Risk doesn’t really matter

Risk is one of the most overused words in fantasy baseball advice. Sometimes we’re told to seek out players with more risk so that we can harness their upsides. Other times, we’re told to play it safe and stay away from risk, particularly in the early rounds of draft (or for the high dollar players in auctions). There are indeed several minor strategies where it is best to minimize or maximize your risk exposure but only if risk-adjusting is low cost. But the truth is, properly understood, risk matters comparatively little.

Unfortunately, risk is often misunderstood. A player’s upsides and his downsides should all figure into his forecast performance. A mediocre player with a small chance of an all-star, breakout season will have (or should have) higher expectations than an identical mediocre player with no chance to breakout. This difference in forecast, average performance is important and valuable. That is, these players should have different dollar values.

But if a player has a higher upside but the same forecasts as another player, then the first player must also have a higher downside (to balance things out, note that, by “forecasts,” I mean their expected average results—the number you’d get if you looked at, say, their Oliver projections). So when an expert says that one player is more valuable than another because he has a higher upside, he must mean that this player is more valuable because he’s riskier. Otherwise it’d hardly be news that a particular player is more valuable simply because he has higher projected stats.

I would much rather have a player that is correctly valued using only three sets numbers: his forecasts given health (e.g. how well the pitcher pitches when he does actually pitch), how much time he’s expected to miss, and the forecasts for the replacement level player at his position. Forget about upside or downside.

Experts who do care about risk usually do so for one of three reasons:

—They are misunderstanding risk and they haven’t adjusted their forecasts for both upside and downside. For instance, forecast home runs should be adjusted for injury concerns. Once done, there is much less reason to double emphasize the injury concerns (or the upside).

—They are double emphasizing the risk of injury, for example, because of strategic concerns. Here’s where you need to be extra careful, since these strategies can only apply in narrow cases. There are only a few if any roster spots that are truly available to players with high upsides (a topic of mine in a past article). That is, there’s a cost to holding on to high upside players, waiting for them to break out—you have to forgo the opportunity to hold on to steadier reserves, a Jason Kubel type rather than a Michael Brantley.

—They are risk averse or risk seeking, excessively so. The risk from any one player contributes very little to the overall risk of your lineup. While it is true that the cost of risk for a first-round player is higher than for a fifth-round player, the difference in cost is or should be small.

Likewise, while you’ll need to make profits on your players in order to win your league, it really isn’t more effective to try to find large dollops of profit from a few players than it is to get small contributions from most of your players. You don’t need to start buying lottery tickets.

Making the case for ignoring risk is tough because the experts’ arguments are literally true without being practically useful. By all means, once you’ve got good, injury-adjusted forecasts, go ahead and contemplate risk-based strategies. But there is a risk worth considering first: the risk that you’ll overemphasize risk concerns on your draft day as much as the experts overemphasize them before.


7 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
AJK
13 years ago

Excellent article.  Completely logicaly and yet completely overlooked by most experts.

JoeC
13 years ago

But most forecasters don’t base their projections on someone getting injured (unless it’s someone who is ALWAYS injured… like, say, Nick Johnson). Most forecasters forecast players to get a full season’s at-bats. “This is what this player could do if they played a full season” is what they’re saying.

THEN the forecasters add on some type of injury risk grade for each player. So no double-counting on the risk factor.

Drew
13 years ago

This article was a good read, but its argument is more applicable to the early rounds of drafts. As the rounds go by, I still think I’d be more likely to take on additional risk. Considering the projections below and assuming all other stats are equal, who would you rather have?

Player A – 15 HR (upside of 17 HR, downside of 13 HR)

Player B – 15 HR (upside of 25 HR, downside of 5 HR)

I’d rather have Player B even though both players have the same forecasted stats. There will be players on the waiver wire – replacement level players – whose stats are only marginally worse than either of these players.  Consequently, I don’t have to participate in Player B’s downside. I can replace him with someone who will likely get me, say, 12 or 13 HRs. The waiver wire provides a natural hedge in case Player B turns out to be a bust. I can take a chance on getting the 25 HR upside without being fully exposed to the 5 HR downside.

Steve
13 years ago

The logic in this article holds in an environment where transactions are not allowed.  In a league with a waiver wire the downside risk is mitigated.

For example, Eric Bedard might be projected for 100 innings, but drafting Eric Bedard nets you, on average, 100 innings + 100 innings of replacement level pitching – DL spot.

Another example: Gordon Beckham’s range of projections ranges from All-Star to AAA.  If he is performing at AAA level, you can dump him for a replacement level player.  So his 10th, 20th, 30th percentile forecasts would actually be capped by replacement level for most of the season.

Mark
13 years ago

Gotta agree with Drew and Steve.  From a pure evaluation standpoint, this article is spot on, but there are loads of game theory quirks that apply to fantasy baseball, many of which involve not just evaluation of players and profit but of the tactical value of risk. 

Part of this is the utility of a concept like “replacement level”.  In evaluative terms, this concept is incredibly useful.  In practical terms, when you are actually deciding to replace a particular player with another one, the replacement level concept you used on draft day is almost meaningless.  You have to invent a whole secondary concept similar to replacement level (or at least some sort of baseline for evaluation) when looking strictly at players on the wire. 

My general theory is that there’s a reverse correlation between cost and the value of risk—that is the lower the cost the higher the value of risk.  You touch on this in the article, but I think you far underestimate its implications. 

Steve’s point is particularly poignant here.  The lower the cost, the greater value there is in risk, thanks to opportunity cost.  If the actual cost of a set of players is approaching zero, the opportunity cost of low risk players approaches fixed—it is easily determined and subject to a limited number of outcomes.  The opportunity cost of a high risk player can be “averaged” but to view it as fixed in this way is an exponentially larger simplification. 

My best example is from 2007.  I had a busted third base position on my roster (don’t remember exactly how) a few months into the season.  I had the choice of adding a single relatively safe bet type player with little upside/downside or dropping a second marginally useful player (again, don’t remember who), and thus slightly increasing my cost, to add two risk-heavy top prospects in Ryan Braun and Alex Gordon.  I chose the latter, and obviously adding Braun paid off while Gordon did not.  I was then able to drop Gordon and recoup most of the extra cost from dropping the second player initially, while reaping huge profit from Braun.  It was of course possible that both paid off and both busted as well, to many varying degrees, but thanks in large part to this gamble I took the crown in the league that year.  A similar situation played out with a team I had in 2009, where Lastings Milledge was the busted asset and Kendry Morales (OF eligible that year) paid a huge profit. 

Another issue I would raise is the difficulty in managing low-risk, mid value assets.  With high risk assets, it is relatively easy to cut your losses and move on when they bust.  With low risk assets, the worst case scenario is often that you wind up with a “bubble” type player who has no remaining downside, a bit of remaining upside, and a very difficult decision on how to handle that roster spot.  A good example of this was Pablo Sandoval last year.  He hit on the lower end of his projection, but it was incredibly difficult to cut ties with him entirely.  Someone who banked on a higher risk player such as Aramis Ramirez probably had a much easier time lucking into a huge profit from a Jose Bautista or Scott Rolen, and if they missed out there, at the very least someone like Chase Headley who incidentally gave you about what Sandoval did overall.  If you had Sandoval, third base was probably tied up far too long for owners to catch a legit surprise season for third base in the latter parts of the year.

Jonathan Halket
13 years ago

Hi guys,
Thanks for the comments.  I agree with all of them, in theory. But, Steve, I think managing risk is actually much harder than you let on.
-1) players that are high risk because of high upside generally have a small sample problem.  They isn’t a long track record of them actually performing at this level.  So, if and when they actually start displaying some of this upside, and you contemplate starting them and dropping another player, you could just as equally become the victim of mean reversion.
Consider Bautista – when exactly in the season did he go from “potentially mean reverting” to “surely maintaining the upside”? And in any case, he was someone who you’d probably start throughout the season in decently deep leagues anyway, so there’s very little option value to his upside anyway.
Likewise for playing the waiver wire and dropping players that display their “low side” – when can we rule out favorable mean reversion?  The difficultly of distinguishing “break (or bust) out” from “mean reversion” lowers the option value.
2) I completely agree that many forecasts don’t properly figure in injury potential.  I don’t consider figuring in this possibility “adjusting for risk” but rather adjusting for “events that may happen” – it is a delicate semantic difference. But some experts definitely double count the risk. Ron Shandler explicitly does it when he councils playing it safe in the early rounds of a draft.

Jake
13 years ago

Mark nailed it.  Great, great comment.

I was prepared to post something along the same lines, but he said it better than I could.