New Way of Doing Projections?

[helloneumn]

Is it just me or has HQ changed the way they are doing mid-season projections? In the past, it seemed to me, especially early in the year, that the "Projected Total" stats did not really change all that much from the Pre-Season projections. The "Projected Balance" stats were adjusted to incorporate the current stats so that the Total stats were relatively unchanged. This year, it seems that early stats are given more weight and that the "Projected Balance" stats basically are a percentage of the Pre-Season Projection.

For example, the 3/24 projection for Pujols was 43 HR & 117 RBI. Now, after his fast start, his projection is 52 HR & 134 RBI. Seems like a big jump based on a small sample size. Now, you might say "Well, that's just Pujols. Maybe he's kicked it into another power gear that we all thought he had." Okay, well then how about Chris Shelton? His preseason projection was 21-60 and now its 31-78! It just seems like HQ has always been slow to change final season projections based on small sample sizes and now they can't change them quick enough.

Has the basic philosopy changed or was I just imagining that was the way it was?

[RON@HQ]

We tinker with the projections algorithm all the time. WIth the launch of our new player database, and daily projections, it afforded us more flexibility in the process as well.

Even still, both Shelton and Pujols got off to such enormous starts that their numbers HAVE to impact their full season expectation. If we stuck with Shelton's original HR projection, that would mean that we'd only expect him to hit another 10-12 HRs over the next 5 months, which is unrealistic.

[helloneumn]

Even still, both Shelton and Pujols got off to such enormous starts that their numbers HAVE to impact their full season expectation.

OK. That makes sense with the counting statistics. But what about things like ERA and WHIP? Or even underlying BPIs?

Take a guy like Chris Capuano who is also off to an "enormous" start. His pre-season projections were 4.30 ERA, 1.42 WHIP, 7.1 K/9, 3.7 BB/9. Now, his projections stand at 4.04 ERA, 1.33 WHIP, 7.7 K/9, 3.55 BB/9. It seems to me that, in this case (as well as most other players I've noticed), you've taken his current performance, added it to a pro-rated set of his pre-season projections to come up with this new projection. This seems especially odd in ERA and WHIP categories since they are probably being affected by a low 23% hit rate at this point.

It works in reverse too. Look at Matt Cain who HQ was high on at the beginning of the year. Pre-season projection = 3.77 ERA, 1.23 WHIP, 8.3 K/9, 3.66 BB/9. Current projection = 4.11 ERA, 1.26 WHIP, 8.1 K/9, 3.5 BB/9. Not only that but his projected balance projected ERA (3.88) is higher than the preseason.

It just seems really early for you guys to have "admitted" that you were off. I would have expected your "position" to be "Its early and while he's performed well to this point, we expect that to even out for the remainder of the season."

[Markus_ls]

We tinker with the projections algorithm all the time. WIth the launch of our new player database, and daily projections, it afforded us more flexibility in the process as well.

Even still, both Shelton and Pujols got off to such enormous starts that their numbers HAVE to impact their full season expectation. If we stuck with Shelton's original HR projection, that would mean that we'd only expect him to hit another 10-12 HRs over the next 5 months, which is unrealistic.

With regards to Shelton I was thinking about Jorge Posada (hmm, maybe 2004?). I don't have easy access to monthly splits for past seasons but I remember Posada having like 8 or 9 HR in April but still finishing with only low 20's - which was in line with his preseason projections.

So I'm not totally convinced Shelton's numbers HAVE to go up. (I'd say they will...but less than HQ seems to feel.)


Helloneumn's question reminds me of the Gambler's Fallacy Thread a few years ago. I read through as much of it as I could stand before my head exploded but never seemed to reach a consensus on whether hot starts should be ignored as normal variation or whether they suggest long term differences (Fallacy hypothesis)?

I DON'T want to rehash that thread (my head in nicely repacked now) but was there a final conclusion? It seems HQ is saying it is not a fallacy to change longterm expected values based on short term anomaly's, based on current projections???

[Michael@HQ]

Mark,

That was a long and horrible thread. It was 2005 and concerned Johnny Damon's batting average but quickly got to the more theoretical issue of whether the regression to the mean effect meant (i) a player would do worse than his pre-season projection the rest of the way or (ii) a player would do the same has his pre-season projection the rest of the way. The issue wasn't clearly defined and it did not produce a consensus.

[NICK@HQ]

If a player exceeds expectations in the early going, and his BPIs support this new level, you probably should adjust.

If a player exceeds expectations in the early going, and his BPIs do NOT support this new level, you should stick to the preseason projections.

It may be as simple as that.

[Markus_ls]

If a player exceeds expectations in the early going, and his BPIs support this new level, you probably should adjust.

If a player exceeds expectations in the early going, and his BPIs do NOT support this new level, you should stick to the preseason projections.

It may be as simple as that.

Now THAT I can handle!

[Markus_ls]

Mark,

That was a long and horrible thread.

That's a good description for some of my past rotisserie seasons!