Playing time churn could force BABS to change
As of this past Sunday, 45 players who had been ranked among the pre-season Top 300 have already visited the Injured List. That’s 15 percent of the top players hobbled in little more than 10 percent of the season.
This is important because the more players who get hurt, the more pointless our pre-season playing time projections become. BABS already projects playing time using very broad strokes – Full-time, Mid-time, Part-time – and it’s possible that those categories are not broad enough.
Full-timers like Daniel Murphy, Miguel Andujar and Corey Dickerson (pictured) are already likely to get downgraded to Mid-time. Mid-timers like Jed Lowrie and Greg Bird might drop to Part-time. And where will that playing time go? Former part-timers who are already seeing an increase in playing time include Rio Ruiz, Christian Walker and Brandon Drury, all of whom already have over 60 AB.
BABS did all she could to prepare us, but all these injuries and role changes have quickly made the pre-season charts moot. So I am going to propose a change to the BABS ranking chart and ask for your input regarding your comfort level with this change.
Up until now, players have been ranked within their asset groups by ADP, but Full-timers were listed first and Mid-timers next. The Part-timers were listed separately down at the bottom of the chart. What I’d like to do is list all players with the same asset profile together.
There are two ways we can do this. In the first, players are still separated by playing time and each playing time group sorted by ADP. This just moves up the Part-timers with the other two groups. In the second version, all players are integrated and sorted by ADPs.
|VERSION 1||Full asset group, sorted
by playing time, then ADP
|VERSION 2||Full asset group, playing
time integrated, sorted fully by ADP
Typically, ADPs will do a decent job on their own of moving the full-timers up the list, but I deliberately chose this asset group to show that it’s not always the case. In Version 2, the market placed a higher value on Greg Allen, even though our projections questioned his playing time potential. Conversely, we saw greater opportunity for Richie Martin but the market did not place value on that. (To date, Allen has 32 AB and Martin has 50, showing once again that BABS is always right… about playing time anyway. The two have one SB between them.)
It may be helpful to note that some of the BABS part-timers who would have been more visible in the proposed new version include Dan Vogelbach (P+), Eduardo Nunez (s,AV), Rowdy Tellez (p,a), Mitch Garver (p,s), Fernando Tatis (p,s) and Clint Frazier (PW).
The downside to either version of this new presentation is that it will create very large asset group blocks. Those folks who already struggle with seeing lists of players as anything other than straight rankings will likely find more difficulty with this. And the fact that this 17-player asset group with Victor Reyes and Tim LoCastro will be listed ahead of groups with players like Joey Gallo and Carlos Correa will give people conniptions.
But that’s what BABS is all about – sorting better skills profiles ahead of lesser profiles at all stages of the market.
So, what do you think? Let me know in the comments below, or send me a note from the Contact Us page. Do you like one version better than the other? Would you prefer that the reports be kept as they are? If I make the change, I would not roll it out fully until next year. However, if there is a groundswell of positive feedback, I might try to get it implemented with at least some of the in-season reports.
You are the users; it’s your call.
I vote for Version 1; second choice is stay as is.
Definitely not version 2. The ADP’s order does not necessarily reflect players with the best BABS skills. I agree with Richard.
I am still formulating my opinion on the change in general, but of the two options definitely option 1. If the changes are implemented in-season for the rest of 2019, RotoLab will still sort them the old way through the year, as it is coded that way. I’d have to think about this quite a bit before issuing a patch, the law of unintended consequences being fully in play here.
I tend to favor version 1. I don’t care much about ADP. Would like to see a sample with multiple assets.
You HAVE to care about ADP (market value). If all the players in an asset group have comparable skill, the only way you can find profit is by leveraging ADP.
Version 2 seems to support what BABS actually is more than V1. The game has changed and the 10 day IL has changed it even further. Version 2 is more simple. Here are all the ‘like’ players and here is how the market is viewing them. (by ADP and not playing time). In simple terms, if all of the players in this group had the same opportunity, similar results should be expected. My two cents, but it’s easy enough to manipulate the sorts in excel to configure her how each individual wants to see things anyway.
Yeah, this is kinda true, especially in light of the point I made about playing time churn. I’ll go with the majority once everyone has a chance to weigh in, but Version 2 is more BABSian. Still, usability has to drive the decision.
If, as Ron writes, we HAVE to consider ADP (market value) to find “profit”, then version 2 most clarifies that consideration.
On another subject, are my questions in the forum so bad/stupid that they don’t deserve a response? The last two I’ve asked, both in “players”, the most recent about a trade, and the other about roster management, have gone un-answered. Not asking Ron, as Ron is usually good, and also busy….just asking fellow subscribers to please lend their expertise. Thanks.
I like the second option.
I prefer the second option because it reflects market value better.
I prefer the 2nd options also.
I like option 1, if you do either though I see the concern of it becoming too long until the next asset group. To help, maybe add a key of what asset groups come before others (similar to how a winning poker hand hierarchy would be) to let the users know when to maybe expect certain combinations of assets.
Thinking about RotoLab, I think the best solution there would be a toggle switch that would (optionally) allow including the part timers in the asset group.
Thinking about how I would use this information functionally, I would think the parttimers would get in the way in the first half of the draft, but having them integrated would be really useful in the second half of the draft. It would be especially useful when trying to fill out the roster (e.g. looking only at 3B, with most Fulltime players and many skilled Midtimers already drafted).
That should be fairly easy to do on a database (like RotoLab). On a spreadsheet, the easiest way would be to adopt the second format above, and use a filter to hide or show part timers.
Being a hard core RotoLab user I agree!
We’ll see how this sorts out then I look at it for next year.
Also a rotolab user and am concerned about how BABS will be presented in the future. I’m leaning toward #2: a player’s skills are intrinsic to the player, his playing time is not. Playing time categories are still predictions. Presumably, players with better skills will be given more playing time when the opportunity presents itself. A toggle between the two sorts would be nice.
Questions for Ron: Do the PT ratings account for the likelihood of opportunity arising for skilled players, or for the pressure being exerted by more highly skilled part-timers on lesser-skilled players, or do they simply reflect how his current team says it will use the player? 2) How large a sample of leagues do you draw from when producing your ADP rankings?
1) I don’t think it is possible to account for the precision inferred from variables you list in a PT rating that is so broad. All players on an MLB roster are assembled to see who will play where and how much. There is not much accuracy you can expect from that.
2) It’s whatever the NFBC feed gives us.
I think I like Version 1 better, but can’t really articulate why. Some of the comments do make me think market bargains stick out a little better in Version 2.
Even though I like Version 1, I think I agree with that statement. In Version 1, the PT is ordered and I look for players with ADPs that are relatively high compared to the players around them, like Gardner. In version 2 it sticks out a little more. But as others have mentioned before, some of these Asset groups would get pretty large. Players with ADPs of 600+ just clog it up.
I like version 2. But I used Rotolab for drafting so it doesn’t really make any difference how your lists are compiled.
As an aside, the PT projections are extremely important but also very difficult to nail. Take the example this year of how BABS lagged the market and ranked Matt Kemp and Hunter Renfroe as FT players in February/March. Therefore, I agree the PT factor is the most important means to upgrade BABS from an excellent to a superb tool.
After digesting this, I think the reason I prefer Version 1 is, Version 2 makes no order distinction on the Playing Time Asset, which IMO is the entire point of BABS. Within an Asset group, these are all similarly skilled players. What differentiates them is Playing Time. Many of them (we are hoping) could put up similar stats if given increased Playing Time opportunities. But sorting on the ADP in Version 2 minimizes that insight, and largely ignores what BABS is trying to tell us. Relying on the ADPs as the primary sort in the Asset Group just screams, Go along with the crowd.
In Version 1, Playing Time is still the most important criteria for assessing the value of each player within the Group. The ADP just tells me which players are more likely to be bargains at each level. Version 2 makes the ADPs the central focus and the Playing Time Asset plays second fiddle. Any list that puts Greg Allen above Brett Gardner based on his 340 ADP versus Gardner’s 383 ADP is far more meaningless than a list that puts Gardner and his F Asset above Part-timer Greg Allen.
This situation further deteriorates when we consider that we don’t get updated ADPs after draft day. We are stuck with what we know as of the start of the season. Like ADPs are going to mean anything useful in a BABS August Report. But we can continue to update and use the Playing Time Asset. Better to make that the more prominent sort, in my opinion.
But what if the Playing Time indicators can’t be trusted?
We can probably trust them as much as the ADPs that are driven by the recency biased masses that log onto NFBC.
Merv nailed what I was thinking as well. While version 2 is closer to a ‘pure’ BABS list, drafting players based ONLY on their core skills might yield a roster full of highly skilled benchwarmers. The assessment of playing time is a fundamental way of differentiating value across a group of similar skills.
Giving it a thorough enough exam, I go with Version 1.
I vote for version two, I like to see all of the players with similar skills and how the market views them in order. I can just look at the PT column to see potential play time.
Version 2 – Simply because ADP is more trustworthy and actionable than playing time.
I can’t say that I agree with that. Given Ron’s article in The Athletic that examined the ADP results and what I dug up for 2018 (see the post a few below), I think the PT Asset hits the mark better. I only looked at 2018, and may look at 2016 and 2017 later this weekend. But 2018 shows the PT Asset hits its mark better. And I somewhat would expect that, because it is a more sober look at the player roles and injury concerns in late spring that drive the PT rating. The ADPs are driven by the owners at drafts and they are always influenced by career years the previous season and players having hot/cold camps.
A question for the “Version 2 tribe”. You are in the middle of the draft and need some Speed. Looking at your Version 2 BABS list above, and Villar, Hernandez, Anderson and Kiermaier are off the board. Do you take Greg Allen? Or do you skip down and take the Full-Time Brett Gardner? If your answer is Gardner, I am not sure why you want them ranked by ADP, because you’re ignoring it. That F playing time rating means more to you than the difference in their ADPs (and that was the point of my post above). If you say Greg Allen, do that a few more times and you can post on this board in July what 9th place feels like.
No offense to the Version 2 tribe, just trying to elicit some conversation on this interesting topic.
I definitely see your point.
If I’m thinking about it in extreme terms, it’s not inconceivable to end up with a post-draft team featuring, let’s say, 6 part time players in my “starting” lineup. We love BABS because she brings forth the skill profiles right before our eyes. However, from a behavioral perspective, I don’t know how comfortable I would feel if, come Memorial Day, I am praying for 4-6 injuries to occur so my part time players can…play.
With that said, I don’t think this dilemma requires us to completely ignore projected PT, as that can become a dicey proposition. Perhaps some of this problem is buffered through the INJ liability – ex.) David Dahl has the INJ liability, so Raimel Tapia gains some “value” and PT potential because of this. Maybe part of the solution needs to encompass this relativity , i.e. a part time player behind a INJ player may have more value than a part time player behind an injury-free player? Could THIS (playing behind a INJ player) be an asset that can be added to a player’s skillset? Just thinking outloud…
I was thinking the same thing. If the PT ratings are projections that take into account the likelihood that, for example, highly-skilled part-time Player A is likely move up to M or F status because the F player he is playing behind is also |-AV,-INJ; or that aging full-time Player B with degrading skills is being backed up by young Player C with p s AV| -EX skills, that has value in a draft. If they are a reflection of how management has said they plan to use a player once the season starts, while useful info, it may be less useful than a thoughtful projection of a player’s PT potential.
If it were possible to see both ratings side by side, this would be off-the-charts useful. Trout is F-F, no doubt, but maybe Player B above is F–, indicating he’s FT now, but is projected to lose his current starting roll as the season goes on and Player C is –M, indicating his current part-time roll is projected to grow into at least medium use.
The great part of my job is being able to research stuff like this at 2:00 in the morning (the sucky part of the job being fixing bugs at 2:00 am after a 14 hour day). I was a little curious on what a few people had remarked about the reliability of the PT Asset and the ADPs. Using the 2018 Final BABS preseason report and the year end report, I looked at the hitters that appeared in both reports, plus a few players that ended up with an F or a P and were not in the preseason report. This resulted in 579 players to examine. Here are the results. The PT projection count for each Asset is on the left and actual 2018 PT Asset finish are the 3 bins on the right
PROJ 2018: F M P
F = 179 107 54 18
M = 108 21 54 33
P = 292 7 32 253
579 135 140 304
The 107 F’s out of 179 projected F’s is surprising, maybe even alarming. Tomorrow I’ll look at ADPs. I have no idea how this will format after I post it.
My March 7 column at The Athletic did the ADP research for you. The important results: “In the active roster draft, a third of our picks performed at par or better; two thirds performed worse than where we drafted them. Fully half of them could have been considered busts.”
OK, that makes me feel a bit better about the PT results. Before I dug into the 2018 PT Asset numbers, I would have thought they would be a little better than what the actual results were. Given the ADP results, they don’t look so bad.
If only 107 of the 179 F’s turned out to be F’s and only 54 of the original M’s turned out to be M’s, it seems to me that the BABS playing time should not be a determining factor but a guide. I use the ADP only as a barometer to when I think I should draft a certain player.
In the Allen vs Gardner discussion, if you’re in the late stages of a draft and you need speed, it’s not a problem to look at the Version 2 grouping and make a decision based on who is available at that point and who might be available a round later. I vote for Version 2.
Yeah, I wish the success rate was a little higher, but this is driven by the MLB’s team liberal use of the IL. The biggest thing I took from the PT results was out of 292 part-timers, 253 ended up there. One thing that seemed to work well (in 2018 at least) is identifying the scrubs. I mentioned this elsewhere, I wish the PT Rating was broken down into F, M, P and none. Not that it would bring anymore accuracy to the rating, but it would let me know if I was drafting a projected 100 Plate Appearance guy or one that is projected for 275.
Do you take Greg Allen? Or do you skip down and take the Full-Time Brett Gardner? If your answer is Gardner, I am not sure why you want them ranked by ADP, because you’re ignoring it.
I’m a V2 guy. Yes I’d take Gardner; but a round later. I’m not ignoring ADP. The ADP ranking tells me I can possibly wait another round to take Gardner. If somebody takes Allen while Gardner is still on the board, BABS says I shouldn’t care.
Version 1 tells you the same thing, as well as the current iteration of BABS. I’m not opposed to version 2, but my concern is we are going to have to wade through a lot of Greg Allen’s in the middle BABS tiers. We will always take an F over a blank PT and in most cases will take an M over a blank PT, regardless of what the ADPs say. I say “blank PT” because it brings up my second point which seems related here.
I had a conversion with an HQ writer this past spring about the BABS system. His thoughts on the playing time Asset was that M playing time Asset covered too much ground. 300-500 PA seems awfully broad to me as well. In a mixed league, I will certainly roster a player with 480 plate appearances, but a player projected for 300, I don’t want anywhere near my roster. In standard Roto, 280 PAs is worth rostering at the end of your draft, but in BABS they have the same profile as a player projected for 100 PAs.
Is there room for another PT: F, M, P and blank? I know the PT estimates in March are not terribly accurate, but as Ron pointed out, neither are the ADPs. But those two are what we have to work with. I’d feel a lot better with either version 1 or 2, if the true deadwood was weeded out. And from my earlier analysis I posted last night, the one thing that can seem to do well if identify the stinkers. Out of 292 project part timers, 253 ended up there.
I like the idea of one more level of playing time. The boundaries of any category would necessarily be arbitrary, but it wouldn’t be hard to run a few studies to see (a) how many players fit into given PT bands, and (b) how good are the correlations between what was projected and what actually resulted. It might also be very useful in trying to understand where PT “leapers” tend to come from: Ms, Ps, or blanks.
Jon, see my post a little earlier, but this is how 2018 shook out:
PROJ 2018: F M P
F) 179 = 107 + 54 + 18
M) 108 = 21 + 54 + 33
P) 292 = 7 + 32 + 253
All) 579 = 135 + 140 + 304
I agree with this as an extension to version 1.
I am also sceptical about ABP as a be all and end all.
Great comments here. Please allow me to add a few thoughts also.
For frame of reference, I have used BABS and RotoLab for the last 2 years, doing many drafts and auctions, both live and online. I love it! I have HQ data also in the RotoLab file but I draft from the BABS tab.
1) I prefer Version 1 vs Version 2. However, I am concerned that the asset classes will become way too large if ALL of the players are added to the class regardless of playing time. Especially true with the pitchers where there are fewer classes since it is primarily ER and K once you get thru the possible save guys, and so many relief pitchers who will have “blank” PT.
2) Playing time is very important part of BABS ratings, but I think the current combination of the F and M players captures 95% or more of the relevant player pool. There will be a few “unknowns “ usually those rookies, especially on the pitching side like this year’s Whitley and Luzardo. Capturing those examples as M solves that issue.
3) ADP is also very important, as Rob stated.
But I do think we need to do more to keep ADP “current” such as the most recent 1-2 weeks of drafts. ADPs from Dec-Jan aren’t as valuable when we reach March drafts.
I can’t speak for BABS, but the ADPs in RotoLab are from HQ which ultimately from the NFBC, so I get what I get. I poked around the NFBC site this past spring to see what was available and I didn’t recall seeing options for picking and choosing which time frame. But I agree, ADPs over the previous 2-3 weeks mean more than extremely early drafts.
I must say this article has lit up the posting board which shows that we have some very knowledgeable people on this web site. I do agree ADP taken in Dec-Jan are not as valuable as those taken a month before the season starts.
I can see the pros and cons of both versions but predicting PT of low ADP players is like predicting the INJ of high ADP players. We don’t know who will get injured or when so projecting a path to playing time for a backup is a difficult task. Usually in my reserve part of the draft I already have a list of players ranked in order of skill set not PT. This includes hitters and pitchers and includes some players in AAA. I draft for skills first, then see how the season progresses. Sometimes I get lucky and other times I have to cut bait to find a replacement in the FA pool.
Maybe we could get BABS a crystal ball for her birthday. LoL.
I’m on version 1. Maybe a simplistic thought on my part, but… ADP is relevant to BABS, but it shouldn’t drive or take precedence over BABS’ primary elements: asset groups and playing time projections. Version 1 stays true to BABS elements while making ADP information available. Version 2 prioritizes ADP/market over BABS.
It’s always better to trust someone in a relationship and accept them for who they are than it is to try to change them.
Like many, I vote for Version 1.
Possibly naive question, however: does BABS need to be more nimble in defining full and mid timers? If we define full timers as players who play in at least 140 games, how many players will meet this standard? And there’s a broad range of mid-timers. How do we separate players who go on the IL for the minimum time from players who are out for two months with an oblique injury (see Judge, Aaron)? Perhaps BABS should consider dividing playing time into full, medium, and short-time playing categories. Defining terms would be a challenge, of course.
The way that I use BABS, version 2 would be best for me as long as the ADP is updated frequently. Thanks for asking for our input.
I like version 2 as this is what I was doing anyway to see who might be available late that the market might be undervaluing.
I have deliberately not read any comments yet- I prefer version one. I want to know who BABS is valuing as the default setting. It’s very easy to then scan ADP’s for inefficiencies. In the end though- it doesn’t matter to the more hardcore of us. I suspect that many, like me, have all this info in a sortable/filterable format. This draft season, I switched between all 3 of these views on my cheatsheets.
My immediate reaction is:
Version 1 for Draft Day.
Version 2 for in-season.
I like Merv’s idea of thinking of playing time as an “asset.” At the draft or the auction, we should definitely place a certain value on that particular asset. That said, BABS’s judgment of that asset is of a moment in time. She is doing the best she can with the info she has. But I don’t expect her to know that Corey Dickerson (pictured) is going to get hurt or what the nature of that injury is. She can tell me whether he has an elevated risk of injury, but that’s not a guarantee either way. If I drafted Dickerson, it’s up to me to figure out the best available player to replace him. I might want to focus on the players who are going to get his playing time, especially if I’m in a deep or only league. Or I might want the players with Dickerson-like skills who have more PT at the moment. Yes, the market liked Allen more than the others, but that was back in March. And neither the market, nor BABS, knew that Cleveland would sign Carlos Gonzalez, and at the moment I think he’s going to get more ABs than Allen. Come to think of it, based on how the rest of my team is performing, I may want a completely different asset group (and Cargo would definitely be in a different asset group).
Is that a straddle? Maybe, but I think we need different tools at different times.
I was just about to type something similar. Well put Jon.
Jon, I tend to agree with your premise, but pragmatically the problem is the ADPs are only generated during the drafting season. After late March they are static. By the start of summer, they are pretty much useless other than to indicate what drafters thought of the players in March. What we do get all summer is updated PT ratings based on MLB role changes and events. But, on the other hand, maybe the underlying skill and the PT projections are all I should care about and just use BABS an information source in season, and not worry about the order within the group.
I have read all the comments which I found very interesting and informative; I prefer version 1 because to me Playing Time is more important than ADP particularity when ADP is calculated mostly on early drafts.
To be clear, the accuracy of the ADPs is completely irrelevant. The market is what it is. If the ADPs say the Mike Trout is a 5th round player, that’s information you bounce off the BABS ratings to determine when to draft him (Round 3 or 4).
So each asset group represents a single skills profile. The playing time indicator has been a rough ranking within that set of skills. But all of this is about leveraging the marketplace, so however the players are listed within an asset group, the information that is important to you is the price you need to pay for those skills – and that is the ADPs, regardless of how accurate they are.
Merv – I liked your last comment probably the most…order within the group during the season is almost useless for me…underlying skills, PT and ADP – all are factors in draft season…and we all subconsciously weigh them a bit differently…but once the season starts, just give me what we projected for underlying skills – the variables of particular leagues, injuries, etc. have moved things differently in different leagues…
For me, during draft season – I’m “iffy” on ADP…I tend to agree with Chris Liss’s theory that ADP becomes “group think” very quickly…and the further draft season goes on, the more pronounced that effect becomes. Is it still valuable information because it’s what YOUR market will largely be using to determine value, yes – but one of my favorite things about BABS is how it’s skill-focused…
To me, this argument is splitting hairs – from a usability standpoint, I think Version 1 is easier to see/read/use during the high activity of a draft or auction…during prep – users can do what they want to setup their own lists and decide if they would prefer to target Greg Allen higher…I want SIMPLE during the draft…and to me, Version 1 wins there pretty easily…
I think that’s the issue that sticks out most with me. Once I zero in on an Asset group, I’d rather have the players projected for more plate appearances above those with less. My concern is some of these Asset groups will have a lot of players to digest under V2. (p) would have 39 hitters if you include the blank playing time. In the heat of a draft, less to wade through is better. With V1, at least the 400+ ADP players that we are most likely not to consider are pushed below the F and M players. With most of us playing in 12 or 15 team leagues, we typically aren’t poking around ADP 400 and lower until our reserve drafts.
In the end, I think Version 1 and Version 2 are going to be pretty much the same for most of the asset groups. The market does recognize the players likely to get the most playing time and drafts them accordingly. I deliberately chose the above group because of the Greg Allen/Richie Martin discrepancy that shows the marketplace is not always right. But I had to dig for it – and I do still think it’s worthwhile to recognize those outliers.
And I specifically chose the (p) group because of how many players it would have in the full group. In practical terms, you are correct. Most groups will be similar between V1 and V2, they will have much less than 39 players, and the ADP rankings will roughly mirror the PT sort.
Just give me an early heads up on what you do next year, so I can patch the code in the fall to match. For 2019, the RotoLab will stay the same as preseason.
There are many ways to sort the players. The primary sorts are asset groups and ADPs -that doesn’t change. From there, we can sort by playing time – or – we can sort by health liability (I played with that for awhile – it’s helpful) or really any of the other variables. But we don’t NEED to sort by those variables because we can see them when we draft our players. When Kevin Kiermaier bubbles up to the top of our ADP-ranked list, we can judge whether he’s draft-worthy by observing his INJ liability just the same as observing his “M” playing time asset. We don’t necessarily need to sort by either of those – we only need to know that they are variables we must consider.