# Day-off risk vs. catastrophe risk

Lots of high-skilled players get big discounts due to injury concerns. Ben Sheets, Rich Harden, and Mike Lowell are just a few of the candidates for the sale rack. The big question is: Just how much of a discount should you apply to these types of players?

Some projection systems, like CHONE and Marcel, forecast playing time (in games or plate appearances). Expected playing time figures heavily in these systems’ forecasts for the players’ counting stats. The more the system expects the player to play, the more runs it expects the player to score (everything else equal, of course).

It is well known that, for fantasy purposes, one should add in the contributions of a replacement-level player when computing a player’s value. So, a player that is forecast to score 100 runs in 100 games is worth more than a player forecast to score 100 runs in 162 games. Why? Because you’ll be able to play a replacement player for at least some of the games that the 100-game player is forecast to miss. That player’s going to score some runs too.

This is one reason why “real-life” baseball valuation stats, like Wins Above Replacement, figure value relative to a replacement player. Of course, since replacement levels in fantasy are so dependent on how deep your league is (in number of teams, bench size and number of positions used), these kind of replacement-level calculations are usually left up to the user.

The temptation is to do something like the following: Let’s say the generic replacement-level player in your league is projected to earn 0.3 RBIs per game. Next, you take, say, CHONE’s forecast of 72 RBIs in 122 games for Mike Lowell. Then you compute the expected RBIs from drafting Lowell as 72 + .3 x (162-122) = 84.

Unless you’re in a daily league (and probably even then), it is unrealistic to expect to be able to replace Lowell in your lineup the minute he gets injured. So you’d like to apply a discount—maybe instead of assuming the replacement player plays 40 games, you assume he plays only 30 games for you.

I’d argue that the discount that you should apply to injury-concern players should vary a great deal depending on the player (and the injury). John Smoltz may be projected to start 20 games while Sheets is projected to start only 19, but I think you should apply a bigger discount to Smoltz, particularly in weekly leagues. Why? Because Smoltz is far more likely to have a start unexpectedly skipped, giving you a big fat zero in his spot for that week. Whereas Sheets is likely to be already comfortably nestled in your DL spot for the bulk of the starts that he might miss. Same thing applies for Chipper Jones versus Alfonso Soriano.

Pure speculation: If you’re looking for some hidden value, I think the players that are catastrophic injury concerns, like Sheets, may give you a little extra profit over players like Harden, who are more likely to have nagging injuries. I bet that lots of fantasy players discount too heavily players that are at risk for season-ending injuries versus players that are given lots of extra days off.

0000
« Previous: Top 10 prospects for 2010: Houston Astros and Arizona Diamondbacks
Next: This annotated week in baseball history: Jan. 24 – Jan. 30, 1813 »

1. Mike B said...

I like this approach to conceptualizing injury risk.

The fact is, “catastrophes” can’t be predicted, so inferring 120 games for a “catastrophe prone” player either dramatically overvalues or undervalues that player – it certainly does not “get it right.”

But we have a bit more predictive power over chronic or nagging injuries, and can more accurately guess the value of a Chipper Jones or Milton Bradley type.

To me, that means taking a chance on Ben Sheets is much more likely to pay dividends than on Rich Harden, given that on draft day Harden’s contribution is more likely to be properly valued.

2. sean said...

Is there a projection system widely recognized by the industry as the best at projecting AB totals?

3. Jonathan said...

Sean – I will do a little research on projecting AB totals.  My guess though, is that this is one area where a computer (that is, a formula based only on data like past ABs and age) is much worse than a human (armed with a computer and some formulas though).  Injury data is hard to systematically incorporate into a projection system. – though for guys who aren’t much injured yet, but are getting older – like perhaps Manny Ramirez.
In other words, a good regression based system probably would be helpful in forecasting older, previously rather healthy players but wouldn’t be very good at forecasting players who are risks for RE-injuring something (at least no system that I know of yet).

Mike B – much in that same spirit, I think we can forecast some injuries risk.  And also humans can figure in some re-injury risk and even perhaps classify that risk as catastrophic or not.  A Francisco Liriano type is more at risk for catastrophe, for instance…

4. Mike B said...

Jonathan,

Exactly.  It is precisely because we can categorize some players as “catastrophe prone” that the distinction exists.

My point is that for such players, projections taking that in to account will likely be fundamentally inaccurate.  Say we predict 20 starts for Liriano because of his past, and on draft day he is valued as a 20 start pitcher.  The likeliest scenarios during the season, though, are either (i) his arm flies off during one of his first starts, making his value less than zero given the draft pick you wasted, or (ii) he goes catastrophe-free and gives you full value, far exceeding the “average” between 0 or 100, weighted for catastrophe, that his ADP reflected.

Players with nagging injuries, however, can be better predicted.  If we expect Rich Harden to give us 20 starts, or 30 5-inning starts, given his track record, to me that seems to be a valuation more likely to be accurate.

We know Chipper Jones will give us 125, not 160, starts.  We do not know whether Nick Johnson will break another leg or three – but either he does or he doesn’t; and if he doesn’t, he is a bargain.

There is less upside to drafting a Harden or Jones, because their value (relative to injury risk) is better captured on draft day.  That said, there is less downside as well…

5. Mike said...

Rich Harden’s ADP is buried under such luminaries as Randy Wolf, Rick Porcello, and Tim Hudson. I’d say he might be a just slight bargain at that rank.

Curiously, Only Harden’s value is hurt by pitching for Texas, Neftali Feliz is still shiny and desirable.

Man, no wonder most people don’t win leagues!