. However, it is important to sometimes take a step back and realize that simple statistics can also sometimes do the trick. For instance, as seen in the chart below fantasy points and pWAR have an 83% correlation between them with a strong, positive relationship between the two that is either linear or even slightly quadratic. While the two statistics provide different information with drastically different formulas, the statistics are still highly correlated with one another.

Scatterplot 3.4

3.5 Additional Candidates via pWAR Summary

  • BlueCaps Keegan Buksa (.447 pWAR)
  • Bobcats Jack Bach (Undervalued pWAR)
  • Bobcats Ryne Willard (Undervalued pWAR)
  • Sloths Ryan Behling (.847 pWAR)

4. Conclusion

There is no one correct way to rank pitchers in the KCL or any league for that matter. All of these differing perspectives and statistics, in addition to the intangibles and “stuff”, should be taken into account to determine what makes one pitcher better than the rest. With all the statistics laid out on the line, here are the power rankings for the top ten KCL Cy Young candidates.

  1. Austin Baker (Sloths) – 21.1 IP, 5 GS, 4-1, 23 H, 9 BB, 32 K, 2.63 ERA, 1.50 WHIP
  2. Griffin McCluskey (BlueCaps) – 20 IP, 5 GS, 1-1, 17 H, 5 BB, 29 K, 2.10 ERA, 1.10 WHIP
  3. Keegan Gagliardo (Sloths) – 15.2 IP, 4 GS, 2-1, 15 H, 3 BB, 26 K, 1.34 ERA, 1.15 WHIP
  4. Will Jackson (Bobcats) – 26.2 IP, 6 GS, 2-1, 23 H, 12 BB, 31 K, 3.41 ERA, 1.31 WHIP
  5. Logan Wombles (Sloths) – 14.2 IP, 8 GP, 1 SV, 11 H, 7 BB, 17 K, 0.48 ERA, 1.23 WHIP
  6. Nathan Garard (Merchants) – 25.2 IP, 1-1, 30 H, 15 ER, 2 BB, 31 K, 4.09 ERA, 1.25 WHIP
  7. Nate Righi (Sloths) – 16.0 IP, 4 GP, 2 GS, 2-1, 13 H, 8 BB, 28 K, 2.19 ERA, 1.31 WHIP
  8. Ryne Willard (Bobcats) – 13 IP, 11 GP, 0-1, 5 SV, 12 H, 8 BB, 12 K, 4.85 ERA, 1.54 WHIP
  9. Cade Sharp (Merchants) – 26 IP, 1-2, 19 H, 16 BB, 22 K, 1.89 ERA, 1.35 WHIP
  10. Mitchell Sampson (Bobcats) – 12 IP, 3 GS, 1-1, 5 H, 5 BB, 24 K, 2.92 ERA, 0.83 WHIP

5. Acknowledgements/Sources

Posted in Kernelytics Blog

Leave a Comment