Monday, March 29, 2010

Rox Talk - Pitching Rotations Part II

Pitching Rotation - Part II

In Part I, I tried to determine if the number one starter had somewhat of a disadvantage because the possibility that he consistently faced a majority of number one starters. As the data shows Jimenez faced an average of 3.3, Marquis a 3.5, de la Rosa a 3.0, Hammel a 2.7, and Cook a 3.7. The data would suggest that over a long season the matchups rarely remain and that there are long stretches where our number one gets to face some fours and fives. The second question therefore was whether it might make sense to have your number one start somewhere else in the order so that perhaps your number one might face more advantageous matchups?

Again looking at WAR data, I calculated the result of each matchup. The data showed that when a Rox starting pitcher faced a pitcher with a lower WAR, the Rox went 58 - 32 and had 36 no decisions. When the Rox starting pitcher was below the opposing pitcher's WAR, then the team went 10 - 17, with 7 no decisions. In the rare case when the WAR's were equal, the Rox were 1 - 1. One thing to note, last year's pitching staff was light's out. By far the greatest staff in Rox history the data I think is a bit unusual but the Rox staff did win 58% of the time overall and when the Rox WAR was greater than their opponents this percentage increased to 64%. So if the Rox starter's won 69 times then if their was someway to optimize your starters so that they faced a pitcher with a less WAR then potentially the starter's might have won 74 times or about 5 wins. Five wins is a nice number and a big difference between 92 wins and 97 wins!

As mentioned, last year was a bit unusual because the Rox starting staff was incredibly dominant. In fact, there were only 11 times when the Rox faced a pitcher with a WAR greater than Jimenez's 5.7. This included facing Wainwright twice, Haren five times, Vasquez twice, and Lincecum three times. Jimenez faced these pitchers six times and finished 3 - 3. In the other five starts the team was 1 - 3 - 1. So what this tells me is that maybe you do want your number one starting against the other team's number one and thus neutralize the other team's big starter. Yeah you might get an additional five wins if your number one didn't face the big dogs but then your other starter's would probably lose 3 - 4 of those games.

Bottom line...just throw your pitcher's into the mix. The randomness of the season would suggest that when all is said and done a good pitcher will help your team to win more often than not but offense helps and a little luck (whatever might that be!)

Friday, March 26, 2010

STATS 1993 Baseball Scoreboard

These books were true gems for me. I discovered them in 1998 and they really made me understand and think about the hidden game within the game. What is nice about these books is that they are not too heavy on over the top stats. They are kind of middle of the road. Enough to keep a stathead happy but also able to speak to the average baseball fan. It was a sad day when STATS stopped publication in the early 2000s. But one day I realized that I didn't have the older ones so I shopped around and found these earlier versions. The 1993 addition is pre-Rox as it details 1992. Some good things I found:

1) Great Foreword by the voice of baseball, Vin Scully. Really has a nice talk on what stats can do for a broadcast.

2) When a team scored after loading the bases they went 897 - 388 in 1992.

3) Dante Bichette's Edmonton (Angels AAA club) 1987 and 1988 stats

1987 360AB - 108H - 13HR - 50RBI
1988 509AB - 136H - 14HR - 81RBI

4) 1992 Reds had dual closers in Charlton and Dibble

5) Some precog on the effects of altitude

Brewer's AAA Team 1988 - 1992 (@Mile High)
@ Denver .289AVG .364OBP .420SLG
@ Milwaukee .264AVG .323OBP .363SLG

6) The big attendance draws for 1992 included Atlanta, Dodgers, Toronto, Cubs, and Boston (Yanks sucked back then...)

7) Temperature and Scoring during 1992

Temp G R/G HR/G
> 80 deg 407 8.49 1.61
70 - 79 1082 8.20 1.42
60 - 69 398 8.28 1.46
< class="Apple-tab-span" style="white-space:pre"> 219 7.84 1.23

8) Average Game Times in 1992

Pitches/G Ave Time Sec/Pitch
AL 275.6 2:53 37.65
NL 264.5 2:45 37.43

9) Bill James "The Favorite Toy" was used to project players chance of reaching 500 homeruns. Guess what Bonds chance was at 28 years of age with 176 homeruns? Only 14%...

10) Rob Dibble had 110 K's in 70.1 innings in 1992...after the All Star break he had 63 K's in 34 innings...that's 16.7 per 9 innings!

Monday, March 22, 2010

Rox Talk - Pitching Rotations Part I

Pitching Rotation - Part I

"Good pitching will always stop good hitting and vice-versa" - Casey Stengel

Is being named the number one starter a curse? Initially, on a gut level, my view is that for the Rox whenever our new number one starts the season he limps out of it with an injury or has a forgettable season. 2008 was Francis's year coming off a great 2007. Shoulder problems ended 2008 for him and 2009 was lost to surgery. Cook was the 2008 pitcher but upon being named number one for 2009 struggled mightily throughout the season and lost the tag early on and limped through Aug/Sept with an injury. So with great trepidation I realize that 2009 is Jimenez's year and he has been awarded that number one tag for 2010. So does being the number one starter cause more stress because the pitcher has to go up against the other teams number one starters?

The underlying question then is what does it mean to be a number one starter? Ideally the number one gets to open the season and be in the pole position throughout the year. Teams hope to ride that best arm to lots of victories. I took last year's schedule and ran through a 5 man rotation. The only condition I had was that the number one starter had to have a minimum of 4 days rest and always pitched on that 5th day. The rest of the rotation fell into the remaining spots. Without too much work I established a schedule where the other starters only had to pitch on 3 days rest four separate times. The result of this was that the number one starter got 36 starts followed by 35 for the number two, 35 for the number three, 32 for the number four, and 24 for the number five. For comparison sake the actual totals for 2009 amounted to 33 for Jimenez, 33 for Marquis, 32 for de la Rosa, 30 for Hammel, 27 for Cook, and 7 by others. So baseball's prevailing strategy then is to get the most out of your best arm and give your best arms a chance for more appearances and thus hopefully win more games?

Sounds like sound strategy but wait (!), this is what every major league team is trying to do so in essence the Rox number one starter is getting saddled with facing all the other number one starters. So two questions spring to mind, 1) Does this pitching rotation play out for an entire season (i.e. number ones face off more) and 2) If number ones continuously pitch against one another would it make sense to not throw your number one starter but have him start somewhere else?

First thing I noted out of the gate while studying this was that starting rotations are rarely fixed. Injuries, off days, and travel schedules allow for a pitching rotation in constant flux. Also while it is easy to establish a rotation in Spring Training, rarely does it end up the same way at the end. Then the next big question is what defines a number one vs number two? Is it wins, innings pitched, or something more exotic like WAR? Of course to analyze this you have to look back at an already finished year which can taint the results because by establishing your opponents rotation based on the end of year numbers you are sort of compromising your analysis if in fact rotation orders determines success. For instance de la Rosa won 16 games in the number three spot last year. So he either faced a lot of other number three starters and was better and thus won more, or perhaps if he was bumped up to a number two starter he might not have won as many games or (and a big or) pitching rotation means nothing and that he just happened to have luck go his way with 16 games he won?

So to answer the first question I simply listed every pitcher the Rox faced and using WAR to established who for the particular team was number one, two, and so forth. I then matched this with the Rox pitcher and took the average over the season. Therefore if an average team had 5 starters who each pitched 32 games (and two more) then the average starting pitching value would be approximately 3.00. For the extreme example above (i.e. number one gets 36 starts) a value of 2.83 is the average starting pitching value. By averaging what your starter faced over a season you can determine if they faced more top of the rotation (i.e. number below 3.00) or faced the back end (i.e number higher than 3.00).

For instance Jimenez last year faced:
1, 1, 5, 5, 5, 5, 5, 2, 3, 1, 5, 3, 2, 2, 5, 3, 1, 3, 5, 3, 5, 4, 5, 5, 1, 1, 5, 4, 3, 1, 1, 3 (average 3.3)

As I said I went with WAR values to establish starter's rank. Wins can be deceiving and innings pitch can be skewed with late season trades or call ups. If a starter was not in the top 5 based on WAR I still used a 5 for that pitcher. Rarely do teams have a defined fifth starter. The spreadsheet below shows the results. Note: In the first week of the year I gave Webb a number one rating as he was slotted to be that had he not gotten hurt...the same for Myers with Philadelphia...he was the opening day starter. The rest of the matchups held true.

So while the theoretical best gives an average of 2.83, the Rox in 2009 was 2.99, and the opposing teams threw an average of 3.25. As the data shows Jimenez faced an average of 3.3, Marquis a 3.5, de la Rosa a 3.0, Hammel a 2.7, and Cook a 3.7. The data would suggest that over a long season the matchups rarely remain and that there are long stretches where our number one gets to face some fours and fives. It is interesting to note that the first three months indicated that our number three starter, de la Rosa, was getting the bulk of hard matchups having an average of 2.1 through June (perhaps that is why he started so slow?). Next week I will ponder the second question on whether it makes sense to not send your number one out on Game 1 and establish the pecking order differently.

Monday, March 15, 2010

Rox Talk - Error of My Ways

Newsworthy Notes....

Helton is locked up until 2013. That will take him to his 39th B-Day and Year 17 of his career. He currently has 2134 hits which came in 7761 plate appearances. Assuming he gets a hit every 3.64 plate appearances, it will take him another 3150 more plate appearances to get the 866 hits he needs. He has played in 1812 games so it will probably take about another 735 games to get there or about 5 more years. My guess is that if he gets close he will play a year or two in the AL as a DH to get the magic number.

In speculation who would win if the 86 Mets played 2009 Yankees...guess only a NYer would care...according to them it would have been the 86 Mets. What if is a neat little site. I ran a simulation of the 2007 Rox vs the 2009 Rox. 2009 Rox won the seven game series 4 - 1. Games really weren't close...2009 version scored 4, 18, 12, and 15 in their wins.

And finally the Sunday New York Times had a nice article about our know all the media attention is starting to scare me...every time we get some love we usually turn around and lay an egg!

Is it a Bad Day?

The 2009 version of the Rox committed 87 errors. I have often wondered if an error in the field translates off the field and into the Batter Box (vice versa does a bad day at the plate lead to more errors?). Looking at the 2009 box score (courtesy of the scores show that 23 players committed at least one error. 18 of the 87 errors were committed by pitchers and Barmes led the team with 13.

One problem that immediately jumps out at me is the total number of bats, on days errors were committed, only amounted to 226 bats by players committing errors. These same 23 players had a total of 4,895 bats during the entire season so basically the error at bats only account for 4.5% of the total at bats. So the data to me is entirely suspect because I could probably just randomly pick 226 at bats and probably get similar results. Furthermore, I then took the 9 pitchers out because they only accounted for 21 at bats and their pitching (a study for another day...) would probably be more affected then their at bats.

So instead of 226 total at bats, we reduced this to 205. Of these 14 position players, they had a total of 4,750 at bats in 2009 (so again this sample size is only 4.3% of the total). Data suggests that on error days the Rox batted 0.215 compared to 0.267 during the season. Slugging percentage was 0.410 on error days compared to 0.462 during the season and strikeout percentage (K divided by at bats) rose to 0.293 on error days compared to 0.230 during the season. Any of these results statistically relevant? Maybe...converting percentage points (percentage difference over total at bats) to hits, it would appear that on error days the Rox hit 249 less hits over an entire season.

On the bright side both Atkins and Tulo seem to make up for their errors at the plate. Both of them have higher averages on error days. Each had 9 errors (27 and 29 at bats, respectively) so again just a small sample size. No conclusions at this point but interesting none the less...will have to track 2010.

Monday, March 8, 2010

Rox Talk - Total Runs Scored vs Time of Game

Random Thoughts for the Week

Dexter Fowler is now 6' 5", guess he has grown 2 more inches since joining the organization... what happens if he keeps growing? That is one big center fielder!

Obvious Research Part II

Sometimes questions pop into my head and I write them down and when I have the time I try to answer the question. This one was pretty obvious but until you graph it ( ;-) ) you never know. So the question was to graph time of a game in minutes versus total runs scored by both teams. The obvious answer was as game time increased more runs should score or the more runs scored should lead to a longer game? My first attempt was the graph below (data from 2005 - 2009).

A monster blob of data. Generally I see an upward trend but the outliers were interesting. I mean the longer games (4 hours plus) generally aren't that high scoring...after that long of a game I think it becomes a game of wills. Who blinks first (22 inning affair in San Diego anyone?). To expand on the data, I took weighted averages for each total and got the graph below. Size of the point indicates the total number of games in those 5 years.

This seems about right. The average game time for about 800 games over this span was 174 minutes (sd, 26 minutes). The average runs scored was 10 (sd, 5). So a run about every 18 minutes! After pondering the data a bit more I wondered what about home/away splits. I mean games at Coors take forever right?

The purple data points represent home games and the grey points are away games. Interesting to note the overall average time for a Coors game is 174 minutes (sd, 26 minutes) while an away game is 173 minutes (sd, 26 minutes) but the average runs scored at Coors is 11 while an away total is 9 runs. It's also interesting to see on the away trend you have an uptick at about 3 runs (the 22 inning game was in San Diego, Rox won 2 - 1) then a decline until about 11 total runs. My guess here what is happening is that in a 1 - 0, 1 - 1, 2 - 1, game the managers do a lot of pitching changes or perhaps a lot more tie games are low scoring affairs.

Either way nice to know...expect a game to take about 3 hours (nice in the Summer not so much in April!)

Monday, March 1, 2010

Rox Talk - 2010 Projections

Reaching into the past to glean something for the future...In 2009 with a preseason projected line up of Spilly, Tulo, Helton, Hawpe, Atkins, Iannetta, Smith, and Barmes, it indicated that this line up could score 851 runs (they scored 804 runs) with the additions of Fowler and CarGo. In 2008 with a preseason projected line up of Taveras, Tulo, Helton, Holliday, Atkins, Hawpe, Nix, and Torrealba it indicated that this line up could score 826 runs (they scored 747 runs). So what about 2010? Using 2010 projections on Fangraphs and the handy dandy runs estimator you get the about 845 runs scored with the line up below.

I think the above line up has to be the most secure of any line up coming into a season. I mean 2008 Rox had no clue who was going to be at second. You also figure in 2010 you will have Smith, Olivo, Spilly, Mora, and Giambi taking some bats throughout the season. Know if you assume the line up above is a playoff contender with 845 runs scored and you assume that a runs score to runs allowed playoff ratio is about 1.16 then Rox's pitchers need to only allow 728 runs to score to keep up their end of the bargain. Also a 845 to 728 spread should give Rox a 92 win season! I have seen a few early season projects ranging from 83 to 86 to 87 to 92. For those bettor's out there if Vegas stays on that 83 line I would take the over big time!
No question expectations are high this year. Our Rox aren't going to sneak into the playoffs this time around. Because of this most of the early projections do have the team winning more than 85 games. Based on projections with Bill James, CHONE, and Marcels on Fangraphs, if you plot runs scored versus runs against you have 799/672, 813/741, and 823/726 respectively. Based on this you get 94, 88, and 90 wins. So I think a Wild Card is definitely in play but what about the division? Things to think about: Will Dodgerland implode because of the owner's divorce and Manny's future, can the Giants score runs and will their pitcher stay healthy, and can the D-Backs find their winning ways with a revamped offense and Webb's new arm? Sorry Padres fans but don't see you guys fighting for the division especially when Gonzalez is traded to the BoSox this summer.
And what is a perfect summer at Coors: For the offense it would be CarGo continues his playoff offensive prowess, Fowler stays healthy and runs, Tulo hits in April, Helton's back holds on, Stewart can hit lefties, Hawpe hits for an entire season, Barmes becomes sneaky good in the 8th spot, and the catcher's find some consistency at the plate. For the pitchers it would be Francis finds his old stride, Cook stays healthy, U-ball becomes a true ace, de la Rosa pitches for an entire year, and Hammel improves upon last year. For the relievers it would be Betancourt and Street...can the dynamic duo close out late inning rallies on a consistent basis?
I give the Rox 90 wins...