Good article, one small comment when you said this:
"There are some flaws with this approach. For one, it doesn’t factor in the game situation. The bases loaded in the 9th inning"
There actually is no flaw with RE24: it's doing what it's designed to do, using the base-out situation only.
To include the inning-score, well we have ANOTHER metric for that: WPA (Win Probability Added)
I mean, you can argue that OBP and SLG are flawed because they don't include the base-out or the inning-score! The reality is that every metric is designed with certain assumptions. It's not a flaw, but a self-imposed limitation. This goes for OBP and SLG, as well as FIP, ERA, or RE24 and WPA, etc.
You are correct that there's no flaw with RE24. I probably didn't make it clear enough that the "approach" which was flawed was using RE24 alone to determine a player's produced value in every situation.
By the end of last week, I was starting to incorporate aspects of WPA into the mix, creating a blend of RE24/WPA, along with some aspects of fantasy sports scoring. The approach I'm seeking is how to best determine what a player actually did in the past, without venturing into a neutral, forward-projecting metric that strips context of the situation.
The original approach was a fantasy-style scoring system, until I settled on RE24 as the best foundation for the job. No matter if I picked RE24, fantasy scoring, or WPA, the evaluation method would always be flawed if it was entirely based on an individual stat.
What I'm working towards might already exist. It would be a blend of RE24 (run value per the inning situation), WPA (situational value per the game scenario), and fantasy scoring (the value of individual stats compared against each other).
Good piece. I went down almost this exact road and settled on RE24 for the same reason you did. It grades the situation you were actually handed instead of pretending every plate appearance happened in a vacuum.
The Janik call is the right hill. He never had the loud line, but he kept the run expectancy climbing and almost never tanked it. RE24 sees that where most stats don't. Same logic on Oviedo over Reed, getting out of a bases loaded no out jam is just worth more than a clean inning nobody threatened in.
Your point near the end is the one that sticks with me though. If you've only got one stat, everything starts looking like the thing that stat measures. So I made myself give RE24 one job and stick to it. It grades the past. I don't ask it to project, because the second I do I'm swinging a hammer at a screw.
The De Los Santos thing I haven't made peace with. Two homers half erased by bad bases loaded luck still feels like it undersells the bat. RE24 is right about what happened. I'm just not sure what happened is the whole story on a guy who can do that.
Good stuff. Machado wants the analytics out of the way. I'd settle for people just picking the right one and knowing when to put it down.
Making RE24 do one job and sticking to it was my approach as well. I was looking for a stat for this single purpose. Next step is looking for a stat that is forward projecting.
As for De Los Santos, he almost gets penalized for being in too many positive situations. Janik kept the RE climbing, but he also was rarely in a situation where the RE was already high. De Los Santos entered multiple situations where the RE was high, and was penalized for not executing every one of them. When he homers, he gets penalized for the expectancy already being high, but when he fails to execute, he gets penalized because the RE was high. In those situations, you need to execute to a certain level every time just to avoid losing value.
I saw another situation yesterday where a player had runners at second and third with no outs. He hit a sacrifice fly for an RBI. The RE was negative, because he went from -23 with no outs to -2- with one out. The run scored, but the RE dropped from 2.04 to 0.67, which more than offset the run value.
Had he walked, his value would have increased 0.65, versus the -0.37 loss on the sac fly RBI. However, if the next person hits a sac fly RBI with the bases loaded and no outs, they would get -0.73, or -0.38 if the runner at second moved up.
RE seems to hate sacrifice flies in no out situations. Even a runner at third, no out sac fly RBI gets -0.13 RE. Yet, I think the traditional thought is that any run batted in would be some level of positive value.
That last line is the whole thing right there. Every RBI feeling like positive value is the traditional instinct, and RE24 just doesn't share it. A run scored and the metric grades it as a loss. RE24 isn't wrong, that's the brutal part. Going from second and third no out to a run in and one out really does lower what you'd expect the rest of the inning to produce.
But "you scored a run and lost value" is a sentence that should make anyone stop and squint.
What it's actually telling you is that RE24 grades the situation, not the run. The run already happened, and RE24 doesn't score what happened, it scores what the inning is worth now. The sac fly is just where those two things split in plain sight.
Your De Los Santos read nails it. He gets the high expectancy held against him on the homers and held against him again on the outs, so the only way to come out ahead is to clear a bar that's already raised every single time. I guess that's RE24 being consistent, but it's a rub when the bar keeps climbing on him for stuff that already happened.
RE24 tells you what the inning became. It won't tell you the run mattered. On a sac fly you kind of need both, and eventually I stopped trying to make one number do it.
Good article, one small comment when you said this:
"There are some flaws with this approach. For one, it doesn’t factor in the game situation. The bases loaded in the 9th inning"
There actually is no flaw with RE24: it's doing what it's designed to do, using the base-out situation only.
To include the inning-score, well we have ANOTHER metric for that: WPA (Win Probability Added)
I mean, you can argue that OBP and SLG are flawed because they don't include the base-out or the inning-score! The reality is that every metric is designed with certain assumptions. It's not a flaw, but a self-imposed limitation. This goes for OBP and SLG, as well as FIP, ERA, or RE24 and WPA, etc.
You are correct that there's no flaw with RE24. I probably didn't make it clear enough that the "approach" which was flawed was using RE24 alone to determine a player's produced value in every situation.
By the end of last week, I was starting to incorporate aspects of WPA into the mix, creating a blend of RE24/WPA, along with some aspects of fantasy sports scoring. The approach I'm seeking is how to best determine what a player actually did in the past, without venturing into a neutral, forward-projecting metric that strips context of the situation.
The original approach was a fantasy-style scoring system, until I settled on RE24 as the best foundation for the job. No matter if I picked RE24, fantasy scoring, or WPA, the evaluation method would always be flawed if it was entirely based on an individual stat.
What I'm working towards might already exist. It would be a blend of RE24 (run value per the inning situation), WPA (situational value per the game scenario), and fantasy scoring (the value of individual stats compared against each other).
Good piece. I went down almost this exact road and settled on RE24 for the same reason you did. It grades the situation you were actually handed instead of pretending every plate appearance happened in a vacuum.
The Janik call is the right hill. He never had the loud line, but he kept the run expectancy climbing and almost never tanked it. RE24 sees that where most stats don't. Same logic on Oviedo over Reed, getting out of a bases loaded no out jam is just worth more than a clean inning nobody threatened in.
Your point near the end is the one that sticks with me though. If you've only got one stat, everything starts looking like the thing that stat measures. So I made myself give RE24 one job and stick to it. It grades the past. I don't ask it to project, because the second I do I'm swinging a hammer at a screw.
The De Los Santos thing I haven't made peace with. Two homers half erased by bad bases loaded luck still feels like it undersells the bat. RE24 is right about what happened. I'm just not sure what happened is the whole story on a guy who can do that.
Good stuff. Machado wants the analytics out of the way. I'd settle for people just picking the right one and knowing when to put it down.
Making RE24 do one job and sticking to it was my approach as well. I was looking for a stat for this single purpose. Next step is looking for a stat that is forward projecting.
As for De Los Santos, he almost gets penalized for being in too many positive situations. Janik kept the RE climbing, but he also was rarely in a situation where the RE was already high. De Los Santos entered multiple situations where the RE was high, and was penalized for not executing every one of them. When he homers, he gets penalized for the expectancy already being high, but when he fails to execute, he gets penalized because the RE was high. In those situations, you need to execute to a certain level every time just to avoid losing value.
I saw another situation yesterday where a player had runners at second and third with no outs. He hit a sacrifice fly for an RBI. The RE was negative, because he went from -23 with no outs to -2- with one out. The run scored, but the RE dropped from 2.04 to 0.67, which more than offset the run value.
Had he walked, his value would have increased 0.65, versus the -0.37 loss on the sac fly RBI. However, if the next person hits a sac fly RBI with the bases loaded and no outs, they would get -0.73, or -0.38 if the runner at second moved up.
RE seems to hate sacrifice flies in no out situations. Even a runner at third, no out sac fly RBI gets -0.13 RE. Yet, I think the traditional thought is that any run batted in would be some level of positive value.
That last line is the whole thing right there. Every RBI feeling like positive value is the traditional instinct, and RE24 just doesn't share it. A run scored and the metric grades it as a loss. RE24 isn't wrong, that's the brutal part. Going from second and third no out to a run in and one out really does lower what you'd expect the rest of the inning to produce.
But "you scored a run and lost value" is a sentence that should make anyone stop and squint.
What it's actually telling you is that RE24 grades the situation, not the run. The run already happened, and RE24 doesn't score what happened, it scores what the inning is worth now. The sac fly is just where those two things split in plain sight.
Your De Los Santos read nails it. He gets the high expectancy held against him on the homers and held against him again on the outs, so the only way to come out ahead is to clear a bar that's already raised every single time. I guess that's RE24 being consistent, but it's a rub when the bar keeps climbing on him for stuff that already happened.
RE24 tells you what the inning became. It won't tell you the run mattered. On a sac fly you kind of need both, and eventually I stopped trying to make one number do it.