Davidi baseball prospectus chapter

[Olerud363]

Mostly for entertainment purposes only. (I am guessing now that Davidi has dumbed it down a bit AA may actually go to one of BP events). Davidi actually makes some good points here and there. But this combined with a lot of nonsense.

 

The absolute highlight

 

- Blue Jays have "An analytics department that is a key part of the organization"

 

The club has lauched basically a super AI named "the BEST". Pronounced "Beest" after Paul Beeston. A database which houses scouting reports, proprieatary analytics data, medical reports, and contractual data in one spot". It was launched by Jay Sartori who has left. So who knows. It may be really good... but the guy who knows how to run it is gone.

 

I wonder if this is what is Despairing the agents?? OK Boros. We're ready to negotiate. You don't talk to us, you talk to "the Best". It's a computer model named after Beeston. And it will Despair you!!!!"

[Olerud363]

It should be Davidi... not Dividi. If a mod wants to edit title please do. I tried but couldn't see how to edit the title.

[Olerud363]

Some other random comments that caught my eye.

 

AA - critisized for being too stat oriented.

 

Moises Sierra - "Sierra's going to hit no question" (Davidi does acknowledge that "he's dumb"!!!"

 

Anthony Gose - Will get a starting job and "could certainly reach his all star potential"

 

Brett Lawrie - Sky high ceiling. Could fall short because he's a meat head -- (reasonable)

 

R.A. Dickey - For 25 million dubious investment. ???? Dubious investment because of the prospects given up. But for his salary he's worth it at 2013 level, and will be a bargain if he bounces back.

 

Loup - "Worrisome decline in peripherals." -- 53 k, 13 bb, 5 hr. 69 ip. This is why Davidi shouldn't be writing this chapter. His 0 homerun rate was godly in 2012 but it was 20 innings. His peripherals are fine. Yes his hr rate increased by infinite... but it was still good.

 

Nolin - Should be on the big-league mound every fifth day this year.

 

Hutchison - Not even given a paragraph. Just a "lineout". Likely to work in aaa. Another reason Davidi shouldn't be writing this chapter. I don't mind that he's optimistic on Nolin. But really a blue jay expert should know that Hutchison still ranks ahead.

[NorthOf49]

I'll add that BP's daily podcast (Effectively Wild) is touring around the MLB right now, talking about a team a day and today was the Blue Jays. I listened to the first half of the episode, which had Sam, Ben and Adam Sobsey talking about the Blue Jays. Will listen to the next half, with Davidi, later today.

--

 

Good post. The BEST existing makes a lot of sense to me and confirms my own thoughts about what AA's "proprietary systems" are. From what I hear they have as many scouts as all other teams have combined and AA has been slipping in comments about an aggregated scouting report approach for a while. He seems to think it's a worthwhile attempt to beat the market. It hasn't worked yet but maybe they're still figuring out how to use the data.

 

But I hope they're not putting all of their analytics bullets into it, or at least are combining it with other inputs. I wonder if Pitch/Hit/Field FX is integrated.

[eastcoastjaysfan]

All of this analytics stuff is wonderful and stuff, but if you don't actually pull the trigger on anything it's all pointless.

 

Let's build a system to beat the market and then not actually act on any of its outputs.

 

Great plan, what could possibly go wrong?

 

All talk, no action.

[Abomination]

All of this analytics stuff is wonderful and stuff, but if you don't actually pull the trigger on anything it's all pointless.

 

Let's build a system to beat the market and then not actually act on any of its outputs.

 

Great plan, what could possibly go wrong?

 

All talk, no action.

 

For all we know, the outputs may have suggested not to go after those other guys, and could ultimately prove correct. We'll just have to wait and see.

[Nox]

The club has lauched basically a super AI named "the BEST".

 

What a f***ing joke.

[Olerud363]

I'll add that BP's daily podcast (Effectively Wild) is touring around the MLB right now, talking about a team a day and today was the Blue Jays. I listened to the first half of the episode, which had Sam, Ben and Adam Sobsey talking about the Blue Jays. Will listen to the next half, with Davidi, later today.

--

 

Good post. The BEST existing makes a lot of sense to me and confirms my own thoughts about what AA's "proprietary systems" are. From what I hear they have as many scouts as all other teams have combined and AA has been slipping in comments about an aggregated scouting report approach for a while. He seems to think it's a worthwhile attempt to beat the market. It hasn't worked yet but maybe they're still figuring out how to use the data.

 

But I hope they're not putting all of their analytics bullets into it, or at least are combining it with other inputs. I wonder if Pitch/Hit/Field FX is integrated.

 

I just can't see that they are using analytics properly at all.

 

Here's a simple example. I want to predict how many games the Blue Jays starters are going to play next year. Simplest system possible. Add up their career average games played and divide by number of seasons for all players. Simple average basically.

 

Then people say well what about his, or what about that. It's not fair to put Goins on the list, he was a callup. Not fair to put Melky on he got a steroid suspension. And Lind platooned some years blah, blah, blah.

 

But I think this is fine. We get an average, it's probably about 115. And that tells you that the Jays need more depth. Probably actually need another full time player and some positional flexibility to cover for inevitable injuries.

 

You can have the most complicated system in the world. And what's it going to project?? 118.56?? Whatever is I bet it's not far from the simple average. Assuming it is done correctly.

 

Now here is what I think. You have these experts systems. Especially in a coorporate environemnt, and they can be tweaked. AA says that Reyes' injuries were "fluke", so you could tweak your system so that an injury can be labeled as "fluke". AA wants to see Reyes and Bautista projected for 145 games. They yell at the programmers until they get the answer they want. What about this, what about that. You have to model Goins differently, and Lawrie differently. etc. Melky. Model the tumour somehow. But it's all ********. They're just rigging the system. Maybe why Sartori left??

[Nox]

Either Davidi or the Blue Jays don't know what computer AI is. Or neither, I suppose.

[CHRIS]

I come across this so many times in my work. Having data is not a solution; using good data to make informed decisions is the goal.

 

I mean, if they have a database that contains all the details of Chapman's dubious bullpen session that led them to the decision that they didn't want to sign him, what good is that? Garbage in/Garbage out and all that.

 

Also, I've clearly missed something involving "despair". Link please.

[flafson]

I wouldn't be surprised if the "AI" is actually Beeston himself and there is no AI.

[Olerud363]

All of this analytics stuff is wonderful and stuff, but if you don't actually pull the trigger on anything it's all pointless.

 

Let's build a system to beat the market and then not actually act on any of its outputs.

 

Great plan, what could possibly go wrong?

 

All talk, no action.

 

BP projects EE to play full time. Nobody else gets 500 plate appearances. So the Jays analytical system, assuming it works and assuming they use it should indicate they need more depth.

 

Guys like Gose, Izturus, Sierra, Thole are going to get significant time at sub replacement. Add in Goins and this could be bad...

 

AA's strategy seems to be that injuries are luck and luck turns at some point.

[Olerud363]

I wouldn't be surprised if the "AI" is actually Beeston himself and there is no AI.

 

These systems are only as good as the knowledge that is programmed into them. Who knows they may have programmed it to think like Beeston.

[Boxcar]

"Super AI" probably translates to a very basic Microsoft Access 2003 document.

[Boxcar]

Yeah, Sartori was a genius. A real statistics savant. Or so I've heard.

[Abomination]

Either Davidi or the Blue Jays don't know what computer AI is. Or neither, I suppose.

 

Yeah, I suspect Davidi may have tried to get too cute in his wording and flubbed that up. Really though, I think machine learning techniques are the future of sports analytics.

[Olerud363]

Either Davidi or the Blue Jays don't know what computer AI is. Or neither, I suppose.

 

To be fair Davidi did not describe as an AI. Just as an analytics database. The context hints at something more than a simple database "the project was in development for 2 and a half years before going live".

 

The description indicates to me that it is an ill-conceived expert-knowledge system.

[Boxcar]

I come across this so many times in my work. Having data is not a solution; using good data to make informed decisions is the goal.

 

I mean, if they have a database that contains all the details of Chapman's dubious bullpen session that led them to the decision that they didn't want to sign him, what good is that? Garbage in/Garbage out and all that.

 

Also, I've clearly missed something involving "despair". Link please.

 

I believe this was the first instance of it being used. Like all good things that are hilarious, looks like it was borne from Angrioter's broken Mexified English. http://www.bluejaysmessageboard.com/threads/1979-Around-Baseball-2014?p=189909&viewfull=1#post189909

[NorthOf49]

I believe this was the first instance of it being used. Like all good things that are hilarious, looks like it was borne from Angrioter's broken Mexified English. http://www.bluejaysmessageboard.com/threads/1979-Around-Baseball-2014?p=189909&viewfull=1#post189909

It found its way into a DJF post too lol: http://blogs.thescore.com/djf/2014/02/13/are-these-the-jays-terms/

[Olerud363]

I come across this so many times in my work. Having data is not a solution; using good data to make informed decisions is the goal.

 

I mean, if they have a database that contains all the details of Chapman's dubious bullpen session that led them to the decision that they didn't want to sign him, what good is that? Garbage in/Garbage out and all that.

 

Also, I've clearly missed something involving "despair". Link please.

 

There's a story about it on djf. But the info probably appeared somewhere here first. Angrioter posted some translations about Jays negotiating tactics.

 

Are These The Jays’ Terms?

by Andrew Stoeten

 

A tweet from earlier this afternoon out of the Dominican Republic from a man named Roosevelt Comarazamy has been circulating, suggesting precisely what the the Jays have — allegedly — offered free agents Ubaldo Jimenez and Ervin Santana.

 

Translated — thanks kindly to @Angelo_riot — it says “Toronto is playing to despair Ubaldo and Ervin’s agents. The Blue Jays will offer $27M for 3 years.”

 

First, some background: Angelo_riot added that Comarazamy is the play-by-play man for Santo Domingo-based Dominican league club Tigres del Licey, and that he has ties with Jimenez, Santana, Jose Bautista, and Esmil Rogers. “He’s legit,” Angelo_rio adds.

[Olerud363]

Yeah, I suspect Davidi may have tried to get too cute in his wording and flubbed that up. Really though, I think machine learning techniques are the future of sports analytics.

 

To be fair AI was my wording not David's. Davidi describes it as "an analytics database" that was "in development for 2 and a half years before going live".

 

Obviously more than a simple database. Sounds like an expert system or atleast the Jays attempt at one. Or maybe they are just so dumb it took them 2 and a half years to get a simple database set up.

[Dr. Dinger]

- Blue Jays have "An analytics department that is a key part of the organization"

 

********. They have an analytics department that is full of geniuses and horrendously ignored when it comes to personnel decision-making.

[NorthOf49]

Olerud, do you know what 'BEST' is an acronym for?

[Nox]

Really though, I think machine learning techniques are the future of sports analytics.

 

Bayesian techniques (a key piece to the core of machine learning) are more the now than the future for the actual smart teams.

[Nox]

Botany Enlightens Shallow Thinkers

[Olerud363]

Olerud, do you know what 'BEST' is an acronym for?

 

No idea. Was not revealed in Davidi's article.

[xposbrad]

Olerud, do you know what 'BEST' is an acronym for?

 

is it like short for something you'd find on Terb? like GFE or BBW?

[G-Snarls]

It should be Davidi... not Dividi. If a mod wants to edit title please do. I tried but couldn't see how to edit the title.

 

Fixed

[Olerud363]

Bayesian techniques (a key piece to the core of machine learning) are more the now than the future for the actual smart teams.

 

I'm no insider but I am guessing one of the problems with the bad organizations is a fundamental misunderstanding of the nature of prediction.

 

AA talks like he is predicting good health for Reyes, Lawrie, Bautista. I look at their career stats, eyeball the averages. These 3 are going to play about 115 games each, hit .265. .340 .450 as a group. I haven't even used a calculator. You use a calculator probably do a bit better. Use unsupervised clustering and a database of lifetime stats of every mlb player ever... better still. Proprietary batted ball data, even better. But all these systems will tell you roughly that these guys will play 115 games each at their career averages.

 

But AA talks like their going to play 150 games at their best... and damn he might end up being right about 2/3. But one of them will play only 40 horrible games. Or one of them will be MVP and the other Aaron Hill 2010. Or they'll all be great and 5 other guys won't. And at the end of the day AA will say "I was right about this guy and nobody could of seen what happened to that guy."

 

The guy who predicts an average of 115 games is right when they play 150, 150, and 40. But the guy who predicted good health for all of them gets to say, "I nailed 2/3 and no one could of seen that X would get Herpes".

[8D]

Link to Shi chat on BP: http://www.baseballprospectus.com/article.php?articleid=22886