Tuesday, March 12, 2013

Complete Hitter KBO WAR for 2012

In December, I broke out my formula for measuring WAR in Korea Baseball Organization (the KBO). For hitters, it didn't contain any kind of defensive metric or positional adjustment, and I didn't have a complete list of players, only looking at a few of them. Thanks to Dan of Mykbo.net, this is the day I fix this flaw. I looked at the top 9 players for each team in hits (I was planning on doing plate appearances, but My KBO's statistics sorts by hits, and I didn't notice that until I was nearly half way done) for this post).

It is called KBO WAR, but we are only actually looking at KBO WAA (wins above average instead of wins above replacement). One of the reasons for this is that, while we created a replacement baserunning method, creating a replacement defensive method just seems a little pointless. Also, when looking at KBO players, at least from an international perspective, we don't really care what a replacement KBO player looks like. While understanding what a MLB replacement player looks like can be helpful (though I have never liked that MLB players' defensive and baserunning components are compared to average while their batting component is compared to replacement), understanding what a KBO replacement player looks like just doesn't have much value if you are trying to see if any are prospects (at least statistically prospects) to move into bigger leagues such as the MLB or NPB. With that said, we may not even care about the average KBO player looks like. A little more on that later.

For positional adjustments, I used the traditional positional adjustments created by Tom Tango.There are some variates out there, such as Baseball Reference uses -10.5 runs for a first baseman instead of -12.5. I just used the -12.5 run value because that is what I have usually used, but I am always open to other arguments I guess. I tweaked the positional values slightly, giving RF a run more than LF. This is because it usually takes a better arm to play RF. I was supported in it being a better position by them having more put outs, but they actually had less assists than left field.

Calculating actual positional value will be a little difficult. I just used percentage of innings, because that makes the most sense to me. Such as, say a theoretical player plays 500 innings, 400 of them at 1st, and 100 in left field. 20 percent of -8 is -1.6, and 80 percent of -12.5 is -10, thus, the player's positional value is -11.6. I think this works because it doesn't artificially award utility, but does take it into account (a problem arises when say a player can play 2nd well, but has to play left field because 2nd is already taken. I can't think of how to get rid of such team biases).

Defensively, I used range factor and a percentage factor instead of a run factor (but will calculate it as runs). I used: Player Range Factor - League Average Range Factor for position, divided by league average range factor times 100. This obviously needs some kind of minimum basis, and that is why we are only looking at starters. It doesn't need to be said that this is a pretty terrible defensive component, especially compared to what we have to work with when it comes to MLB players (and those metrics are pretty flawed as well), but it gives us something to work with.

For catchers, I subtracted the averaged passed ball ratio by the player's passed ball ratio and multiplied it by 100 to move the decimal up 2 and add it to the difference in caught stealing average times 100. For example, if a catcher has a 32.6 CS % and .007 PB ratio, the catcher has a .9 defensive value. Yes, this weighs stolen bases higher than passed balls. This is for several reasons, one being the difference between wild pitches (data we don't have for catchers) and passed balls is somewhat subjective.

I didn't look at pitchers, because I am only doing the positional players here, but adding the defensive component to pitchers would be interesting.

As you will see in the spreadsheet, I put something called WAE, Wins Above Excellence. Because we don't even even care about average players from a international perspective, there should be some kind of cutoff for when we should start caring about them from an international perspective.
Any way we do this will probably be arbitrary (Especially since we don't have a ton data on position players that have played in both American and Korean leagues, and I usually distrust translations). A couple of years ago, using an even more crude method, it seemed that there weren't any internationally interesting Korean position players. Because I didn't know where to set the baseline (I am open to suggestions!), I just left it blank for now.

Of course, WAR, other than FIP WARs for pitchers, doesn't take in account what we might call luck, randomness, or BABIP luck. There have been several attempts to try to neutralize the BABIP out of hitter statistics, and I have tried several of my own. Here, I am using Hitter FIP, which turns into a runs created per 9. Instead of using innings as the baseline, I am using games played. You still add the 3.2. This has obviously problems, such as ignoring doubles and triples, discounting speed, but may be useful in a basic predictive sense. It is unrelated to WAA or WAE, but I am including it for fun.


Remember, most of this stuff only works on full time players, and again, that is fine, because we don't really care about part time KBO players or above replacement but below average bench KBO players.Without further words or explanations (I am sure many of you skipped the method and went straight to the results, which is fine), here is my list of the 72 (9 from all 8 2012 teams) players,sorted by WAA:
Let me know if you find any errors, etc. I have noticed that the speed score is well below 0 combined, which is a problem and skewed the WAA to a negative. I have left it, but just note that players are probably better runners than the metric is giving them credit for and they are probably worth a little more on a whole than the WAA gives them credit for. If anyone wants to fix this component, be my guest. Defensive and offensive value is positive, which it should be since these players are starters, or at least play a lot. You would expect starters to be, on average, better than league average. The positional adjustment, as a whole, is negative, which it should be. In case you were wondering, the average hFIP for the 72 players was 3.93.

This approach is somewhat crude and has plenty of flaws, but hopefully it will help us quantify and rank the talent in the KBO. You will notice that there is a 2nd sheet called "defensive guts", which is just the defensive averages for each position. If you would just like the spreadsheet, feel free to look at it or download it here.

1 comment:

  1. There are various online sources to provide you informative details on this topic, but this is one is very helpful

    CPR in Wichita

    ReplyDelete