Closer Reliability Report-2008 Season-Week 1

The first week was full of both saves and save opportunities. On the surface, most saves came from the teams' announced closers, but from the thirty teams, there were Save Opportunities (SO) from 50 different pitchers! Four announced closers (five if BJ Ryan is included) didn't even have a save opportunity.

Roughly that means that each team used one to two different pitchers in the first week alone to obtain a save. 1.67 to be exact, but 2/3 of a pitcher does not an effective fast ball make!

There were 75 SO in the first week and 51 of those were converted into saves. That is a conversion percentage of greater than two-thirds, or %68.00.

Announced closers were responsible for 47 of the SO and 43 of the saves. Other relief pitchers accounted for 28 other opportunities, but only 8 saves. So while non-closer relief pitchers had 37.33% of the SO, they only converted 28.57% of those (or 15.68% of the total saves). In contrast announced closers converted 91.49% of their SO.

So on the first week there were many opportunities for both closers and non-closers alike. However, the true closers had more opportunities and were more proficient at finishing their games. This make sense, closers' save opportunities generally come in the last inning of the game and they are of a more reliable stock then their relief teammates--hence the reason why they are indeed closers in the first place.

As the weeks go on, and more pitchers are used, more opportunities will come from multiple sources. From a fantasy sense, especially in deeper leagues, it may make more sense to bench RP that you just have on your roster for potential saves, and have more pitchers who can earn you other stats such as W, WHIP, ERA, etc.

Closer Reliability Report-2008 Season-After Opening Day

Opening day has absolutely showed me that a way is needed to project saves from those who aren't closers. The sample size is limited since it only takes into account the first 2 days of the season (or first four days taking into account the two games in the Tokyo Dome)

In those few days there were a total of 16 save opportunities and 7 total saves. Of those opportunities, 8 were from the projected closers. The projected closers however did have 6 of the seven saves.

What does this mean?, absolutely nothing since the sample size is very limited. But as the season progresses trends should start to develop. The next update will be after the first full week of the season.

Closer Reliability Report-2008 Season

This season I will be verifying the validity of a MLB Team's announcement of a particular player as their closer, especially at the beginning of the season. For past years, 20 Save Opportunities or greater is defined as a closer, for the current year, anyone who posts a save will be tallied.

Throughout the year I hope to answer certain questions...Does age or years of service as closer predict reliability?, Do certain statistics such as ERA, K/BB ratio, or dominance predict reliability of a closer?, Do certain types of pitchers (i.e. power, finesse, etc) have greater predicted reliability?, are closers affected by a change in park or league after a trade?, and also what predicts a future closer. Will also evaluate the first season closers, of the 30 projected closers, there are who are six (or 20%) being named for their first full season--so there should be a high amount of volatility.

The average number of season as a closer is 3.4, with the maximum at 13 by Trevor Hoffman.

Table 1.
Opening Day Closers

Name	Team	L	DOB	SaC
Borowski,Joe	CLE	AL	5/4/71	3
Capps,Matt	PIT	NL	9/3/83	1
Cordero,Chad	WAS	NL	3/18/82	3
Cordero,Francisco	CIN	NL	5/11/75	5
Corpas,Manny	COL	NL	12/3/82	1
Gagne,Eric	MIL	NL	1/7/76	3
Gregg,Kevin	FLA	NL	6/20/78	1
Hoffman,Trevor	SD	NL	10/13/67	13
Isringhausen,Jason	STL	NL	9/7/72	8
Jenks, Bobby	CWS	AL	3/14/81	2
Jones, Todd	DET	AL	4/24/68	8
Lidge,Brad	PHI	NL	12/23/76	4
Lyon,Brandon	ARI	NL	8/10/79	0
Nathan, Joe	MIN	AL	11/22/74	4
Papelbon,Jonathan	BOS	AL	11/23/80	2
Percival, Troy	TAM	AL	8/9/69	9
Putz,JJ	SEA	AL	2/22/77	2
Rivera, Mariano	NYY	AL	11/29/69	11
Rodriguez,Francisco	LAA	AL	1/7/82	3
Ryan, BJ	TOR	AL	12/28/75	2
Saito,Takashi	LAD	NL	2/14/70	2
Sherrill, George	BAL	AL	4/19/77	0
Soria, Joakim	KC	AL	5/18/84	1
Soriano,Rafael	ATL	NL	12/19/79	0
Street,Huston	OAK	AL	8/2/83	3
Valverde,Jose	HOU	NL	7/24/79	2
Wagner,Billy	NYM	NL	7/25/71	10
Wilson,Brian	SF	NL	3/16/82	0
Wilson,CJ	TEX	AL	11/18/80	0
Wood,Kerry	CHC	NL	6/16/77	0

Coming Next...Week 1 Results

Fantasy Baseball Abstract-Predicting HR or AVG, what is more reliable?

Question: What is more reliably predicted in fantasy baseball-Batting Average or Home Runs?

Methods: I felt that the best way to answer this question is to realize how far away a prediction was away from 100%. The percentage of the prediction would be defined as 100% and the percentage of the actual result would be the total actual result divided by the total predicted result. The difference between these would be defined as the Confidence Factor or CF. The closer a CF is to zero, the greater reliability it reaches. The reliability percentage would be defined as actual result divided by projected result. The closer the result gets to 100% the more accurate the prediction would be.

It was broken into two categories, both exceptional players and reliable players. Exceptional HR hitters are defined as projected 25 or greater; and exceptional AVG players are defined as a projected 300 average or greater and 500 ABs. Projected reliable players are defined as a projection of 500 ABs or greater.

Data: For HRs, exceptional players were chosen who had a projected HR total of 25 or greater. This was factored in to their regular season stats regardless of injury. Of the 53 players to hit 25 HRs or greater, their projected total was 1789 HRs and their actual total was 1375 HRs, for a 23.1 CF. Another way to say it that it was 76.9% reliable. 3 players however (Joe Crede, Russ Branyan, and Morgan Ensberg) had ABs less than 50% of their projected totals due to injury or decreased playing time. Factoring in these three players, the projected HRs drop to 1700, and the actual total was 1349. This brings a CF of 20.6 or 79.4% reliability.

The results for reliable HRs are as follows. Of the 161 players that had 500+ projected ABs, a total of 3432 HR were projected and 2735 were hit. This is a CF of 20.3 or a reliability rate of 79.7%. Thirteen players who had at least 500 ABs projected ended up with less then 50% due to injury of lessened playing time. These players were Rocco Baldelli, Russ Branyan, Jorge Cantu, Joe Crede, Chris Duffy, Morgan Ensberg, Andre Ethier, Shea Hillenbrand, Nick Johnson, Mark Kotsay, Juan Rivera (LAA) Ryan Shealey, and Chad Tracy. The remaining 149 players were projected to hit 3175 HRs but actually hit 2676. The CF for this group is 15.7, or a reliability rate of 84.3%

For BA, players were chosen who had a projected AB total of greater than 500 in both categories. Exceptional players were defined as having a greater than 300 average.

Exceptional BA players were projected to have an average of 311, and they actually hit 299. This has a CF of 3.8, or 96.2% reliability. None of the players projected to hit above 300 and have more than 500 ABs had greater than 50% of their ABs lost.

The reliable BA results were as follows. The projected results were 285, and the actual results were 275. That is a CF of 3.5 and a reliability rate of 96.5%. Twelve players who had 500 ABs projected reached less then half that amount, they are the thirteen listed above with the exception of Russ Branyan who did not have 500 projected. This group’s projected BA was 286, and their actual BA was 281. This is a CF of 1.7, or an impressive reliability rate of 98.3%.

Conclusion: Why such a great disparity? I believe that easily quantifiable numbers (i.e. last year’s ct%, gb/fb ratio, etc) tend to favor the prediction of Batting Averages over Home Runs. Factors that cannot be predicted, such as wind, temperature, and humidity affect how far a ball travels, and the difference of 3 to 5 feet can mean the difference of between a home run and an out. However, these minutiae of difference would not affect a non-HR hit as much. There is a greater probability (assumed not proven) that a hit ball would land in the field of play and NOT be an out than hit over the fences. Then logic would follow that difference in feet or inches would not make as much of a difference in simple average as compared to HRs.

This is a game of inches, but the science isn’t specific enough to measure those inches. Instead we use quantifiable metrics, historical and implied trends, and some dumb luck to predict the future. But the Sabremetric Statistician’s crystal ball is a bit fuzzy when it comes to factors outside the game that affect inside the game. These factors however do not affect balls in play as much as balls hit over the wall.

AND this is only looking at data over the 2007 season. Every year I will add results to this, further proving or disproving the theory that batting average is more reliably predicted than home runs.

Addendum I: While reviewing this abstract I noticed a potential flaw in my logic. Over the past few years I have shown that BA and ABs have a direct correlation. Also the reliability of these players is much greater thereby making the reliability of these predictions easier to project. Therefore I will also project the CF and reliability rate of 25+ HR and 500 ABs.

46 Players fell into both categories. They were projected to hit 1579 HRs but hit 1225. This gives a CF of 22.3 or a 77.6% reliability rate. 2 players had less than half of the projected ABs, Joe Crede and Morgan Ensberg. Taking them into account 1517 HRs were projected and 1209 were hit. This raises the CF to 20.3 and the reliability rate to 79.7%.

Therefore the theory still holds true.

Addendum II: I noticed that I used Batting Average but Home Run Totals. To avoid this possible contradiction I will use HR average for the players who fell into the rankings from Addendum I.

The average number of HRs project to be hit was 34.3, and the actual average hit was 26.6. The CF for this was 26.7 and the reliability rate was 73.3%. Again factoring out Crede and Ensberg the averages change to 34.5 projected and 27.5 hit. This lends to a CF factor of 20.3 or a reliability rate of 79.7%.

No marked difference is shown taking account average rather then total.

Debate I: Perhaps however, a CF of 20 for HRs is not as important a CF less than 4 for BA. The Batting Average category is normally won by a matter of points, and the Home Run category is one by more than a HR or two, more often won by difference of many home runs.

Table I:CF and Rel% of players with >50% projected ABs

	CF	Rel%
Exceptional HR	20.6	79.4
Reliable HR	15.7	84.3
Exception BA	3.8	96.2
Reliable HR	1.7	98.3