Projections Idea

[david_hume]

I'm considering using a weighted average of a few different projection sets to generate the numbers I will use this year. What do people think of this method?

[Todd Zola]

Honestly? And this is not because projections are part of our business, but as I wrote in a piece for SI.com last week (which we will be co-posting here soon), the difference between most systems is playing time. The skills are different, for sure, but the playing time is the more relevant variable.

So my personal suggestion is to choose any system, or systems to provide the "rate of performance."

HR/PA
BB/PA
K/PA
RBI/PA

etc.

Then add in your own playing time and you basically have YOUR OWN projections.

Perhaps a slightly different way to approach the same thing is use different systems to highlight guys projected for different playing time, then spend your time deciding which playing time assignment is more relevant.

We may say 600 AB.
Someone else may say 500 AB.

Averaging systems will give the player 550 AB. But if YOU feel the player will get 500, I would use 500 or else you will end up rostering a player you really do not think will return what you paid. And the opposite is true. If you think he will get 600 and use the 550 average, you will pass on a player when he may have been actually available at value to you.

Here is the link to the SI.com piece....

http://sportsillustrated.cnn.com/2010/f ... index.html

[viper]

I'm guessing that the HR/PA, etc. are pretty much a formula based on years, age, history, etc. Then you get together [two or more] and discuss PA. The ratios likely stay fixed through the projection season but the pA time changes based on signing and what you see in spring training.

Once you pretty much agree on PA, you then come up with actual numbers for each category. How much are minor modifications made based on some type of consensus adjustment - basically the old smell test for your numbers smells poorly.

[Guest]

The biggest strength of your method will be its biggest weakness - you will average out all of the outliers.

What this means is you'll have a projection set more tightly packed together than probably the individual sets suggest. This by itself isn't an issue.

But you'll want to have a separate area on your draft list which outlines guys you and/or your draft sets are higher or lower about, because in the heat of the draft your list isn't going to help you with that.

It's a perfectly fine idea but just remember consensus projections require a little more thinking from the user, is all. Other than because I think our projections are as good as you can get, I can't argue with the approach.

[kjduke]

The biggest strength of your method will be its biggest weakness - you will average out all of the outliers.

Completely agree. It will be like drafting using ADP. Not a bad method if you're short on time or if your edge is in other aspects of the game, but you won't have any edge in the projections.

I think the best method is in Todd's first response to this topic - projecting stats per PA - which allows you to adjust on the fly to news. Using absolute projections will handicap you when situations change. The only disagreement I have with Todd's general method (from another thread) is in projecting PA/R and PA/RBI - I think there is value added in considering both a player's team lineup, and where that player hits in the lineup. Batting 4th vs 6th in the lineup makes a significant difference in the rate of R/RBI and in the ratio between the two, as does of course hitting in a relatively stronger lineup overall.

[Todd Zola]

The only disagreement I have with Todd's general method (from another thread) is in projecting PA/R and PA/RBI - I think there is value added in considering both a player's team lineup, and where that player hits in the lineup. Batting 4th vs 6th in the lineup makes a significant difference in the rate of R/RBI and in the ratio between the two, as does of course hitting in a relatively stronger lineup overall.

To clarify, I agree. I guess I "assumed" it was understood the RBI or R per PA was influenced by these factors. I should know by now to remember the ass-u-me axiom, especially in the more theoretical discussions.

We combine what each individual event "should" lead to with respect to runs and RBI with the team context.

[Todd Zola]

Now to clarify my clarification...

We have been asked in the past how we change projections if someone like Matt Kemp is moved from 7th to 3rd in the order.

My answer has been the additional PA will sufficiently add to the counting stats, so we really don't make any further adjustments.

Again here, I assumed the ORIGINAL runs and RBI considered team context. Whether we make an additional change depends on the nature of the change.

It would have to be something extreme like what was talked about with Jose Reyes a few years back -- going from leadoff to 3rd (which could happen with someone like Grady Sizemore) to make a real subjective adjustment.

[kjduke]

If you look at the difference in lge averages in PA per R/RBI for batting 7th vs 3rd, for example, then adjust for that particular player along a regression line based on his hitting stats, you get a good idea of how the R and RBI ratios should change for any lineup slot.

[Guest]

The issue isn't with an understanding of the repercussions of the lineup change as much as it's an issue with actually adjusting projections which already probably have a built in 10-15% error just to start for a lineup slot change that may or may not stick for an entire season.

But yes, the point is a good one.

[david_hume]

thanks all for the great replies. i hope to have a more detailed response soon, but for now...

The biggest strength of your method will be its biggest weakness - you will average out all of the outliers.

this is my biggest fear, and the problem i foresaw with this method. i definitely don't want to just flatten out everything and draft based on ADP.

the concern that got me thinking about all of this in the first place is that i think i tend to pay up for the guys which the particular data set i am using likes, and i'm not always aware i am paying up for them. perhaps the solution to this is that i need to know the player pool and market values better. another thing it stresses to me is the importance of having guys who YOU think are going to outperform the market and/or the data set you are using. so, okay, those are some good things to focus on.

more later.

[david_hume]

So my personal suggestion is to choose any system, or systems to provide the "rate of performance."
...
Then add in your own playing time and you basically have YOUR OWN projections.

okay, let's say i do this. my next question is, what's the best way to turn these raw values into dollar values? i am familiar with the old school SGP method, but i gather there is a lot of opposition to it these and people are mostly using other methods. so what are they? any links appreciated.

[AllstonRockCity]

If you made your own projections you could plug them right into the CVRC and that would spit out your dollar values. You would just need to format them exactly as they appear on the 'players' tab of the CVRC.

[Todd Zola]

So my personal suggestion is to choose any system, or systems to provide the "rate of performance."
...
Then add in your own playing time and you basically have YOUR OWN projections.

okay, let's say i do this. my next question is, what's the best way to turn these raw values into dollar values? i am familiar with the old school SGP method, but i gather there is a lot of opposition to it these and people are mostly using other methods. so what are they? any links appreciated.

In my youth, I led the crusade against SGP and still feel the method is flawed from a theoretical sense. But I have softened my stance from a practical sense as I have learned there are far more important things than having a static number attached to a very fluid entity, a theme being discussed in several threads.

We are working on getting the archived essays posted, one of which outlines our procedure in great detail.

Platinum subscribers have access to more essays, which have streamlined the method as well as the CVRC, an excel based program that calculates the values automatically. Of course, it is fed our projections, but you can easily type in your own.

[JP Kastner]

I take a difference stance on this and have written about this before. I keep a ranking of common projection systems. One can convert dollar values into a ranking and one can convert ADP into a ranking. There are certain key players that each system values more. If I know my opponent is using ADP and player X is going 5th round and I think he is worth 3rd round, I should pick him up 4th round. The difference is one round of value. The same things goes for someone I know is using another projection system like BBHQ.

If I don't know what method they are using, there are certain "indicators". Mike Jacobs was a good example last year for BBHQ. If my opponent drafts Mike Jacobs earlier than I value him and ADP values him, then I'm writing BBHQ next to his name and now I what numbers he is using. I basically have his cheat sheet in front of me.

Smooshing the data together doesn't give you much.

[david_hume]

In my youth, I led the crusade against SGP and still feel the method is flawed from a theoretical sense. But I have softened my stance from a practical sense as I have learned there are far more important things than having a static number attached to a very fluid entity, a theme being discussed in several threads.

We are working on getting the archived essays posted, one of which outlines our procedure in great detail.

Platinum subscribers have access to more essays, which have streamlined the method as well as the CVRC, an excel based program that calculates the values automatically. Of course, it is fed our projections, but you can easily type in your own.

I do care about the practical application of the various valuation methods, of course. But personally, I also care a lot about the theoretical puzzle, so I am quite interested in reading any rants you wrote against SGP and all of the stuff about your valuation methods.

Is there a way to use CVRC with a mac? I've avoided it so far because I am under the impression it is PC only.

[Todd Zola]

Part of the free archives will be some anti-SGP harangues.

As opposed to some anti-Aaron Harangs.

The MACROS will not work with the Mac.

However, all the macros do is bring the final result out from the bowels of the program and put it neat and clean on its own page.

If someone has even intermediate Excel acumen, it is very easy to make it Mac-usable.

And someone with beginner's acumen and some patience can be taught how to do it pretty easily.

[david_hume]

Perhaps a slightly different way to approach the same thing is use different systems to highlight guys projected for different playing time, then spend your time deciding which playing time assignment is more relevant.

We may say 600 AB.
Someone else may say 500 AB.

Averaging systems will give the player 550 AB. But if YOU feel the player will get 500, I would use 500 or else you will end up rostering a player you really do not think will return what you paid. And the opposite is true. If you think he will get 600 and use the 550 average, you will pass on a player when he may have been actually available at value to you.

Here is the link to the SI.com piece....

http://sportsillustrated.cnn.com/2010/f ... index.html

As I re-read the SI piece today, the following occurred to me:

The reason I don't attempt to put together my own projections from scratch is that I am confident that the Mastersball guys (or the authors of whichever set I use) know more about the skills of each player and the other elements (age, ballpark, injuries, etc) that go into projections that I do. But how is this any different for playing time? If Todd tells me a guy will get 600 AB and another reputable site projects 500, who am I to decide?

This it not an annoying philosophical question. Okay, maybe it is. I'm not sure.

I just don't want the 600 vs. 500 decision above to be reduced to, "I've always liked David DeJesus, so I'll give him 600, but Vernon Wells really screwed me two years ago, so I'll give him 500."

[Todd Zola]

As I re-read the SI piece today, the following occurred to me:

The reason I don't attempt to put together my own projections from scratch is that I am confident that the Mastersball guys (or the authors of whichever set I use) know more about the skills of each player and the other elements (age, ballpark, injuries, etc) that go into projections that I do. But how is this any different for playing time? If Todd tells me a guy will get 600 AB and another reputable site projects 500, who am I to decide?

This it not an annoying philosophical question. Okay, maybe it is. I'm not sure.

I just don't want the 600 vs. 500 decision above to be reduced to, "I've always liked David DeJesus, so I'll give him 600, but Vernon Wells really screwed me two years ago, so I'll give him 500."

I don't think you are so much deciding on simply A versus B. If we say 600, that means someone else gives those extra 100 to someone else. Maybe you look at the other numbers and say "no way will So-And-So get that many AB, the 600 looks more reasonable."

Besides, a lot of players will have similar at bats. The ones that differ will be guys like Jose Reyes (returning from injury), David Freese, Cliff Pennington, Josh Hamilton, etc. Maybe you don't have an opinion, but the point was this sort of thing is something we really do not have any more credibility than anyone else that really follows the game. I think it is known that we don't have clubhouse access to all 30 teams -- or maybe the ability to actually talk to all 30 teams.

If you trust me, or Ron Shandler, or Jeff Erickson, or whoever, that's fine. But if someone takes Jose Reyes because he was ranked higher by one of us that thought he was pretty much recovered and he ends up getting hurt, that someone should not then say "jack-offs, there was no way he was going to get 600 AB, they really screwed up that projection." If that someone did not think he would get 600AB, they should not have drafted him valued at 600 AB.

Again, most players playing time is also a weighted 3 year average with injury considerations as well as current roster constitution on that team. So most AB will be reasonably similar. It's the different ones that really matter.

[AllstonRockCity]

Part of the free archives will be some anti-SGP harangues.

As opposed to some anti-Aaron Harangs.

The MACROS will not work with the Mac.

However, all the macros do is bring the final result out from the bowels of the program and put it neat and clean on its own page.

If someone has even intermediate Excel acumen, it is very easy to make it Mac-usable.

And someone with beginner's acumen and some patience can be taught how to do it pretty easily.

I'm a MAC user and thanks to Todd the CVRC works for me!

(now read that again and pretend its late at night and you just saw an infomercial for the CVRC). Still not funny? well fine, its still true