Quote:
Originally Posted by TomVeal
Thinking about this issue further, it occurs to me that the suggestion of a WPA-based system has merit, since it would measure performance in a way that isn't tracked currently. The game already recalculates win probability after every play, so perhaps the programming wouldn't be too daunting.
|
I raised the issue of the Player Of The Game logic in my introductory post on these forums, so I'm glad to see others noting some of the existing shortcomings. (Disclaimer: the rest of this is mostly rehashing my earlier proposal; also, tldr
)
Based on earlier threads, it has become more clear to me that the issue with PotG selections may not be as evident to people who sim their games and review the boxscores afterwards; it's more clear when you actually play out the games and watch them unfold. Since the current PotG algorithm ignores game context (more on that below), it's often difficult to reconcile the selections that appear.
The current OOTP PotG is based on Game Score -- for starting pitchers, it uses Pitching Game Score (using the established methodology for calculating game score; somebody has already linked to it in this thread), which is entirely based on their pitching line, and for batters, it uses "Batting Game Score", which is based on their batting line. (I'm guessing that this is some proprietary formula -- have never seen Batting Game Score methodology published elsewhere -- and don't exactly know how it's calculated, although with enough data points we can probably reverse-engineer it.) And TBH I frankly have no idea how it works with relief pitchers.
Anyway, with the starting pitchers and the batters, it's pretty straightforward -- highest Game Score gets the PotG.
There are a few fundamental issues with this.
1. This approach ignores League Run Environment.
If you're playing in a deadball era, starters are going to regularly throw up superficially impressive pitching lines. As such, Pitching Game Scores are going to regularly be very high (and high Batting Game Scores are going to be infrequent), and PotGs are predominantly going to be a starter.
Conversely, if you're playing in a slugger's paradise, where a 5 AB / 2 R / 3 H / 4 RBI / 1 HR / 1 BB batting line is fairly commonplace, the opposite is going to be true -- you're going to have lots of high Batting Game Scores, few high Pitching Game Scores, and the PotG is going to usually be a batter.
(My current dynasty has a ~9R/G environment, and it seems to be slightly tilted towards starters as PotG)
The problem with ignoring League Run Environment is that impressive individual performances relative to their environment are not recognized. For example, if most games in a league wind up being 2-1 affairs, throwing up a 7 IP / 5 H / 2 ER / 3 BB / 6 K pitching line is no big deal, relatively speaking. But a hitter who goes 2 for 5 with 2 doubles and 2 RBIs has done something noteworthy in this environment, but he's not going to be PotG.
2. This approach ignores game context.
As Game Score is based solely on the pitching line or batting line, it doesn't factor in game context at all. That is, someone who goes 3 for 5 with 2 HR and 4 RBI has the same Game score, whether he's the decisive difference-maker in a 4-3 squeaker, whether this all comes as part of a pile-on in a 17-2 drubbing, or if this occurs from the losing side -- a vailiant effort in a 13-7 loss. Game Score also makes no distinction between a 2-run homer that gives a team a 2-0 lead in the top of the 1st, widens a 6-run lead in the 7th, or walks his team off with a come-from-behind win in the bottom of the 9th. (Or whether it comes when his team is 9 runs down, etc.)
I think this is where is where PotG selections go awry most frequently. It may not be as obvious if you're simming the game and reviewing the boxscore after-the-fact -- it superficially seems "correct" that the player with the flashy pitching/batting line gets PotG -- but if you're actually watching the game play out inning by inning, occasionally the PotG selections will be confounding and will not seem true-to-life. After all, in the case of a hypothetical 5-4 win, the next day's recap in the real world would open with "Joe Smith hit a 2-run walk-off double in the bottom of the 9th to lead Richmond to a 5-4 win" as opposed to "Ron Clark scattered 7 hits over 7 innings as Richmond came from behind in the bottom of the 9th to win 5-4"; typically, the Player of the Game would be bestowed on the player who had the greatest impact on the game, as opposed to the player who strictly had the best batting line or pitching line. Much of the time, it's the same person -- i.e. the guy with the 3 HR game usually is the one chiefly responsible for his team's win -- but there are many cases where it's not.
As others in this thread have suggested, I'd like to propose using Game WPA as the methodology for determining Player of the Game. Advantages:
- it takes into account game context
- it takes into account league run environment
- it uses a common scale for both hitters and pitchers (Batting Game Score and Pitching Game Score both work on a loose 0ish-to-100ish scale [not truly], but +5 points in a Batting Game Score is not necessarily equivalent to +5 points in a Pitching Game Score; on the other hand, a WPA of +0.1 by a pitcher legitimately has the same value as a batter +0.1 WPA.)
- by definition, hitting and pitching contributions are weighted equally (since WPA is a zero-sum metric)
- Player of the Game for someone on the losing team is still very much possible, but more infrequent
Strictly speaking, I don't think there are any drawbacks for using WPA to determine Player of the Game.
Practically speaking, the one downside I can envision is potential puzzlement from people who sim their games out and don't see the context in which player contributions took place -- "How did this guy who went 1-for-4 with 2 RBIs get Player of the Game over this other guy who drove in 4 RBIs? Is this broken?"
(The only other concern about using WPA I would have is whether it's derived from the various win probability percentages that are actually displayed onscreen during the game, as I'm pretty sure the table being used contains a few errors -- i.e. instances where win probability for the losing team incorrectly goes
up when the losing team makes outs, etc. -- which would result in some individual WPA values being off.)
Anyway, I hope this was easy to follow, and hope that the OOTP team will consider switching to WPA for PotG. Very much looking forward to all the new additions in OOTP16!