Projections are everywhere. Rare is the obsessive fantasy player who doesn’t use projections of some form or another, particularly among those who frequent these pages. Some players are even uncompromisingly strict with the value output that their chosen projections produce, refusing to go one dollar past what the system recommends.
What we don’t normally do is draft exclusively from projections. If you think about it, that would be extremely hard. How can we completely remove our preconceived notions about individual players? Whether it’s an injury that we’re not convinced a player is over or prospect pedigree that we just can’t forget, it’s hard to let projections overrule our instincts. That’s before we even consider the fact that overruling the projections is sometimes the right way to go. While they will usually do better than us in the aggregate, slavishly following the projections and drafting a mediocre innings-eater late in the draft is normally a suboptimal draft strategy. You’d often be better off taking the high-upside play who doesn’t have the track record.
I have often wondered what it would be like if we did have to draft from projections alone. How would we make decisions? What would we target or avoid? How much does actually knowing who the player is influence our interpretation of projections? Exactly how to do this was something I’d never quite fleshed out, but the answer hit me while we were tossing around ideas in our BP Slack. Kevin Jebens suggested that we draft the best past seasons that we could remember without looking them up and then see who would have the best fantasy team, and I both remembered this projection idea and realized that we could draft an entire league without even knowing who the players were.
The concept was simple: get a random season of PECOTA projections; strip out the names, teams, and any other identifying features; and draft exclusively from these “nameless” projections without making any attempts to look up stats to figure out what year it actually was. In order to ensure that I could take part without having an unfair advantage, I got an independent third party to pick any PECOTA spreadsheet, hide all the names, teams and any other potential giveaways like comps before giving the data to me.
We would draft starting lineups only, with 13 hitters—with just one catcher rather than two—and nine pitchers. The best lineup by standard 5-by-5 roto scoring would be crowned the winner. I encouraged people to focus on drafting based on the projections alone, rather than trying to solve the puzzle of which specific year it was, and thus defeat the point of the exercise by identifying the real players behind the anachronistic names.
Two caveats. After some debate, while I thought specific ages might make it too easy to identify certain players, we agreed that ranges were a reasonable inclusion as a relevant aspect of projections. Players were therefore put into four categories: under 25, 25-29, 30-34, and 35 or older.
The second issue, which you might have seen coming, is having a draft with no names at all is quite a challenge if you are going to identify who has been picked. I could have simply given each player a number or random string of characters so we could do this, but that sounded both dull and confusing. Instead, I went to Baseball-Reference, ran a Play-Index query to find players who played in the majors over 100 years ago, and randomly sorted those before adding them to the projection spreadsheet. This may be the only time you see Nap Lajoie, Wee Willie Keeler, Bullet Joe Bush and Bones Ely in the same draft (or, in most cases, any draft).
Nine other BP contributors volunteered for this unusual challenge: Jesse Roche, Jon Hegglund, Kevin Jebens, Mark Barry, Mike Gianella and Tim McCullough from the fantasy team, along with Collin Whitchurch, Kaz Yamazaki, and Martin Alonso. With no ADP, no news, and no names, it was time to build a strategy based purely on the projections.
For those who want to follow along, the full draft spreadsheet that we all used to look up projections and make picks is available here. The first aspect of the mystery projections that stood out to me was that this is definitely not the speed-starved situation in which we currently find ourselves, nor the record-breaking homer environment. A total of 3316 stolen bases were projected with 4808 home runs. Last season we had over 1000 fewer stolen bases and almost 2000 more homers.
If anything, it felt like power might be at a premium. Only seven hitters were projected to hit at least 30 homers. Without personally experiencing baseball in the early days of PECOTA, I wasn’t entirely sure which period this most closely resembled, but it was apparent that more recent assumptions about being able to get reliable power late weren’t going to hold up.
Another change from drafts in more recent years was the plentiful supply of high-volume starters. A total of 42 different pitchers were projected for at least 180 innings, and a total of 12 were slated to go over 200. Volume felt crucial to me in this scenario because we couldn’t stream, but I also didn’t want to play in the back end of this pool. Many of the 180-plus inning pitchers had ERAs close to, if not over, four, with WHIPs in the high 1.20s or even over 1.30.
It would also be really hard to pick out the late-round pitching breakouts based on projections alone. We had no pitch data, no information about mid-season repertoire changes, no articles highlighting how one small tweak could unlock a pitcher’s potential. There were seven pitchers who had over 210 projected innings, four of whom also had elite ratios. I resolved to get one of these four starters, if possible, and to focus on building up my rotation earlier in the draft. I didn’t want to play the late-round guessing game about breakout starters, nor did I want to sacrifice volume because I was scared of drafting bad ratios.
Much of what informed my planning came from working on the depth charts, which drive all the PECOTA playing-time projections. I know how difficult it gets to project time in certain situations, and making speculative picks who we could later drop if things didn’t pan out wasn’t going to work here.
It felt like that uncertainty was a much bigger issue here. As Mike said, the main challenge with drafting from projections only was that it introduced “a lot of luck in terms of guessing major injuries or extreme drop offs in pitching performances.” With no way to compensate for mid-season injuries or unforeseen performance slumps from apparently reliable players, the myriad factors that can affect playing time were heavily on my mind. Was the projection low because they were injured at the time, had an extensive injury history, were they in a platoon or otherwise crowded offense, or perhaps even a prospect yet to debut? I hoped to largely avoid these situations, so I didn’t take on too much volume risk.
That also went for closers. We all know how easily closing situations can change, and in this draft hundreds of saves would likely go unused because those who accumulated them in real life weren’t in the role to begin the year. Grabbing at least one elite-level projected closer who had as firm a hold on the job as projections can give was also high on the agenda.
Much discussion took place over age, both from comments made during the draft and feedback that participants gave me when we finished. Multiple players mentioned that this was one area where bias could still play a big role. Some, like Jesse and Martin, eschewed the 35-plus age category altogether and heavily focused on the players in their twenties. In fact, Martin admitted that not having any name bias meant that his age bias “went into overdrive,” and he passed up 35-plus players with better projections for younger picks.
Others, including Mike, Mark and myself, didn’t avoid the older players. Mark said that, for him, ages “might be a tie-breaker, but I didn’t have a plan to draft a younger team.” I spent a while thinking about the implications of ages. On one hand, those in their mid-thirties could have been about to enter their final season, and we might end up with a total bust. On the other, those players would almost certainly have a long track record, and if PECOTA projected them confidently, it reassured me that they were a proven quality player, not to mention the fact that they were probably more secure in a lineup or rotation.
Some of the younger players would be future stars and had little risk of age-related collapse, but they might also be rookies or sophomores with virtually no track record in the majors. The error bars would be wider. That felt particularly true for the under-25 category. There would almost certainly be some players in this bracket who would turn out to be breakout stars, or who were already well-established because they’d hit the majors at 20 or 21. Those players would be outweighed, however, by those who struggled to adjust or simply put up respectable performances for their age but did nothing remarkable for fantasy. PECOTA might also be speculating optimistically based on minor-league performance but have no major-league basis for the projection. Given that we didn’t know the year, I had no idea how good the system would be in situations where minor-league stats formed the bulk of data for projections.
Ultimately, the best strategy felt like diversification. I didn’t want a whole team over 30, but if these players had strong projections, I wasn’t going to shy away from taking them. I also wasn’t going to have an entire team of uncertain youth, but getting a couple in the youngest category who might break out way beyond their projection felt reasonable. Mostly, I wanted to play in the middle two categories of 25-29 and 30-34, as the safest combination of track record without age-related decline. Like Mark, I felt like age was a good tie-breaker given similar projections.
To ensure that I wasn’t just eyeballing stats and missing good values or overrating certain characteristics, I decided to calculate dollar values for these projections. Perhaps this was taking things a little more seriously than most others, but I could hardly phone it in. I used an SGP calculation spreadsheet, tweaking some old SGP values to what I felt was a rough approximation of the environment. The values confirmed my early feelings. While there were plenty of late-round hitters who I felt comfortable drafting, the last third or so of the positive-value pitchers were either high-volume, poor-ratio types; promising ratios but missing at least 20 percent of the season; or relievers with the kind of saves totals that suggested they weren’t totally in control of the closer role.
I know that I wasn’t the only one to produce valuations. Mike noted that he ran these projections through his 15-team valuations and sorted by dollar values. Reinforcing the point about a high stolen-base environment, Mike said that he had to move away from drafting solely by these numbers because they placed “too much of an emphasis on speed,” and there were so many stolen-base options later in the draft. Martin also calculated his own values, weighting them himself by the priority that he gave them and standardizing them all before multiplying by playing time to get an overall score.
Tim and Mark also sent me feedback, and they didn’t work off values so much as identify specific targets. Both said that they looked for well-rounded hitters who could contribute everywhere, and batting average was a big factor in their decisions. Tim identified a weakness at the middle-infield positions and made sure to fill second base and shortstop inside the first 10 rounds. Mark filtered for targets by slugging percentage and playing time, then stolen bases, while looking at K/9 and innings on the pitching side.
I did a bit of other prep, too. I used the stats available to us to calculate strikeout and walk percentages, and then K%-BB% for pitchers. We didn’t have these available in the projections, but I was able to accurately calculate them for hitters and cobble together a fairly good approximation for hurlers. As one of the most predictive single stats at which to look, I thought that this would help me to identify any potential bargains who were undervalued because of poor ratio or wins projections. The strikeout percentages weren’t as high as we’re used to, so I aimed to draft a whole staff with a strikeout percentage above 20 percent and targeted K%-BB% marks above 15 percent.
With the league being only 10 teams, I wasn’t too worried about positional scarcity. I figured that I could draft a strong projected team by ensuring that I nabbed a couple of premium power options, focusing on getting five or six of the starters who looked truly reliable for both ratios and volume, getting at least two closers without shopping in the bargain bin, and then filling out my offense with the plethora of late-round hitting options.
I drew the ninth pick, which wasn’t ideal from a projection standpoint. It meant that I missed out on the obvious top talent, who was none other than Wee Willie Keeler, immediately made the No. 1 pick by Mark. Willie was valued at over $47 with an incredible .305/.391/.513 line, 24 homers and 38 steals, plus a league-leading 110 runs. The next player was a full seven dollars below Keeler’s value. I also wouldn’t get the top pitcher, Howard Ehmke, projected for a ludicrous 2.39 ERA and 1.02 WHIP with a strikeout for each of his 214 innings plus a couple to spare. Jesse grabbed Ehmke with the third pick.
By the time my pick arrived, the best five-tool contributors had all been taken. The one option I felt would still give me an opportunity to set myself apart was one of those seven 30-plus homer hitters: Otto Knabe. He was projected for only a .255 average but with a top-five OPS of almost .900 and a league-leading 36 home runs—three more than the next-highest player. I liked the vote of confidence implied by the obvious outlier of this projection, perhaps a little too much, as it wouldn’t be the last time I let that determine a pick. I also considered whether punting average would be viable down the line.
I knew that the four elite pitching contributors I mentioned earlier would all be gone by the time the third rolled around, and Collin took Rasty Wright on the turn, the only high volume starter with a K/9 in the double-digits. Mike had also nabbed Fuller Thompson, the starter with the highest projected innings total at 227 2/3. I selected Bruce Hitt, who had an innings total just slightly below Thompson’s and a 2.94 ERA, with almost a strikeout per inning. The age tie-breaker came into play almost immediately, as Hitt and Dan Griner had almost identical value, but Hitt was in the 25-29 range and Griner was 30-34. Sure enough, Tim took Griner a few picks later.
As the draft unfolded, it became clear that people were not employing the same values or the same strategy as me. It was quite refreshing to be in a draft where a lot of people seemed to be doing something different. It meant that I was able to get the top-ranked second baseman at the end of the third, and a second of those seven 30-plus homer hitters early in the fourth.
It also meant that I was left staring at picks who my values suggested should have gone rounds before, weighing up the right time to finally pounce. One obvious example was Solly Hofman, an outfielder in the under-25 category projected for an incredible 71 steals. He also had a relatively low playing-time projection of just 487 plate appearances and clearly couldn’t hit, with a .628 OPS. However, those steals made him my 18th-ranked hitter and 25th-best player overall, and once again the sheer outlier appealed to me: PECOTA’s second-best steals candidate was at 50. I finally succumbed at the end of the seventh. This not only compensated for my first three hitters all being single-digit stolen-base guys, it put me in a very nice position to win both steals and runs outright.
The other extreme value ended up being the first closer off the board, Ed Appleton. I am not a player who drafts closers early, and I don’t think that I have ever taken one as early as I selected Appleton, at 52nd overall. He ranked 19th overall by my valuations and as the sixth-best pitcher, with a 1.62 ERA, 0.95 WHIP and 45 saves, plus 107 strikeouts in 67 2/3 innings. I couldn’t help myself. Even though relievers are volatile, I felt like this was the best chance of avoiding some of that volatility and owning one of the league’s best closers.
With the core set, I proceeded more or less to plan, making sure that I picked up plenty of volume without sacrificing quality—especially to compensate for Hofman—and avoiding hitters with poor plate discipline, while targeting the good K%-BB% pitchers. It was clear that I was drafting pitchers earlier than the other teams. I was tracking the standings as we went along, and while the team wasn’t set apart in ratios early, the strategy started to pay off later in the draft, as other teams lost ground with almost every new pitching pick they made.
My early sluggers and league-leading stolen-base pick meant that I wasn’t last in the offensive counting stats, despite prioritizing speed. While I hadn’t punted, the batting average had turned out to be a problem, but I was starting to move past other teams more easily in the other four categories as they turned to pitching.
The question was, of course, would it be enough? Would a projection victory translate into a real-life victory anyway? Find out next week.
Thank you for reading
This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.Subscribe now