Bartolo Colon's xERA

[BLMJ]

As of 5/26 (before his start today is included), Projected Total xERA seems to have a bug:

Actual to Date: 4.06
Projected Bal: 4.66
Projected Total: 3.18

I noticed this earlier in the season with other players, so if it's already been discussed then disregard this. Just tryin to help.

[RobR]

That was fixed this morning.

[BLMJ]

Well it's changed, but I'm not sure it's fixed:

To date: 4.40
Balance: 4.58
Total: 4.18

Although, having picked him up the lower the better!

[RobR]

That's OK. Due to the data that's available, we use different versions of the xERA formula for projected data and YTD data, so the Total number won't always fall in between the YTD and projected numbers. You might see similar behavior on PX and SX.

[BLMJ]

I'm confused...

I gather that there is a formula used for YTD xERA for small sample sizes and then a different formula used for Bal and Total xERA. How they don't all add up doesn't make sense.

Would it be correct to say that applying the balance of season formula to the YTD stats the xERA would be well below 4 (which is how the total value is 4.18)? If that's true, I guess I'm not sure why you'd want to show his xERA being 4.40 as that seems misleading. Either he's done well enough that his current xERA is bringing down his total (from the Bal projection of 4.58) or he hasn't and his total should be well above 4.18

Like I said, I'm confused

[RAY@HQ]

It's not a sample size issue, there are some subtle differences in the xERA formula used to calculate YTD xERA v. Proj xERA. These are due to data limitations... to use the same formula for proj xERA that we use for YTD, we would need to project some information that we don't project.

RobR can explain in more detail, I know he's been looking at this stuff very recently.

[BLMJ]

I can understand there are potential reasons to use different formulas that can certainly be justified. I don't understand why the total doesn't equal the weighted sum of the two. Why doesn't the calculated total use the YTD xERA listed on the player page?

If there is a strong justification for not using the YTD xERA shown and that there is a better way of incorporating the YTD stats into the total projection, then isn't that the xERA that should be listed on the player page?

If we were talking a small difference then ok. With 12 IP pitched, 48 more projected at an xERA of 4.78 and a total xERA of 4.33, the formula will need to be using a YTD xERA of 2.53. Thats 1.87 runs lower than the YTD xERA shown (4.40). Doesn't add up to me.

[BLMJ]

JP Howell actually has slightly better BPI projections for Bal than he does for his total, but his Total xERA is lower.

YTD: xERA 3.74
Balance: Dom 8.4, CMD 3.0, HR/9 0.9, BPV 89, xERA 3.88
Total: Dom 8.3, CMD 2.7, HR/9 0.9, BPV 87, xERA 3.37

Could it be the normalizing factor for Balance and Total are not the same?

[RobR]

Not ignoring this, just haven't had time to compose an answer yet.

[BLMJ]

No problem, thanks again for taking the time

[RobR]

It's not abnormal for there to be slight disconnect between the "Actual to Date", "Projected Bal" and "Projected Tot" xERA. Because we don't have the same data sets available for actual and projected stats, we use different formulas. This would mean that identical Actual or Projected performances would yield different xERA. For this same reason, the Total xERA won't always fall in between the Actual and Projected xERA. Once we have enough sample it shouldn't ever be that far off either.

[BLMJ]

I guess I'm not being clear... I understand you'll use different formulas for projected and actual. I accept that and am not concerned with the differences in the two.

Regardless of how those two components are calculated, isn't the total calculated by:

Total = (a*Actual + b*Balance) / (a+b)

where a and b are the IP's for each. The answer has to be that this is not the process used or else there would be no discrepancy. My question is then why not, and what is the process?

[RobR]

xERA is a formula. It is fed the input stat parameters for the period of interest. So we pass YTD stats to one formula. We pass projected stats to a second formula. We pass the sum of YTD + Projected stats to the second formula for Total.

YTD xERA = f(x, y, z, ...)
Projected xERA = g(a_ytd, b_b_ytd, c_ytd, ...)
Total xera = g(a_ytd + a_proj, b_ytd + b_proj, c_ytd + c_proj, ...)

[BLMJ]

Interesting... I would not have thought to do it that way but I'm sure you have your reasons.

Thanks for the responses.