CSV Download Files - Missing Variables & Changes - Please Change Back

[Timmeh49]

I know you HQ guys will get this figured out eventually... However, I spent some time trying to deduce what might be messed up.

And as best as I can tell, of the raw stats (PA, AB, R, H, 2B, 3B, HR, R, RBI, BB, K, SB, CS), the only one that is messed up, I think, is AB. In particular, the PA field looks pretty normal. You guys seem to assign players discrete PA totals in multiples of 34 (66 players have 34 PA, 46 have 68 PA, etc.). Using the given AB field, here's a plot of the composite AVG at each PA level in the 2024 file (downloaded earlier today), along with the composite average from the Mar 19, 2022 file for the same series of PA levels.



Weird, right? The players given the least playing time have the highest AVG in the 2024 file, whereas for the 2022 file, AVG increases with PA level. FWIW, other AB-based ratios (CT%, SLG, etc.) show similar trends: the best performers on a rate basis in 2024 are the ones projected for the fewest PA.

Then I thought... what if AB is the messed up field. So I decided to recalculate AB as simply the listed PA minus the listed BB, and then calculate AVG for each PA tier with this new AB definition. It looks alot better.



Again, I have faith you guys will find the errors and fix everything, but it does look to me that the AB numbers are wrong, and then other errors seems to flow from that.

[viper]

PA is more than AB plus BB. You also must include HBP, Sac Flies & Sac Bunts (all not available in BBHQ). My league leans heavily on PA and BBHQ never had those projections in the past. And they still don't.

Consider the case of Yordan Alverez. CSV downloads has him at 612PA, 537AB,157H, 292BA and 81BB. His BA is exactly 157/537 so one must assume his AB is correct. Now 537AB+81BB equals a minimum on 618PA. Whoops! That is 6 over the projected PA (much more than any potential rounding error). And we still haven't added in the approximate 14 HBP+total SACs that other sites project. Alverez is not an isolated case.

I would probably trust all the numbers they show but not PA in the slightest. For 99.9% of fantasy leagues, AB is all that really matters. My league is a lone wolf so I cannot use any BBHQ hitting stats in my preparation. Equating the hitting stats to the pictures in a Playboy magazine, I bought Playboy just for the articles - didn't we all. The BBHQ articles are the best you can find

[blueboy714]

Yes, send any refund requests to support@baseballhq.com and I'll help you out there.

The sure amount of these daily changes to the CSV download files is unbearable. I am going to be request a refund. Totally disappointed this year with the quality of BHQ's work. I have been a subscriber to BHQ for close to 3 decades and buying Ron's books well before that. I've even attended multiple First Pitch Forums - but this has been totally unacceptable.


Thanks.

[RAY@HQ]

I know you HQ guys will get this figured out eventually... However, I spent some time trying to deduce what might be messed up.

And as best as I can tell, of the raw stats (PA, AB, R, H, 2B, 3B, HR, R, RBI, BB, K, SB, CS), the only one that is messed up, I think, is AB. In particular, the PA field looks pretty normal. You guys seem to assign players discrete PA totals in multiples of 34 (66 players have 34 PA, 46 have 68 PA, etc.). Using the given AB field, here's a plot of the composite AVG at each PA level in the 2024 file (downloaded earlier today), along with the composite average from the Mar 19, 2022 file for the same series of PA levels.



Weird, right? The players given the least playing time have the highest AVG in the 2024 file, whereas for the 2022 file, AVG increases with PA level. FWIW, other AB-based ratios (CT%, SLG, etc.) show similar trends: the best performers on a rate basis in 2024 are the ones projected for the fewest PA.

Then I thought... what if AB is the messed up field. So I decided to recalculate AB as simply the listed PA minus the listed BB, and then calculate AVG for each PA tier with this new AB definition. It looks alot better.



Again, I have faith you guys will find the errors and fix everything, but it does look to me that the AB numbers are wrong, and then other errors seems to flow from that.

I was on a call with the programmers this AM about this issue. I think you're on the right track here.

[Timmeh49]

I was on a call with the programmers this AM about this issue. I think you're on the right track here.

Thanks Ray. It sounds like this will be straightened out soon.

[circle360]

I know you HQ guys will get this figured out eventually... However, I spent some time trying to deduce what might be messed up.

And as best as I can tell, of the raw stats (PA, AB, R, H, 2B, 3B, HR, R, RBI, BB, K, SB, CS), the only one that is messed up, I think, is AB. In particular, the PA field looks pretty normal. You guys seem to assign players discrete PA totals in multiples of 34 (66 players have 34 PA, 46 have 68 PA, etc.). Using the given AB field, here's a plot of the composite AVG at each PA level in the 2024 file (downloaded earlier today), along with the composite average from the Mar 19, 2022 file for the same series of PA levels.



Weird, right? The players given the least playing time have the highest AVG in the 2024 file, whereas for the 2022 file, AVG increases with PA level. FWIW, other AB-based ratios (CT%, SLG, etc.) show similar trends: the best performers on a rate basis in 2024 are the ones projected for the fewest PA.

Then I thought... what if AB is the messed up field. So I decided to recalculate AB as simply the listed PA minus the listed BB, and then calculate AVG for each PA tier with this new AB definition. It looks alot better.



Again, I have faith you guys will find the errors and fix everything, but it does look to me that the AB numbers are wrong, and then other errors seems to flow from that.

Good Stuff

You are like me

I was doing that the other day too.

I like puzzles. So it makes sense

Im also with you.

Ive been a customer for 2+ decades.

I'll give Ray and company the benefit of the doubt that it will get figured out. I know some people will complain. But this stuff happens al the time in different fields. See cable going out. Something happens to your router and wifi. etc etc

[Timmeh49]

Good Stuff

You are like me

I was doing that the other day too.

I like puzzles. So it makes sense

Im also with you.

Ive been a customer for 2+ decades.

I'll give Ray and company the benefit of the doubt that it will get figured out. I know some people will complain. But this stuff happens al the time in different fields. See cable going out. Something happens to your router and wifi. etc etc

+1

Like you and others wrote, there's only so much that can be done via internal testing. I know there is a tendency for us users to think "how they have missed that". But testing resources are limited. In the end, the best testers are us.

[RAY@HQ]

Posted an update on CSV files in today's site update. Other than the issue discussed immediately above (PA/AB/low-PT guys with weird stats), I think we're back to having a stable and usable set of CSVs. Unless there's a really good reason, I'm going to try to leave these files alone in terms of structure/column layout, at least til after Opening Day. Then, after everyone's gone through a cycle with them, we can assess another round of changes.

[RAY@HQ]

I know you HQ guys will get this figured out eventually... However, I spent some time trying to deduce what might be messed up.

And as best as I can tell, of the raw stats (PA, AB, R, H, 2B, 3B, HR, R, RBI, BB, K, SB, CS), the only one that is messed up, I think, is AB. In particular, the PA field looks pretty normal. You guys seem to assign players discrete PA totals in multiples of 34 (66 players have 34 PA, 46 have 68 PA, etc.). Using the given AB field, here's a plot of the composite AVG at each PA level in the 2024 file (downloaded earlier today), along with the composite average from the Mar 19, 2022 file for the same series of PA levels.



Weird, right? The players given the least playing time have the highest AVG in the 2024 file, whereas for the 2022 file, AVG increases with PA level. FWIW, other AB-based ratios (CT%, SLG, etc.) show similar trends: the best performers on a rate basis in 2024 are the ones projected for the fewest PA.

Then I thought... what if AB is the messed up field. So I decided to recalculate AB as simply the listed PA minus the listed BB, and then calculate AVG for each PA tier with this new AB definition. It looks alot better.



Again, I have faith you guys will find the errors and fix everything, but it does look to me that the AB numbers are wrong, and then other errors seems to flow from that.

Just to confirm you were on the right track here.... this is now fixed, and here's the explanation from the developer:

In the calculation for projected at bats (plate apps - walks) I'd incorrectly used the baseline version of walks, instead of the version adjusted by plate appearances. So the walks were too high, causing at bats to be too low, which was throwing other calculations off

With that, I think the CSVs are in a good and stable place.

[Snider]

Ray - Can you help me understand why for example the first 3 hitters listed in alphabetical order - CJ Abrams. Jose Abreu, and Wilyer Abreu have slightly different at bats and projections today versus just a day or two ago with the exact same playing time projections? Not a criticism but just a question and me just trying to understand the theory behind the projections and the files. In prior years unless something unforeseen - stable playing time meant stable projections. Thank you

[RAY@HQ]

Ray - Can you help me understand why for example the first 3 hitters listed in alphabetical order - CJ Abrams. Jose Abreu, and Wilyer Abreu have slightly different at bats and projections today versus just a day or two ago with the exact same playing time projections? Not a criticism but just a question and me just trying to understand the theory behind the projections and the files. In prior years unless something unforeseen - stable playing time meant stable projections. Thank you

That change would be related to the issue I just posted about immediately above... that issue affected all batters. It was less visible in guys who had near-full-time PA projections, but likely still had an impact.

Someone else (Viper, I think) had pointed out recently that projected PA was differing from (projected AB+BB). That was also the same issue. PA ties out to AB+BB now for all batters. (As he noted, we don't project the other minor components of PA)

With that issue solved, the CSVs should be stable going forward.

[Timmeh49]

Just to confirm you were on the right track here.... this is now fixed, and here's the explanation from the developer:

In the calculation for projected at bats (plate apps - walks) I'd incorrectly used the baseline version of walks, instead of the version adjusted by plate appearances. So the walks were too high, causing at bats to be too low, which was throwing other calculations off.

With that, I think the CSVs are in a good and stable place.

Thanks, Ray. And nice to know I was on the right track.

But I guess I'm going to be "that guy" and say... There still seems to be something a little bit off with the low PA guys. Now, I think some of them are too low...

For example, Christian Arroyo is projected for a slash line of .103/.139/.134 in 102 PA. This seems awfully pessimistic for a 28 y/o with a career line of .251/.298/.392 in 995 PA.

Having a reasonable projection -- even for these lower-ish PA guys -- is important for me (and probably others), because I also calculate a "full time" dollar value (scale up to 550 PA for players projected to fewer than 550) for all players. This helps me identify end-game purchases or reserve picks.

So I'm thinking there might still be something amiss.

Thanks for considering.

[RAY@HQ]

But I guess I'm going to be "that guy" and say... There still seems to be something a little bit off with the low PA guys. Now, I think some of them are too low...

For sure, a fair point. And it allows me to explain a little more of the projection process.

That .104 BA for Arroyo is in fact, exactly the baseline projection I uploaded to the site for him. So tha'ts not a site issue, that's a projection adjustment issue. These site/CSV calc issues have obscured the projections issues, so I couldn't separate those out. Now that the CSVs are clean, it's easier to spot issues like Arroyo.

Now, truth be told, I may not clean up every projection like Arroyo right away, as long as he stays at 5% projected playing time. But, if it gets to be mid-March and it looks like he might make the Opening Day roster, and our MIL PT analyst bumps that 5% to 30%... then you'll see that projection cleaned up pretty quickly.

Everything I wrote in the last two grafs is normal pre-season projection maintenance for me. That's the bottom line here... the process should be normal going forward.

[Red]

The CSV files downloaded from "STATS/2024 Projections" menu do not match the CSV Center download. The MLBAMID column is one of the columns missing missing from the "STATS/2024 Projections" spreadsheet. Have not studied to see if other data is different or missing.

[RAY@HQ]

Yes, the formats are known to be different. As we posted in the explanation on the Projections page, that page is really meant for ad-hoc searches, think of the HTML version of the projections on the old site. If you want a CSV download, you're better off with the ones in the Download Center as opposed to the CSV button on the projections page. (The exception would be if you've created a specific search/filter or something on the projections page and want to download that result.)

In terms of the data in the tables, they should match up... in most cases. The exception being that the CSV download center only refreshes once daily, where if I happen to run the projections process by hand during the day, the projections page will pick up those changes immediately (as will the Custom Draft Guide, Playerlink, etc)