Adjusting Fantasy Baseball Projections for a Delayed Opening Day

Ariel Cohen digs into to how COVID-19 will effect the 2020 fantasy baseball season. He looks at how projections and rankings should change, in light of MLB's delayed start date. See which players rise and fall as a result of coronavirus.

With a heavy heart I write this column. I truly wish that there would have not been a reason to write this column. Coronavirus (COVID-19) has now postponed the 2020 Major League Baseball season. In an effort to employ social distancing, local governments have made the tough call to forbid the gathering of large crowds. The NBA, NHL, PGA, XFL and NASCAR have all suspended their seasons. The NCAA tournaments have all been canceled.

And for us fantasy baseball players … major league baseball has pushed off the start of its season until least mid-May. It will likely be even longer. A lot longer. Many of our home fantasy leagues have now chosen to postpone our annual drafts or auctions. The NFBC has canceled or postponed many of its high money league contests.

But many leagues have still chosen to continue with their regular draft schedule in the coming weeks. Some of my home leagues will continue as scheduled. Just this past Sunday, I took part in the Tout Wars Head-to-Head points league. I understand the merits of postponement. It isn’t a regular year. I also understand the merits of continuing to draft this March. Grappling onto elements of normalcy has its benefits. Whichever group you're in, this article will address how to modify projections and expectations for 2020 given this past week’s rapid turn of events.

Risk Management in Fantasy Baseball

For those of you who are unfamiliar with professional background – I am a certified actuary. I am a Fellow of both the Casualty Actuarial Society and the Society of Actuaries.

In a nutshell, an actuary is an insurance professional who is well versed in the mathematics of insurance. Actuaries price insurance contracts, decide on the amount of reserves that companies need to hold for claims, inform management on the ROEs of various lines of business and deal with many other assorted insurance-related tasks.

One of my goals as a baseball writer and analyst is to bring concepts found in insurance and apply them to fantasy baseball.

At the top of my mind this week is the broad concept of risk. During a fantasy draft or auction, every time that a player is acquired, the owner takes on all of the player’s skills. Power hitters will offer the ability to hit home runs. Speedsters will offer a high stolen base output. Combo players will offer a blend of the two, etc.

Simultaneously, when a player is acquired - the owner also takes on all of the risks associated with the player. Some examples of player risk include (but are not limited to):

• Injury Risk
• Performance Risk
• Durability Risk
• Age Risk
• Experience Risk
• Regression Risk
• Playing Time Risk
• etc.

Every fantasy owner will acquire players with varying degrees of a number of these risks. There are no “sure things” in baseball. Nolan Arenado has been an amazingly stable fantasy asset over the past five years. He may exhibit a low amount of risk relative to others, but there are a multitude of possible scenarios where his final achieved value could be far lower than expectations.

A typical fantasy owner should acquire a few players who exhibit high risk. Rarely is a team devoid of high-risk players. Risky players often come with a cost discount at a draft/auction commensurate with the amount of risk that the player holds. It is the job of the fantasy owner to determine which risks are correctly accounted for by the market, which are overpriced, and which risks are accretive investments.

Generally speaking, drafted players are mainly uncorrelated. Their production (and hence value) is largely independent of one another.

Suppose that you have drafted Ozzie Albies and Nick Castellanos. An injury to Albies should have zero relationship to whether or not Castellanos gets injured. Or at least, that’s the way we draft. If I draft Giancarlo Stanton – whether or not he comes back fully recovered sooner or later should have zero correlation with Trey Mancini coming back sooner or later from his cancer surgery. [We hope they both heal soon.] We assume that all players, and (more importantly) all risk types are independent from one another.

We don’t want to draft too much injury risk in the aggregate on our teams, or too much of any risk for that matter. We want just enough to give us some upside – and hopefully, we buy those players at the appropriate discounts.

Injury Risk

Guess what we just learned? Injury risk is not independent. Injury risk was not independent for any 2020 draft that occurred prior to this week. Not even close.

The injury risk taken on by rostering Eugenio Suarez has every bit to do with the injury risk taken on by rostering Alex Verdugo. What do I mean by that?

Provided that baseball is played in 2020, every single player who is currently injured and was expected to return sometime in the middle of 2020 – will likely earn a profit! Every single discount that a player received in drafts, will either go away completely or be severely mitigated for any going-forward March drafts.

I acquired Aaron Judge for a mere $14 in the Mixed LABR auction two weeks ago. Unless Judge is out for an extended period of time, or requires season-ending surgery – in all likelihood, this will result in a profitable outcome for me. The bet that I made on Judge – acquiring him via his injury discount – will likely pay off. In the Mixed Tout Wars auction that occurred on Saturday, Judge was acquired for $23 by Gene McCaffrey. Though there are differences in league format and value between LABR and Tout Wars, much of the price difference between the two auctions arises because of COVID-19.

Every single injured player’s value has changed in the past week. From Mike Clevinger to Michael Conforto to Aaron Judge to Adalberto Mondesi to Giancarlo Stanton to Eugenio Suarez, etc. – every single injured player’s worth has been altered … and in the same direction: Up.

Injured players were correlated. Highly correlated. Injury risk was not independent.

Had a fantasy owner stocked his team with 15 of these injury risk players (acquired at discounts) – he would theoretically now have a large excess in team aggregate value. I happen to know and play with fantasy players who constantly accumulate many injured players at discounts. They typically hope that one or two players breakout and vastly outperform their acquisition price. 2020 would be the year of striking gold for such a drafter.

Yes, I do understand that I am talking about a team profiting from the COVID-19 pandemic. No one had that intention. Once again, it is a sad situation to be even talking about any positive effects on fantasy baseball from an awful widespread disease. But as a fantasy analyst, these are the facts and we need to contemplate how to go from here.

How to Adjust Projections

Now onto what to do going forward. What is the best way to update our own values/rankings of players to reflect the 2020 prospective landscape?

It all starts with projections. We need to alter/update our former projections for every single player in baseball – in order to allow us to properly evaluate their value change relative to the rest of the player pool.

To this end, I’d first like to classify all players into the following categories for today’s analysis:

• Injured Players
• Pitchers with Innings Caps
• Suspended Players
• Prospects who will be called up on a certain date
• Prospects with service time manipulation / who need more seasoning
• Standard Players - All Other Players

There are many ways to partition the list of MLB players, but for today - I would like to illustrate how to deal with the valuation of these classes of players.

Let’s start with the easiest group – the “All Other Players.”

Standard Players

First off, let’s choose an estimated start date that baseball will resume. We need to come up with a best guess of the percentage of games that will be missed during the season. Will the regular season be extended in October? We need to make an educated guess based on the latest information available to us.

For today’s analysis, I will assume that:

• June 1 will be the league’s start date.
• There will be no additional games played in October (to keep things relatively simple).
• Roughly one-third of the season will be missed. [I’m choosing to be optimistic.]

These are all assumptions that are almost impossible to predict. I won’t pretend to have a good idea of when MLB will commence. But let’s go with these assumptions today to help us run through the math of how to adjust projections.

Let’s call the percentage of the MLB season that will be missed the Missed%. The date that the season will be resumed will be referred to as the League Start Date. Numerically using our stated assumption:

Missed% = 33.3%

For now, let’s assume that rate stats will not differ. That is, we will assume that a player who was expected to hit one home run every 40 at-bats, will do exactly that in a shortened season.

The math goes as follows:

• New AB = Old AB * (1 – Missed%)
• New HR = Old AB * (1 – Missed%)
• New SB = Old SB * (1 – Missed%)

I prefer to refer to the formulae as follows:

• New AB = (1 – Missed%) * Old AB
• HR = HR/AB * AB [New or Old]
• SB = SB/AB * AB [New or Old]

Since HR & SB rates, etc. do not change post-translation – if we compute all counting stats as a function of playing time, we only need to adjust a player’s playing time. The key translation, and the math of our specific example:

• New AB or IP = (1 – Missed%) * Old AB or IP
• New AB or IP = 66.7% * Old AB or IP

Since we are essentially looking for a factor to apply to playing time, let’s refer to the multiplier as the Adjustment%.

• Adjustment% = (1 – Missed%)

For all rates, it should follow that:

• New BA = Old BA
• New OBP = Old OBP
• New SLG = Old SLG

There is no work to be done on any of the standard baseball averages.

Injured Players

As described above, all injured players will gain in value relative to the rest of the player pool. The reason stems from the fact that each projection considers zero production until the player has returned.

Let’s take Miles Mikolas. ATC projections previously projected the following:

The limited 115 innings projection was partially due to a forearm issue.

* Auction Values are per NFBC roster settings (15-team, 5x5 scoring)

At this time, I would like to give credit to Reuven Guy, who is my fantasy partner and is my co-host of the Great Fantasy Baseball Invitation Podcast – Beat the Shift. In real life, Reuven is an orthopedic PA. On twitter (@mlbinjuryguru), he is on top of player injuries and prognoses. Late last week, we went through all currently injured players – and determined each of their expected “comeback dates.” We put down an estimate of when they might be healthy enough to play in the major leagues.

Let’s also convert the Comeback Date into a percentage of the season that the player was supposed to miss due to injury. Here are some basic percentages that we can use:

InjuryMiss% will represent the portion of the 2020 season that each player was expected to initially miss.

The key to understanding how to adjust Mikolas’s projection (for example) is the following:

115 IP = 0 IP prior to Comeback Date + 115 IP after Comeback Date

There are now two possibilities for all injured players:

1) A player’s comeback date is set on or after the league start date. [InjuryMiss % >= Missed%]

In this scenario, all of the player’s innings will be as previously projected. If Mikolas was expected to return on June 15, for example – all 115 innings would still be projected for him. In this scenario:

• Adjustment% = 100%
• New IP = 100% * Old IP

There is a change to this type of player’s projection. Since the rest of the league has a 66.7% adjustment factor, the player becomes far more valuable. In fact, these players will become a whopping 50% more valuable relative to the standard player (using our original June 1 League Start Date assumption).

2) A player’s comeback date is set prior to the league start date. [InjuryMiss % < Missed%]

Let’s set Miles Mikolas’s comeback date at May 1, 2020. That translates to a 16.7% InjuryMiss%.

The adjustment math is as follows:

• Adjustment% = (1 - Missed%) / (1 - InjuryMiss%)
• New IP = (1 - Missed%) / (1 - InjuryMiss%) * Old IP

For Mikolas, his Adjustment% would be:

Adjustment% = (1 – 33.3%) / (1 – 16.7%) = 80.0%

The 80.0% is a higher factor than the common 66.7% factor, which gives Mikolas about a 20% relative value increase over the standard player. His new IP total goes to 92, with all rate stats following suit.

New Miles Mikolas Projections:

Suspended Players

All suspended players will now LOSE value relative to others. Suspended players still have to serve punishment for the same number of games, regardless of how many baseball contests end up taking place. If for example, all MLB games are canceled in 2020 - suspended players would then have to serve their sentence in 2021, etc. They would be worth zero in 2020.

If a player was previously projected to be suspended 50% of the season, he will now be suspended for a larger part of the season. Hence, he will lose value relative to the others.

Take Domingo German as an example. His suspension was set to expire roughly in the first week of June.

Similar to the injured players above, it is important to understand that all of German’s projected innings occur after his Reinstatement Date.

76 IP = 0 IP prior to Reinstatement Date + 76 IP after Reinstatement Date

Let’s set Domingo German’s reinstatement date at June 7, 2020. That translates to a 37.5% SuspensionMiss%.

The adjustment math is as follows:

• Adjustment% = (1 - SuspensionMiss% - Missed%) / (1 - SuspensionMiss%)
• New IP = (1 - SuspensionMiss% - Missed%) / (1 - SuspensionMiss%) * Old IP

The Adjustment% is of course, subject to a minimum 0% value – the smallest amount of time that a suspended player may be able to play in a season. For injuries, the Adjustment% was never less than the standard adjustment; for suspensions - it is no greater.

Domingo German’s Adjustment% would be:

Adjustment% = (1 – 37.5% - 33.3%) / (1 – 37.5%) = 46.7%

Pitchers with Innings Caps

There are a number of starting pitchers in the major leagues that were not projected for a full season’s worth of volume due to innings cap impositions. These pitchers might be rookies/sophomores who have not yet pitched a full season, or perhaps they might be recovering from Tommy John surgery or other health issues.

Let’s take A.J. Puk for example.

Projected innings for this class of player look like the following:

121 IP = 121 IP prior to Innings Cap Limit + 0 IP after Innings Cap Limit

Projections have assumed that all of the innings for A.J. Puk are frontloaded. When the season finally commences, Puk would pitch the first 121 innings, and would stop pitching thereafter.

For pitchers with innings caps, I introduce one more intermediate (but complicated) step into the process.

Without going into too many details (as this isn’t our main focus for the day) - I first generate the percentage of capacity innings that a pitcher would pitch in a full season. This is either based on organizational depth charts, on IP per start metrics, or based on skills, etc. Most starting pitchers will generally project out to 80-95% of possible capacity. Assuming that capacity is 200 IP these days, fully healthy pitchers are generally projected between 160-190 innings.

For A.J. Puk, I set his Capacity% to 80% (he would be a ~160 innings pitcher this season without team imposed innings limits).

The adjustment math is as follows:

• Adjustment% = Capacity% * (1 - Missed%) / (IP / 200)
• New IP = Capacity% * (1 - Missed%) / (IP / 200) * Old IP

The Adjustment% is subject to a minimum 0% value, and a 100% maximum value.

For Puk:

Adjustment% = 80% * (1 - 33.3%) / (121 / 200) = 88.0%

Prospects

Dealing with prospects is more difficult. There is no one-size-fits-all rule. Earlier in this article, I identified two classes of prospects:

• Prospects who will be called up on a certain date
• Prospects with service time manipulation / who need more seasoning

Prospects who will be called up on a certain date

As captioned. Treat these prospects as injured players. Assume a prior “call-up” date and substitute it as an injury return date. You will then be able to use the formulae above for injured players.

Prospects with service time manipulation / who need more seasoning

These are prospects whose organization has a monetary reason for holding the player back for some time. Perhaps keeping a prospect down some 25 days is needed to garner a year more of team control, or to push off arbitration by one season. Whatever the case may be – these prospects will be in the minors for some (fixed) set of time before coming up to the majors. Or, some prospects simply need an additional amount of fixed time in the minor leagues in order to hone their skills. No matter what the actual start date of the MiLB season, these are the players who will stay in the minors for a select time period.

Treat these prospects as suspended players. Assume a prior “call-up” date, substitute it as a reinstatement date. You will then be able to use the formulae above for suspended players.

There are many other possibilities of how one can model prospect adjustment. Hopefully, these two basic examples will cover most cases for you.

Limitations / Notes / Future Enhancements

In no particular order, here are a few of the limitations/simplifying assumptions of today’s article, as well as some possible variants or future enhancements:

• The model above assumed that the season would end after September as usual. It is highly possible that the 2020 regular season (if played) would be extended well into October. That would shift/change some of the precise formulae.
• We assumed a deterministic/static starting date for the opening day of baseball in 2020. Of course, at this point in time, we do not have a firm grasp on it. An actuarial model would pick a midpoint, assume the starting point to be variable, and run the playing time with stochastic simulations.
• I refer to playing time in the above as at-bats for hitters. More precisely, one should be using plate appearances instead of pure at-bats.
• I refer to IP in many of the adjustment equations above. You may substitute in plate appearances for the equivalent hitter formula.
• One class of player not covered is the players who would be “losing their roles.” Whoever is the 5^th starter in St. Louis might end up losing playing time once Miles Mikolas comes back from injury, etc. One might want to consider adjusting their projections downwards as well.
• We assumed above that homerun rates (or stolen base rates) would be identical if the season started at any point in the year. Of course, that isn’t exactly true. HRs grow in the heat of the summer, and SB are quite variable from month to month throughout the season. Some additional adjustments would be needed to be more precise.

Conclusion

Hopefully, you have gained a further understanding of the types of players requiring a rankings change - which could be either up or down. Should you participate in drafts in the next few days, consider modifying your strike price for players as described above. I have guided you on the mathematics of how to adjust playing time, should you be inclined to do so on your own.

We have also seen today that injury risk is not independent from player to player. There always exists the possibility that an entire class of risk will be devalued or may collectively appreciate in a single moment. This is a crucial concept to understand. It arises far more in life than you might imagine.

Finally, I want to aid you in knowing just how much a player's value has been translated. Below is a listing of injured, suspended and innings cap players with adjusted valuations. I provide a pre and post NFBC (15-team, 5x5 roto) value for affected players.

Risers and Fallers, Assuming a June 1 Opening Day

I wish you all safety and health in these difficult times. Hoping that baseball is back real soon …

More 2020 Fantasy Baseball Advice