Player Projection Adjustments

[Stat Boy]

I think, based on years past, you really don't fiddle with stat projection adjustments on most of the players at this point unless it's to allow for an injury or playing time projections. But how do you (or do you ever) adjust projections based on news coming out of Spring Training?

For instance, what about Spencer Strider, who's apparently added an effective curveball to his repertoire and been lights out this Spring Training. Do you perhaps improve his projection based on the combination of legitimate news and results? Or similarly, if there's a batter who's noticeably stinging the ball more who went to Driveline over the off-season. Would you maybe adjust your thoughts on his projection?

I realize this is a slippery slope and certainly runs the risk of reading too much into spring stats. But, at the same time, certain players -- such as the examples above -- might warrant an upwards tick in projected stats, no?

[RAY@HQ]

My philosophy is that those kinds of changes are way more likely to do harm than good in terms of projection accuracy. I very rarely make those kind of adjustments.

Plus, it's good to be consistent about these things. If I don't make those adjustments, and you know that, you're free to make your own adjustments if you want to.

[Stat Boy]

I assumed that was the case. And I assume most other sites have a similar philosophy. It's also why I think most projection systems don't go out on a limb and tend to bunch together, for the most part, with only minor differences. HQ does have the Speculator and their UP and DOWN predictions, which can be helpful. But I honestly think that if there is a recognizable change in a player's skill set -- like a new pitch or a new approach -- that is being demonstrated in ST, it's sort of derelict not to factor that in among all the other data used to project a player. Again, I recognize the slippery slope nature of doing that and the lack of any track record to show it's relevant -- but it might make sense to at least keep a running tally of those kind of situations for the next couple of years and see if it can be predictive or not. But, in the meantime, I'll just make my own judgment calls as these stories arise. Thanks for clarifying.

[RAY@HQ]

The percentage of new pitches in spring training that get shelved before Opening Day might surprise you.

[Stat Boy]

That's true. And why I understand how this idea would be problematic and would need to be done on judiciously on a case-by-case basis as opposed to a blanket thing.

But specifically this spring, I'd say Spencer Strider's curveball and Tylor Megill's splitter have had demonstrable impacts thus far. And while the caveat of, "It's only Spring Training" is always there, if I were putting my name on a projection for this coming season I'd want to allow for those pitches seriously affecting their results. So I'll go out on a limb and say both those pitchers will significantly outperform what HQ is projecting for them this season. I think Strider will have an ERA under 3.00 and Megill under 4.00, both a full .50 under what HQ is projecting for them.

[RAY@HQ]

Regression is the most powerful force in our game. I can tell you that I would never improve Strider’s projection because of the new curveball. How much better can it make him, vs all the things that could pull his results in the other direction?

[RABohms]

That's true. And why I understand how this idea would be problematic and would need to be done on judiciously on a case-by-case basis as opposed to a blanket thing.

But specifically this spring, I'd say Spencer Strider's curveball and Tylor Megill's splitter have had demonstrable impacts thus far. And while the caveat of, "It's only Spring Training" is always there, if I were putting my name on a projection for this coming season I'd want to allow for those pitches seriously affecting their results. So I'll go out on a limb and say both those pitchers will significantly outperform what HQ is projecting for them this season. I think Strider will have an ERA under 3.00 and Megill under 4.00, both a full .50 under what HQ is projecting for them.

How much does the Strider projection really matter? He is a first round pick and the first pitcher off the board. Are you taking him at 2 behind Acuna or 5 or 6 behind the top handful of bats? Cheaper arms like Megill and other spring training risers are the guys you have to jump ADP to make sure you get them if you want them.

But also when Tylor Megill gives up a spring training homer to Tim Anderson, do you go back and soften the projection

[Stat Boy]

RABohms, you're taking drafts, I'm talking auction value. I think it makes more of a difference in auction, say, with Strider as to how high you're willing to go.

[Stat Boy]

Ray, how much better can a curveball make Strider? A lot. A lot more strikeouts and a lot lower ERA. I'll take the bet right now with you on the under for HQ's ERA projection for Strider. You in?

[RAY@HQ]

No thanks, not my thing.

[JonE]

It doesn't mean much for anyone to say they'll take the over on Strider's projection, because Strider could outperform his projection even if his curveball doesn't change a lick. Underperforming wouldn't mean anything either. If he finishes at $36 or $40, the projection is still correct: Strider performed at a top-of-the-league rate.

But that's besides the point. When it comes to Strider in particular, I think Bohms is right even despite the need for an auction value. If he was projected at $38, and the best pitcher on the board, but a curveball would make him (let's say) a $43 pitcher, then he's still the best SP on the board. When you talk about the best pitcher on the board, there are a lot of other factors affecting his final purchase price. (Is your auction pitcher-crazy or pitcher-hostile? Do top players generally go for more than their projected value, or less?) You might end up paying $36 or $41 for him whether his actual projection is $38 or $43. It seems to me you're talking about the difference between "better" and "a lot better," which is not a very useful change from a strategy perspective.

Obviously this would matter more if we were talking about Megill, and if we knew that a new pitch would raise his R$ from something like $2 to $8. But that kind of change is not really projectable at this point, based solely on a handful of innings against mixed-level competition. I haven't seen any studies that demonstrate a hard connection between spring pitch mix and in-season performance, so I'm not surprised that BHQ doesn't try to force such changes into their projections. I'm assuming you've seen such studies and that they are private; once they get into the mainstream BHQ can add them to the formula. I understand why Ray doesn't want to make individual guesses based on news that might turn out to be noise.

[Stat Boy]

All fair points, and maybe news like this belongs to the Speculator column, or the UP projections alone. I just think, in general, HQ and other sites with projections, tend to bunch around a consensus middle ground. There will be a few outliers -- as is usually highlighted by the ADP comp columns, where bargains are found based on HQ's projections vs. general ADPs. But, for the most part, the differences tend to be marginal. I just think less quantifiable factors -- players improving their mechanics at Driveline, new pitches, et. al. -- are rarely factored into projections, and can actually impact player performances a lot, much like recoveries from lingering injuries can. But I understand that they're not as dependable as things like barrel rates or hard-hit balls, etc., to base a projection on. So I get that. But it does seem like there's room to start tracking these things from year-to-year and see what legit changes in a players's approach might be more sticky than others. It seems at least a little worthy of some research.

The difference -- at least for me -- in an auction is when a player like Strider (or any top, expensive player) comes up for bid, there's always a clutch moment when they reach whatever your bid limit is. So if everyone's going off more or less the same projections and Strider, for argument's sake, is valued at $34 based on his assumed projected stats, but you have him potentially worth a lot more -- like $42, the confidence in going considerably higher than the room is there. Obviously, that applies for any player. Someone in the room will ultimately be willing to go higher than everyone else to get said player. But when you have a potential outlier like Strider, how high makes sense? And to my mind, if one has the curveball making him even better than the high level people already have him at, you might be willing to be the one to go higher than everyone else in the room.

If it makes no sense for HQ to incorporate that, I get the argument that you and Ray are making, and it's left for the individual player to adjust the projections as they see fit. I totally get and appreciate that argument. I just think there's an untapped area of X factors that maybe should influence projections that were mostly made in the fall, but rarely if ever are taken into account. But I also understand the potential of those factors being too subjective to truly incorporate for the site.

[ToddZola]

This is the classic projection versus prediction discussion.

(c/p from the Internet thingie)

Projection and prediction are two different concepts.

Projection refers to estimating future trends based on past data. It is based on historical data.

Prediction involves making an educated guess about what might happen in the future.
______________________

(back to me)

If there ever is enough data to conclusively say, "When a player adds a curveball (or whatever), his skills improve by X%", that can be incorporated into a projection.

I'm not holding my breath.

[Timmeh49]

It doesn't mean much for anyone to say they'll take the over on Strider's projection, because Strider could outperform his projection even if his curveball doesn't change a lick. Underperforming wouldn't mean anything either. If he finishes at $36 or $40, the projection is still correct: Strider performed at a top-of-the-league rate.

But that's besides the point. When it comes to Strider in particular, I think Bohms is right even despite the need for an auction value. If he was projected at $38, and the best pitcher on the board, but a curveball would make him (let's say) a $43 pitcher, then he's still the best SP on the board. When you talk about the best pitcher on the board, there are a lot of other factors affecting his final purchase price. (Is your auction pitcher-crazy or pitcher-hostile? Do top players generally go for more than their projected value, or less?) You might end up paying $36 or $41 for him whether his actual projection is $38 or $43. It seems to me you're talking about the difference between "better" and "a lot better," which is not a very useful change from a strategy perspective.

Obviously this would matter more if we were talking about Megill, and if we knew that a new pitch would raise his R$ from something like $2 to $8. But that kind of change is not really projectable at this point, based solely on a handful of innings against mixed-level competition. I haven't seen any studies that demonstrate a hard connection between spring pitch mix and in-season performance, so I'm not surprised that BHQ doesn't try to force such changes into their projections. I'm assuming you've seen such studies and that they are private; once they get into the mainstream BHQ can add them to the formula. I understand why Ray doesn't want to make individual guesses based on news that might turn out to be noise.

+1

I like Ray's approach: stick with the tried-and-true methodology. And then leave it to the user to decide how to bid.

FWIW, I think this line from Ron's classic essay on valuation is very relevant here:

For any player ... the one piece of information that is far more important than the most accurate projection... is

how the other owners in your league value that player.

[usualsuspects]

If there ever is enough data to conclusively say, "When a player adds a curveball (or whatever), his skills improve by X%", that can be incorporated into a projection.

I'm not holding my breath.

I think that the pitch level data does exist, it is just tedious to find and use. Sarris for instance seems to think that it takes about 80 balls thrown in anger to determine the effectiveness of a pitch. And there are a bunch of articles and spreadsheets over at the Athletic trying to incorporate that into ERA and K% projections. Also Fangraphs has a lot of various pitch effectiveness stuff in their charts. So you could perhaps see that if Joe Blow apparently has a wipeout slider but only threw it 10% of the time, versus a mediocre sinker he threw 40% of the time, what might happen if they reversed those ratios? Until their arm fell off from too man sliders, anyhow. However unless you have some advanced computing skills and high level access to data this has to be looked at one pitcher at a time and one pitch at a time at Baseball Savant or FG or something.

In the particular case of Strider, if he is going to throw a curveball, he is going to throw it instead of some pitch he already throws. So if the curve is better than the pitch it replaces, his results ought to improve. Or, the even more difficult case to prove even if looking at pitch level data, does throwing a curve make his other pitches better. Tunneling and whatnot. I could be wrong but I do not think HQ is particularly at the vanguard of using this pitch level data. BP, FG, Savant, the Athletic are other places to look.

Not that it has necessarily been shown that all that adds a lot of "predictive accuracy" for roto purposes.