The Weighted Means Spreadsheet, PFM, and depth charts were updated Monday morning. Here are some of the fixes and improvements:
- you’ll see two new tabs in the Weighted Means Spreadsheet–it now has both playing time projected and PECOTA raw projections.
- a lot of work has gone into the translation postprocessing of the PECOTA data, and GB%, BABIP, and defensive projections appear to have been significantly improved. There are still individual players that have interesting results, which we’ll continue to look at.
- a clarification: the column in the weighted means spreadsheet GB Out %, not GB%–that is, GBO/(GBO+AO), not GB/BIP.*
- four comparable players are now available, and their names are formatted to be easier to read.
- players have been added. Please comment on additional players we’re missing and we’ll get them in there too.
- player R and RBI now scale to the team run environment.
- a few pitchers have been assigned >30 starts, which was treated as a hard cap in previous runs.
We’ll keep you posted about additional updates… please let us know what else you are seeing in the comments.
* We would prefer GB/BIP, but for large chunks of our data, we don’t have GB info on hits, so we don’t have GB/BIP information. We can get GO data for a lot more pitchers than we can get GB data - so that’s what’s used. Once you get past pitchers with < 10 IP, the correlation between GBO% and GB% is about 0.96--pretty close to identical.
Thanks, Dave. Just scrolling through the spreadsheets a little bit, it appears that you guys are projecting a high run environment? Many of the position players seem projected for big years relative to some other systems (e.g., CHONE).
Are those actual numbers for the 438 projected players, or for the league as a whole? I suspect that the 438 projected players would be expected (on average) to outperform the league average since the league-wide figures would include replacement players.
This may be a dumb question, but can someone clarify precisely the difference between "playing time projected" and "PECOTA raw projections," and how the adjustment is made?
I think it means, for instance, that PECOTA might project promising young 1B Joe Shlabotnik to have 600 PA in 2010 with a 20.0 VORP in the raw projections. But "playing time projected" is adjusted by human intervention, and because Shlabotnik is actually playing for St. Louis behind Albert Pujols, there's no way he'll get 600 PA, so he's adjusted down to something more probable?
Can you explain or point me to an explanation of the environment in which PECOTA figures its "raw" projections? Exactly how raw are we talking here?
Are these projections against league-average opponents and do they account for league and divisional difficulty, park factor, etc.? I'm guessing every player gets the same "neutral" context in the raw projections and then gets expected actual context in the playing-time projected set, but please correct me if I'm wrong.
As far as I know, all of the projections are done in "translated space" where performance at every level has been translated and normalized. Those are expressed in EqBA/EqOBP/EqSLG form. They're then translated to the league- and park-specific environment of the big-league team, so those triple-slash numbers move a bit.
The playing-time specific projections are based on our current assumptions about team rosters and lineups. They're subject to constant revision as events unfold in the spring - injuries, callups, job battles, trades - and those can have a bearing on the team's scoring and run prevention, and thus the projected standings.
Actually, it was in Battlefield Earth where their transportation motors didn't really move objects themselves but translated the space around an object from one location to another. That translated space also swapped its contents with the other space, effectively moving the object.
A lot of those things are going to have to wait until the PECOTA cards are done - right now we only have the weighted mean forecasts, and some things (Beta, Upside, etc.) require the full range of forecasts.
Can someone check to see if the SGP option is working right in the PFM? I'm getting only tiny changes in output across the values. Also, SB seem to be extremely highly valued, to the extent that one category speedsters are getting into the top 20. Is there a way to bring this down? I thought that's what the SGP level was supposed to do.
And the positional adjustment seems to output the same results for level 1, 2, and 3, so it only seems to go from 'off' to 'crazy,' rather than a gradual increase in the value for shallow positions. Ex: Pos Adj =0, Mauer =$17.60, Pos Adj = 1, 2, or 3, Mauer = $49.03
It's not binary now, you do get different values for each setting, but the change is tiny, and not consistent. Bourn, a one category SB guy, goes from $27 at SGP=0 to $28 at SGP=1, and $25 at SGP=2 under my settings.
What I'd like, is some way to de-weight SB in these forecasts. I know that the PFM is supposed to give a prediction for earned value, not price, and its easy to devalue one category guys in drafting/bidding, but the five category guys are harder to revalue.
Here's how my dollar values are affected by SGP after six rounds of a mock straight draft:
SGP=0
McClouth - 48.23 Pence - 50.36
SGP=1
McClouth - 63.56 Pence - 27.26
SGP=2
McClouth - 33.55 Pence - 55.25
This is totally illogical to me. Shouldn't these numbers follow some kind of trend as SGP increases? It's like level 0 and level 2 are the same, and level 1 adjusts way too heavily.
I notice there is a BPID on the "hitters" and "pitchers" tabs in the spreadsheet. If you can't incorporate HOWEIDs in all your fantasy content, can you at least incorporate consistent BPIDs in all fantasy content?
It would make sorting and matching up player info so much easier!
If I could second this ... It would REALLY be great if there was a single key ID attached to each player in both the preseason projections and mid-season down-loadable statistics reports.
Mentioned it before, but you guys have David Ortiz as a +12 1B, pushing his WARP much higher than it should be, when it should actually be subtracting from his value. He defense, when displayed, is below average.
I assume the defensive numbers are projections as well, so perhaps PECOTA is picking up on a weird comp somewhere.
Just saw Gammons talking about your projected standings on MLB network. Cool, except they never mentioned 'beta' version, only that the projections can change daily. For the sake of BP's reputation, I'd reconsider the release of beta versions, and make us wait the extra few days for the real thing. As a compromise, its pretty obvious which subscribers are willing and qualified to help with beta testing if that help is needed. Thx for the updates.
This misses the point. As has been pointed out, even if the data inputs were absolutely perfect when first released, the projections only have meaning in that instant. Tomorrow... well, not so much. The projected standings SHOULD still change daily, or nearly so - and sometimes significantly. Every time a free agent is signed, injury reported, trade made, every manager's announcement that a player has "earned" a starting spot, etc. - player projections will change with each, and thus the projected standings may change as well. How can this be difficult to understand?
As for the official preseason PECOTA projection (not prediction, remember), I suppose that's the one immediately before the first pitch of Opening Day. But if in the next moment Beckett tears his rotator cuff on a pitch that breaks Jeter's wrist, well, that's the shelf life on these things. So if you're sensitive about the impact of these projections on BP's reputation, better to just avert your eyes. This ain't no art museum, it's the sausage factory floor.
Well, I see you're probably in a keeper auction points league. Those are very uncommon and it looks like the system is not designed to determine dollar figures for your league structure. You may need to create dollar figures yourself using the projected points each person is going to accrue. There are websites that can show you how to do this. Maybe check with fantasybaseballcafe.com if they still exist.
Where's Alfonso Soriano? Just ran a simulation of a 12 team league drafting 5 OF.... and he's nowhere to be found. I know he's not the monster he once was, but he should be in the top 60 OF, no?
I'm actually glad Soriano's forecast isn't too optimistic, because outside of the injury he had a lot of issues in 2009 that could hinder his future performance. I'm curious to see what the PECOTA cards say about him though, as his upper level projections could still have a hint of the old Soriano to them. I would be okay with that.
The PFM positional filter (at the top of the results) doesn't recognize "Swing" as being both a SP and RP. In other words, filtering on just SP (or just RP) excludes Swing, instead of including Swing. Also, there no longer appears to be any pitcher categorized as CL, so at the very least you may want to delete that category from the filter.
Saves projections in the playing-time weighted sheet still look wonky to me. You guys have Joakim Soria leading the league with 42 saves. Meanwhile, Heath Bell has 14, and Brian Fuentes and Mariano have 21 each. Shoot, the whole Cubs pitching staff only has 18 combined. I guess that could happen, but it seems unlikely.
Would like to echo the interest in SS/Sim numbers... they're going to come too late for me to make my borderline keeper decisions (Ronny Paulino or Jesus Flores?), but having them before the draft would be really great. No one else gives us sim players a real good metric to go by...
25 inches on the ground and now the weather-peeps are saying up to 20 more Tuesday into Wednesday. Please tell me good news is on the way in the form of new PECOTAs. My inner Jack has been stirring.
i just noticed that edgar renteria is projected with the most ABs of anyone in the most recent PECOTA set. does this seem odd to anyone else? his projected 628 would be a career high at age 34. previous three years: 494, 503, 460. offseason elbow surgery. what gives?
Any update on when the PECOTA cards are going to be done? The last thing I remember seeing was that you guys hoped to have them up last week, but realisitcally it would be this week. Well, it's Friday and still no sign of them. Just wondering when they might come out. I sure hope we don't have to wait nearly as long as we did last year.
61 comments have been left for this post.
Thanks for the update Dave. Could you possibly provide an ETA for the Upside data?