Since “Moneyball” brought the use of data-driven baseball analytics into major league front offices, the projection systems for both the public and the ones that create internal forecasts for teams to use have benefited from one thing above all else: consistent data.
College players always had a certain number of plate appearances or innings pitched. The five-month minor league season provided similarly static baselines for prospects, and the 162-game major league season did the same at that level.
But all the changes to the 2020 baseball calendar at every level, plus the specter of more changes in 2021 as the coronavirus pandemic continues, have left teams trying to fill that data void that Orioles assistant general manager for analytics Sig Mejdal referred to as “the fuel to that projection engine.”
“The impact of the missed 2020 season is going to be around for quite a while,” Mejdal said. “I think we’ve been spoiled that our models are built on past data, of course, which year after year has been a very similar condition. Players play entire seasons, get a full amount of at-bats, and that is fuel for the projection engine and now we have the majority of professional players missing an entire year — and it has nothing to do with injuries.
“As long as 2020 will be in the projection system, we’re going to have to take care to adjust and consider the fact that these guys with zero at-bats can’t be treated like past players who had zero at-bats in a year. That’s going to be around for a long time.”
For a team like the Orioles, who under executive vice president and general manager Mike Elias and Mejdal have used a preexisting (though now significantly upgraded) projection system named “Omar” to help guide decision making since they were hired in November 2018, that has meant different adjustments at different stages. All of those, Mejdal said, are works in progress.
When the team needed to rely on barely a month of early-season games for scouting and data purposes for college prospects for the five-round MLB draft in June, domestic scouting supervisor Brad Ciolek said the Orioles had to weigh projections on what the rest of their seasons would look like “pretty heavily” when making decisions.
The professional forecasts impact major and minor league free agency and trades, with the recent Rule 5 draft an example of how they had to operate without the 2020 season’s data to rank their players.
“We’ve been baking in greater variability,” director of pro scouting Mike Snyder said after the Rule 5 draft last month. “Next season, I anticipate we’ll probably be surprised more often than we are in a normal year, where players might make a positive leap that the league as a whole didn’t necessarily see signals of happening.”
Michael Weis handles the amateur forecasts for draft preparation, while Ryan Hardin is responsible for the professional projections, Mejdal said. On the pro side, forecasts for both the Orioles’ players and those on other clubs need to factor in a year with no minor league statistics. In both instances, it’s a matter of how those gaps in the data are filled in, Mejdal said.
One plate appearance at any level, Mejdal said, can be “revealing in a tiny way [of] your underlying skill,” and it’s incumbent on those who use that piece of information to get the most they can out of it.
Baltimore Orioles Insider
“Whether you fill it in and keep your old model or scrap the model or create a new one, it’s still the same — you still have A, B, and C,” Mejdal said. “You’re just determining how you weigh, and how A, B and C interact.”
In each instance, Mejdal said there’s no best way to move the balances to spit out a particular outcome.
“It’s changed, so we have to change,” Mejdal said. “This has never taken place before, so there has to be some amount of educated guesswork, but we still can do our best to remain as data-driven as we can and adjust our models as appropriately as we can.”
Almost a year of working under such conditions doesn’t mean they’re solved, though. And conditions for all of baseball aren’t necessarily settled yet. Colleges could still play fewer games even as many leagues plan to continue their normal preparations for the 2021 spring season. According to Baseball America, the start of the Double-A and Class-A seasons in the minors could be delayed by at least a month and possibly more.
Every time such a change is reported, the Orioles’ analytics team works on their pivot.
“When you see the changes taking place or announced, their attention immediately goes to how will this affect the model and what do I need to do to change the model to accommodate for this change?” Mejdal said. “It’s a model, so there’s going to be subjective decisions that you make in the creation of the model. Like so many of the model builds, we will get together and we’ll discuss it and come up with what we think are a few of the better ways of structuring the model, and then we’ll let the past data create the parameters within that model.
“I want to give all the credit to the analysts. They’re immediately on this, and they’re actually already creating models for different scenarios that are going to take place in 2021 or that might take place in 2021, both for college and professional baseball because obviously, we still don’t know what baseball is going to look like in 2021, both amateur and professional.”