ottoneu Values Refreshed
Back when we launched FG+ earlier this year, I included a table that put dollar values on players for year one and future year ottoneu leagues. Since then, I have mentioned a few times that I wanted to refresh those values with new projections.
Well, I have auctions coming up the next two weekends and finally got my act together to get my spreadsheets in shape; today, I release them into the wild!
Actually, those of you who follow me on twitter (@chadyoung) may have seen these already, but today I will give a bit more info on what you are looking at.
Before sharing the link, I’ll start with an update on my methodology (you can see the original here).
The biggest difference is that I used a combination of three projection systems – Steamer, ZiPS and PECOTA – rather than only a single system. I do this because I like to smooth out the outliers from each. Some people have told me they think this is a mistake, but it’s a process that works for me. If all three (or even two of the three) agree that a guy is a candidate for a big season, it shows up. If only one does, it shows up but much less clearly. Instead of getting caught up in the particulars of a single system, I am using a modified version of the “wisdom of the crowds” and it has served me well in the past. I do not, by the way, weight them equally – I have a preference for ZiPS in general and Steamer for pitchers, and I take that into account.
The next change is that I made a change to the way I calculate Points Above Replacement (PAR) for points leagues. As discussed with reader Chok in the comments of the article linked to above, I was likely under-weighting the value of a player leading off (or just playing every day) and getting large numbers of plate appearances, by breaking everything down to a per-PA ranking. I still use Pts/PA as my primary method to determine value, but now any plate appearances past #600 are compared to a replacement level of 0. So, for his first 600 PA, Ryan Braun gets credit for his 1.613 Pts/PA minus the OF replacement level of 1.117 Pts/PA, but for the other 76, he gets the full 1.613. I think this will help the issue of undervaluing guys who play a lot and overvaluing platoon players. For catchers, I used 500 PA instead of 600.
I also limited the player pool, only using players who were ranked by all three systems, leaving me with 644 hitters and 652 pitchers.
And, finally, rather than basing inflation on a theoretical set of keepers based on owners only keeping the players my spreadsheet says they should keep, I based it on the actual results of the cut deadline in my leagues.
With that said, here is the link to the spreadsheet.
A couple interesting notes:
1) Inflation is definitely less drastic than it appeared before. This is to be expected – we are not perfectly logical (at least not defined as only keeping guys whose value is mathematically better than his price) in our keepers. We keep players who we think are underrated (I kept a $30 Matt Moore because I believe he will produce at a much higher level than the rankings do); we keep players who we think have more value in the future than they do now (my $5 Trevor Bauer has negative value in the rankings, but I expect he will far outperform the $7 he costs me NEXT season); and we keep players because we fear we can’t fill needs otherwise (I don’t love a $4 Ryan Dempster, but pitching will be hard to come by in the original league). All of this results in more dollars per production being spent before the auction, limiting inflation. That said, in my leagues that are one and two years old, you can see a couple dollars inflation on the top guys and that jumps to closer to $10 in the long-standing original league.
2) Positional strength varied by format. In 4×4, the replacement level values were, from lowest to highest, SS, 2B, C, OF, 3B, 1B. In 5×5, C, SS, 2B, OF, 3B, 1B. For points leagues, it matched 4×4. The big difference here is where catchers rank. I assume this has to do with catchers providing little speed and with the catcher position having a number of low-average, high-power players who play better in 4×4 and points leagues.
Alright, dive in, let me know what you think!
A long-time fantasy baseball veteran and one of the creators of ottoneu, Chad Young's is the Managing Editor for RotoGraphs, and can be heard on the Keep or Kut Podcast. You can follow him on Bluesky @chadyoung.bsky.social.
This is great! Any chance we’ll see AL and NL only values by this weekend (please please please!)?
Since ottoneu doesn’t officially support AL or NL only, I don’t usually spend much time on those. I probably won’t release AL/NL only values. I’d think that if you looked at other AL/NL only values and see how much they differed, you would see a similar impact here, just starting with a different base.
Any chance you could share the spreadsheet with the projections from each? Correlating the players is not the easiest task… lol
see my answer below. Don’t think I can do this. Sorry!
Adam Dunn higher than David Wright? Something doesnt pass the smell test.
Dunn is over Wright only in one format (4×4) and they are ranked within $2 of each other (which is minimal in ottoneu world). Keep in mind that Dunn has OF eligibility and will likely hit between 10 and 20 more HR than Wright. Last year, Dunn was +20 in HR, -13 in R, -.058 in OBP and -.034 in SLG. My composite projections have all those gaps closing (+9, -9, -.023, -.015). With the gaps slipping so much, those extra HR hold more value than any other single stat, and that plus Dunn’s eligibility in OF (which is more valuable than 3B eligibility) bumps him up past Wright.
Because my methodology is heavy on letting the projections speak for themselves, I don’t always 100% agree with how things come out – but in this case, I have no issue with Dunn>Wright in 4×4.
As a newcomer to Ottoneu this is very valuable, both for the thought process behind the system as well as for the projected auction values, so thank you!
In looking at some of the data, I was wondering where your personal preferences are more likely to show – in the fanpoints projections, for example, or in your positional rankings? Consider Ryan Zimmerman and Adrian Beltre – although you value and rank Belter higher, Zimmerman shows up with higher projected points. What accounts for this?
This is just a guess, but Zimmerman over Beltre might have something to do with age-related regression built into the projection systems (Beltre is 33, Zim is 28). While the projection systems see a soon-to-be 34 year old, Chad (and other humans) see Beltre coming off 3 straight very good seasons with no signs of slowing down. Zimmerman is also a bit injury prone. Of course, the projections also see strong recent performance and recent playing time to an extent, but maybe don’t weight it as highly.
First of all, my personal preferences actually have no bearing here. I show my preferences in how I balance the projection systems and other choices I make in methodology, but I do not touch the projections. So Zimmerman>Beltre is purely a result of the projection systems saying that, using my methodology, Zimmerman is the better bet.
but I’d also note that Zimmerman only passes Beltre in 4×4, not in 5×5 or Pts. Think this is due to OBP vs. Avg balance – in 4×4, Zimmerman projects to have a higher OBP and more runs, but Beltre has the advantage in RBI (balancing runs in 5×5) and in AVG, which also shows itself in additional hits in Pts leagues.
Finally, Beltre being ranked higher than Zimmerman in points despite fewer total points comes from balancing number of PA. My methodology assumes that with Beltre you are going to get about 21 games of a replacement level 3B to fill out that spot to 600 total, and that is worth more than 24 points. If you add those points in, Beltre ends up with about 833 points, which outpaces Zimmerman.
Hey I was wondering if you have the composite projections of Zips,Steamer, and PECOTA.
I’m not an excel wiz, so doing that is difficult to me, even though I have PECOTA access.
I thought about doing this but considering that PECOTA projections are not free and I can’t control who sees my spreadsheet, I don’t feel comfortable sharing a spreadsheet that includes (even in composite form) their data. Hope this doesn’t cause too many problems.
Did you do them by hand or there is some excel magic I can do
I have a BP subscription and I’ve been really looking for a way to do this….so it sucks
I should also note. I was looking for a way to create a composite ZiPS,Steamer,PECOTA with OPS,Net Steals, and Net Saves.
I basically did it by hand. You can concatenate columns (first name, last name, age for example) to create a unique string for each player, and then do vlookups off that, but it won’t always match since some projections will use different names (Matt vs. Matthew Joyce or Moore) or different ages (seasonal vs. age today). So I had to make those matches by hand by editing names/ages until they matched.
I had the same problem with the lack of uniformity. Some have different names and formats, which make things…difficult.
What would be the formula for an OPS/Net Steals league compared to you 4×4 and 5×5 formulas?
I guess I’ll just do my own.
I’m new to Ottoneu this year, so can someone please explain this very directly because I am awfully confused on both this and the previous value sheet about the system. I would assume that the $Yr1 auction values are for new 5 by 5 leagues drafting this year, but it appears these values are differnt than say, the ESPN auction estimates. They seem a bit high from where my league drafted, but whatever. What is confusing is how they go negative starting around Will Middlebrooks. This makes no sense to me and I’m probably missing some saber methodology thing here. I’d really appreciate some basic explanation of how to read this.
Nothing fancy going on here. The Yr1 values are, in fact, for a new league. They go “negative” because that is about the point at which players become “replacement level” or lower.
In 5×5, there are about 350 players with values of $0 or higher. My methodology assumes that between prospects and “flyers” (guys who are not expected to have real value but have upside – see Scott Kazmir for an example of a player who is suddenly getting some love despite clearly not being projected as anything), most teams have around 10 players that are going to project negative.
Let me know if that makes sense.
Also, for this netting out higher than your league actually drafted, that does not surprise me – going from salary caps and rosters around $260 and 25 to $400 and 40 is a pretty big shift, and I think most new owners tend to underestimate how big a shift they can make. The result is that a lot of first year leagues seem to bid lower than is necessary.
Curious what an empty roster spot would “earn”?
On my own work, I had an empty SP slot as the 8th best SP option (excluding keepers) at one point. It didn’t pass the laugh test, so I dropped it, and further refined some stuff that might have eliminated it. An empty MI spot was nowhere near competitive.
Thanks Chad – that helps a lot
Chad, I noticed that the “above replacement” auction values add up to somewhat less than $4800. Can I assume you are allocating some portion of the aggregate league budget to players that are not expected to make positive contributions (e.g. prospects)?
Exactly. Can’t remember the exact numbers off the top of my head, but should be roughly $40 and 10 spots per team, or something along those lines.
I have been working on my own excel spreadsheets using only the PECOTA projections from BP and I stumbled upon this and like the combinations better so thank you.
My question is on the “Points” tab. Are those fangraphs points? Are all weights and point values represented? For instance, does it include caught stealing, doubles, and triples projections as well?
yes, that is fangraph points and yes CS, 2B, 3B, and anything else is included.
Are these rankings adjusted for positional scarcity? I’m playing in a league that uses FanGraphs points scoring, but is considerably shallower (less teams and smaller rosters), so I’m wondering if this ranking has value to me (dollar values would be off, but rank order is useful) or not.
positional adjustments are made based on an assumption as to the number of players at each position who will be owned. So based on the 40 man rosters with an assumption of about 30 players meant to add value this year, I assume, for example, that 32 SS will be owned, so the replacement level SS is #33. If you used smaller rosters, fewer SS would be owned, so all of the SS would become less valuable (because there would be better players in the FA pool). This would impact all positions, but maybe no equally. If there are 25 good SS and then the bottom falls out, shrinking rosters could mean that suddenly you go from a FA pool of horrendous SS to a few solid options, which could cause SS values to drop a lot, while maybe at 2B the impact is smaller, meaning that they would actually gain value relative to the SS.
does that make sense?
Woo-hoo, I got mentioned in a Fangraphs article!
Your replacement level for starters is right where I calculated it (about 3.8-3.9 points per inning) but your replacement level for relievers is quite a bit lower (5.8 as opposed to 6.8). In other words, I saw the top (300 IP x 12 teams) or so innings pitched by relievers last year to flatten out at around 6.8 points per inning, and am assuming that relievers will do about as well this year as they did last year. Are you factoring in Holds? You may be hedging for reliever unpredictability more than I do, which is probably smart.
You also have some really odd outliers. Hector Noesi for $1.11 with only 3.18 points per inning (was worth like -$22 last year). Travis Wood for $10.40 at 3.78 points per inning (was worth -$16). Brian Duensing at $11.29 for 3.5 points per inning (hard to value his starts and relief innings, but 3.5 is well below replacement level for either).
Also, just in terms of strategy, I always assume that $40 is allocated to roster spots (minimum $1 each), and $360 is allocated to gaining an advantage, on each team. So, I do a lot of =int(dollarValue)+1 type calculations, which I think might be closer to how valuation should work. It’s not possible to spend $0.01 on a player. If he’s worth drafting, he’s $1 or more.
I do account for holds but I really hedge on those. I think my leaders in holds in this version have like 15-17 or something. I know there will be players with more, but the projection systems don’t project them and I have not found a good way to to accurately project them either. The equation I am using right now projects about the right number of holds and about the right number of pitchers getting holds, but spreads them out way, way more than they actually will be. This leads to the lower replacement level, and it also likely deflates values for any players who you are confident will be hold machines. I probably underrated Pestano, for example.
As for those outliers, as I mentioned above, I don’t mess with the projections that the systems pump out. If they expect Noesi to regress and Wood to step up, my values will reflect that.
Perhaps you could join me in lobbying Niv to create a points system that is a little more purely sabermetric and does away with the somewhat ridiculous statistics of saves and holds altogether. I think the easiest way to do that would be to have IP worth a little bit more (like +1) and to have GS (game started) be a modest penalty (like -5). Essentially, the starter would owe you five innings at 5 points each and then anything beyond the fifth would be 6 each, like relievers would get. That way, a workhorse pitcher who pitched a complete game would be credited for coming in “in relief of himself”. A rested bullpen is valuable to a team, so starters’ contributions would be perhaps better reflected this way, and even relievers on sub-500 teams with few save and hold opportunities would be valued appropriately.