2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

[TripleThreat1973]

Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

Just letting everyone know that this project is still going forward.

When entering the rating, I noticed two things (after I entered ALL of the batters).

[1] Some minor leaguers were too high due to THT (Oliver's ) projections receiving too much weighting, or being the oonly system projected for those players.

I went back and combined (evenly weighted) 5 projection systems, and minor leaguers are much more "in check" now (especially the guys that killed at A).

[2] I didn't have anything for fielding and just kept the fielding ratings as is.

Each fielding rating has 3 metrics used for its rating. Each formula is based on the amount that the player is above/beloiw average (usually a rating of 65 to 70 is average).

I'll explain later.

For MiLBers, I will start with "replacement level" fielding, and then use their "defensive runs" projections as a modifier to see how far above/below the replacement level their fielding ratings will be. Kind of a compromise between (a) having MiLBers ready for major ;league fielding (reality), and (b) having them be able to progress in a multi-season franchise. The guys that are MiLBers that will "see time" in MLB in 2011 already have fielding metrics available.

[3] I now have formulas (multiple) for control. Again, there is not just all-encompassing formula for pitch control, pitch break, etc ... unless you want to limit yourself to one metric to govern the entire rating (I don't). So, multiple metric and formulas are used for each rating, and the data are averaged together to get a complete and accurate rating for each category.

I'd love to have this done by the end of this upcoming weekend.

[ImBaconBeckham]

Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

What you're doing is awesome TripleThreat.

I tried to do the same last year for 2K10 - using sabermetric principles to re-do ratings. The problem was that it got far too time consuming for one person to accomplish. I hope that you have a lot of time on your hands.

If you need any help in the future, I'll be glad to help. I know a good amount about sabermetrics.

BTW, how do you edit the ratings on the Show to specific numbers? They're all bars, not numbers like 2K's.

[Jason_19]

Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

Triple, thank you for posting all of this. I agree with a lot of what you have posted.

how do you edit the ratings on the Show to specific numbers? They're all bars, not numbers like 2K's.

When you go into the edit screen, the numbers are displayed.

[TripleThreat1973]

Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

Yeah, I wish I had started this is as soon as projections came out. Unfortunately, they all come out at different times.

Starting this right at the end of the season would be ideal.

I actually do NOT want to know how many hours I have spent on this ... its probably over 150+ hours.

Its probably not as bad as it sounds because I dont watch much tv or play many video games. So, Im still able to do a lot of other things in life, and be able to do this ... simply because I dont spend 3 hours a night with tv and-or video games.

-----------------------------------------

Another comment about Bahnzo formulas and projections. It is also possible that the ratings came in low because of the nature of projections. No one is projected to hit more than around 40 homers. We know somebody will hit more than 40 homers, but it will be due to abnormal HR-FB rates. Likewise the highest BA is .314 (porjected), although we know guys will hit higher than .314 ... but theyll do it because of BABIP rates that are much higher than their career norm (Yeah, in other words it will be "luck", or in other words "due to something they cannot or willnot repeat).

So, its not projected. So at least when using projections for a single season, using formulas that allocate rating points based on the degree above-below league average seemingly will work better.

In that regard, one formula or metric is not likely to get it done as well as a handful of formulas and metrics averaged together for a single, but complex, rating.

[Jason_19]

Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

Yeah, I wish I had started this is as soon as projections came out. Unfortunately, they all come out at different times.

Starting this right at the end of the season would be ideal.

I actually do NOT want to know how many hours I have spent on this ... its probably over 150+ hours.

Its probably not as bad as it sounds because I dont watch much tv or play many video games. So, Im still able to do a lot of other things in life, and be able to do this ... simply because I dont spend 3 hours a night with tv and-or video games.

-----------------------------------------

Another comment about Bahnzo formulas and projections. It is also possible that the ratings came in low because of the nature of projections. No one is projected to hit more than around 40 homers. We know somebody will hit more than 40 homers, but it will be due to abnormal HR-FB rates. Likewise the highest BA is .314 (porjected), although we know guys will hit higher than .314 ... but theyll do it because of BABIP rates that are much higher than their career norm (Yeah, in other words it will be "luck", or in other words "due to something they cannot or willnot repeat).

So, its not projected. So at least when using projections for a single season, using formulas that allocate rating points based on the degree above-below league average seemingly will work better.

In that regard, one formula or metric is not likely to get it done as well as a handful of formulas and metrics averaged together for a single, but complex, rating.

I know what you mean.

I never rate players based on projections from other people. I rate players based on my own projections. I can see why some people might do this because there aren't too many people who know a lot about too many players, so I don't mean that as a knock. This may sound like bragging, but my projections are always - not sometimes - more accurate than the the top experts' projections overall. I am not a professional scout, but I have coached and scouted at different amateur levels and am very confident in my ability to evaluate talent and to project stats for players at every level.

With that being, I usually create my roster sets and there is no way get accurate results. You could travel into the future and find out what every player's stats will be at every level and you still won't be able to represent that in a game. The only thing that you can do is a rate a player based on his skill sets so that when you actually play with that player his attributes will accurately represent his ability.

The range of simulation stats that come from a certain attribute varies way too much. You can have two players in the same division who both hit in the #3 spot, have the same types of players (players who have the same attributes) around them in the order and one of these players (the #3 hitters) can be rated in the 40s and 50s in every category and can hit .320 with 35 HRs and have 15-25 SBs with 75 Ks. The other player who is rated in the mid 70s in every category can hit .250 with 10 HRs and have 8 SBs with 175 Ks.

Now, if a player is playing way above his level of talent in a season, I will adjust his attributes , based on my attribute scale, to almost fit in with the players who actually have that level of talent to consistently put up those kinds of stats. I do this because it simply gives the video game version of this player a better chance to perform the way that he is actually playing in the season. I know that these players could still put up those same types of high stats when simulating with their more deserving lower attributes, but if someone plays with that player, he has a very small chance to perform up to that level. I would rather see too many high simulation stats - because you're going to get a ton of those anyway - than have a bunch of players who do not have the ability to play the way that they are actually playing in a season when you play with them in a game. I hope this makes sense.

As for minor league players, I don't have a certain formula to lower ratings. I know that most people have a formula to lower the ratings of minor league players. I simply rate players based on their skill level.

[Karle]

Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

Triple Threat,

I have been reading with great interest your posts on this thread.

In regard to Major League Equivalents for position players I have been using the MLE calculator on the Minor League Splits site. However, for pitchers I am really at a loss as to how one calculates MLE for H/9, HR/9, BB/9 and SO/9. I note you use some of these for values calculating control and break on pitches and I assume these numbers are calculated off of MLE numbers rather than the unadjusted values. Am I corect about this?

I attempted to obtain MLE numbers for these pitcher values by reducing the number (for strikeouts) or increasing the number for the other three by 10% for AAA numbers, 20% for AA numbers and 30% for Advanced A or A numbers. However, I am sure this was only a quick and dirty way to get more appropriate (but not necessarily correct) numbers and there must be better ways to do it.

I am trying to produce the Detroit minor leaguers for the OSFM set for 2011 which I have volunteered to assist Knight with and there has certainly been alot of number crunching just for this one franchise. I cannot imagine doing the same for all 30 teams.

As for Baserunning Ability, I actually went through all the recaps of the games for Detroit's AAA, AA, A+ and A teams and collected the advanced baserunning information which baseball-reference.com provides for major league players. I used this information, though I made up my own formula to get at least good relative numbers.

Thanks for all your work on this.

Karle

[Rwings832]

Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

Triple Threat,

I have been reading with great interest your posts on this thread.

In regard to Major League Equivalents for position players I have been using the MLE calculator on the Minor League Splits site. However, for pitchers I am really at a loss as to how one calculates MLE for H/9, HR/9, BB/9 and SO/9. I note you use some of these for values calculating control and break on pitches and I assume these numbers are calculated off of MLE numbers rather than the unadjusted values. Am I corect about this?

I am trying to produce the Detroit minor leaguers for the OSFM set for 2011 which I have volunteered to assist Knight with and there has certainly been alot of number crunching just for this one franchise. I cannot imagine doing the same for all 30 teams.

As for Baserunning Ability, I actually went through all the recaps of the games for Detroit's AAA, AA, A+ and A teams and collected the advanced baserunning information which baseball-reference.com provides for major league players. I used this information, though I made up my own formula to get at least good relative numbers.

Thanks for all your work on this.

Karle

Karle,

If you need help with Detroit, I can definately share with you about some of the players for the OSFM...I do see quite a bit of the Mud Hens games.

[Karle]

Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

Rwings,
I will definitely keep you in mind. Last year I subscribed to MiLB.tv hoping to obtain information on Toledo player accessories and equipment as well as lower level teams, but was very disappointed both with the quality of the picture as well as the fact that the lower minor league Detroit Franchise teams were not covered.

Karle

[TripleThreat1973]

Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

I know what you mean.

I never rate players based on projections from other people. I rate players based on my own projections. I can see why some people might do this because there aren't too many people who know a lot about too many players, so I don't mean that as a knock. This may sound like bragging, but my projections are always - not sometimes - more accurate than the the top experts' projections overall.

Okay, but let's be real. Almost all projections work the same way ... a weighted average of the previous 3-5 seasons.

Some apply more aspects like park factors, league factors, aging curves based on player type, etc. The projections for established major league players are almost identical. It's straight-forward.

Marcels is the simplest system. It is strictly a 3-year weighted average, assuming that all players regress to their mean. It is typically very high in the "most accurate list". It just illustrates the point that "most players always regress to their mean". Fluctuating thngs like LOB%, BABIP, that drastically affect ERA (but not FIP) are not projected because of the "luck" nature of them. Same thing with batter BABIP, which is a strong influence on batting average.

Projection young players and MiLBers is where the tricky stuff comes in.

The only thing that you can do is a rate a player based on his skill sets so that when you actually play with that player his attributes will accurately represent his ability.

That would be a strong case for using projections versus "last season's stats", since the last seasons stats are reliant on aspects that are "unrepeatable" (high/low BABIP, unsustainable HR/FB rates, etc). In other words Bautista isn't likely to sustain his 2010 HR/FB rate which was double his rate of any other season. Likewise Cahill experienced a BABIP 100 points lower than his career average, and Josh Hamilton just went bonkers on everything.

Projections take that aspect out and give a season stat line that the player would expect to experience with "luck neutral" aspects.

In the actual game, all sorts of stuff can happen. As I mentioned in this thread. Mark Reynolds hit anywhere from .220 to .287 with the same ratings. It happens.

I'm after what you describe, that a player will generally perform to his skill set. Even the team you play as in a franchise will be highly influenced by your skill/sliders/patience, etc. In other words you can play with the team with the worst vision rating, and lead the league in the fewst strikeouts, if you have decent ability.

The idea is that when you sim 10 seasons, the same expected players are at the top of the leaderboards. There will always be risers and fallers due to the inherent randomness of what can happen in baseball.

If you were to start a franchise with the cardinals and Wainwright was lost for the season in Spring Training, you'd say "That's bullcrap, he's one of themost durable pitchers in baseball." You'd be right, but &%&$ happens.

As for minor league players, I don't have a certain formula to lower ratings. I know that most people have a formula to lower the ratings of minor league players. I simply rate players based on their skill level.

If using MLE's with appropriate accomodations made (neutralized), a very slight reduction (2-10%) would suffice. If you're using just regular minor league stats, you'd need to research park/league factors due to the drastic and dramatic hitter/pitcher friendly leagues in MiLB.

[Jason_19]

Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

Okay, but let's be real. Almost all projections work the same way ... a weighted average of the previous 3-5 seasons.

Some apply more aspects like park factors, league factors, aging curves based on player type, etc. The projections for established major league players are almost identical. It's straight-forward.

Marcels is the simplest system. It is strictly a 3-year weighted average, assuming that all players regress to their mean. It is typically very high in the "most accurate list". It just illustrates the point that "most players always regress to their mean". Fluctuating thngs like LOB%, BABIP, that drastically affect ERA (but not FIP) are not projected because of the "luck" nature of them. Same thing with batter BABIP, which is a strong influence on batting average.

I completely understand what you're saying. I just don't think that weighted averages are worth much. Plus, you can't even calculate them for a large portion of the most talented players in the league because they haven't even been in the league long enough.

I was simply saying that I never rate a player based solely on his stats. I basically rate MLB players based on their skill sets in relation to what I project them to do and the results that I get are always far more accurate than any of the publications that project stats, no matter formula is used. I rate the MiLB players based solely on their skill sets/ability.

If you were to start a franchise with the cardinals and Wainwright was lost for the season in Spring Training, you'd say "That's bullcrap, he's one of themost durable pitchers in baseball." You'd be right, but &%&$ happens.

Once again, I completely understand your point. With that being said, this is where specific knowledge comes in. I would think that this would be mind boggling accurate because Wainwright has some of the worst mechanics of any pitcher in the game.

http://www.chrisoleary.com/projects/Baseball/Pitching/RethinkingPitching/Essays/AdamWainwrightInvertedW.html

http://www.chrisoleary.com/projects/baseball/pitching/RethinkingPitching/Essays/AdamWainwrightsPitchingMechanics.html

With all that being said, I still think that what you're doing is great and I appreciate it. I just don't think that it will ever produce accurate stats or accurately rated players, especially for the younger players.

[TripleThreat1973]

Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

I understand what you're saying. Projecting players with 5 years or less MLB service is the most challenging aspect. For example, Mike Stanton's projections are all over the place.

I'm a pitching coach (former pitcher) and a big fan of O'Leary (He does a lot of the dirty work in terms of images/video). In regards to Wainwright, O'Leary's biggest concern is the height of his elbow, which may or may not be a big deal. The big concern with Wainwright (especially in hindsight) is his worload combined with being a slider thrower. At some point in time, all slider throwers that are not genetic freaks (Randy Johnson) have elbow trouble. The good news is that elbow trouble beats the hell out of shoulder trouble (Brandon Webb), and a lot easier to come back from, especially with TJ. Wainwright for his career, throws 17% sliders, almost as much as he throws his dominant curveball.

He's always had the "high PAS elbow", not quite the same as the inverted W (Prior, Reyes), which is also why his curveball is so good. It's a trade off. Kevin Brown was dominant because of all the torque he put on his arm. It created a ton of sinking action, but also had a high risk of injury. The alternative is to have a world full of Jeff Suppan's. Heh Heh.

But, really, any pitcher that throws a lot of slider will be hurt at some point, and out for a season. It comes with the territory. High risk, high reward.

I do like how Chris always points out that a MLB pitcher, who has used the same mechanics since likely HS, needs to change his mechanics in order to continue being productive or reach his potential. There's no way a MLB pitcher can undo that much muscle memory and comfort at the highest level of competition. Carpenter's in the same boat on that one. 3-4 years of dominance, followed by a lost season. Rinse and repeat.

As far as predicting injuries ... there's always some guru predicting a major injury for everyone. There's a LOT of guys just waiting for Timmy to get hurt, so they can throw themselves a parade ... while ignoring the almost 1000 IP of dominance out of a small frame body.

The bigger concern with Wainwright is likely that he jumped right from closer in 06 to 200+ IP season, and has thrown 200 IP in 3 of his 4 seasons as a starter. Still, even with his mechanics, he could reasonably come back and throw 4-5 consecutive 200 IP season before another injury. We really don't know how much is mechanics, genetics, wear and tear, etc. Scott Kazmir, for example, has very good mechanics, and just has a dead arm from all the innings at a young age. Even then, plenty of young pitchers with good mechanics are brought along slowly, and still have career ending-altering injuries. It's not an endeavor that many guys enjoy long careers. And then there's Rich Harden, who does 3 major things wrong ... *grin*

With AW50, I wonder if there was serious evaluation done at the end of the season, and whether they figured "rest" would fix whatever was ailing him. Obviously, he was probably hurt at the end of the season, and rest didn;t fix it. Unfortunate, especially as I'ma huge Cardinals fan.

But, really what I was getting at, is that looking at stats or projections, Wainwright would project for 200 IP of dominance almost any season, or in the game's language "a high durability rating". As far as the game goes, Greg maddux would be just as much of a injury concern as someone with a similar durability rating.

I do agree completely, that even with "perfect ratings", there will be variability in performances.

With injuries, two people can look at the same stats and see a lot of games played without injury and think [1] wow, he's very durable, and another thinks "whoa, he's due for an injury". Pujols, for example, will probably get hurt sometime in the next 5 years, even while jogging out ground balls, and having a flawless, almost effortless swing. Nature of the beast.

Good discussion.

[grappa]

Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

I have really enjoyed reading this thread and I too edit rosters on my own. I only own a psp, so I have to do all of the ratings and edits on my own. I really enjoyed Bahnzo's tool and am sad to see it go.

That being said, I have downloaded Bahnzo's spreadsheet, but cannot really figure out how to use it.

I see there has been talk about compiling a new spreadsheet that can be used that will calculate all of the ratings, as long as you enter them in.

I have about 5 teams left for a 1987 set and would love to do a 2010 season (because I'm a Giants fan) and then carry that season over into a projected 2011 season.

Has there or does anyone know if there will be a spreadsheet available and is there anything I can do to help.

Thanks a lot.

[grappa]

Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

I had a couple of questions about how people figure out some of the ratings. I am sorry if these have been answered, but I couldn't find them in the thread.

1.) Speed: I saw there was the Bill James Speed Rating multiplied by a modifier...what's the modifier that people are using?

2.) Baserunning ability: I have no idea how to go about calculating this.

3.) Baserunning aggressiveness: Same as above.

4.) Clutch: I am still a little unclear about how to go about clutch rating for pitcher and hitters. Just put the better players closer to 75 and the others closer to 55?

5.) Stamina: I thought I read that TT was finding that starters weren't pitching late enough into ball games and that relievers were getting too many outings.

Thanks ahead of time and sorry if I overlooked these answers if they are already posted.

[ImBaconBeckham]

Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

I had a couple of questions about how people figure out some of the ratings. I am sorry if these have been answered, but I couldn't find them in the thread.

1.) Speed: I saw there was the Bill James Speed Rating multiplied by a modifier...what's the modifier that people are using?

2.) Baserunning ability: I have no idea how to go about calculating this.

3.) Baserunning aggressiveness: Same as above.

4.) Clutch: I am still a little unclear about how to go about clutch rating for pitcher and hitters. Just put the better players closer to 75 and the others closer to 55?

5.) Stamina: I thought I read that TT was finding that starters weren't pitching late enough into ball games and that relievers were getting too many outings.

Thanks ahead of time and sorry if I overlooked these answers if they are already posted.

There's a clutch rating on FanGraphs, you can standard deviate all the samples from a certain amount of years and then devise a rating system off of that.

Or you can look at WPA/LI for pitchers (especially relievers).

The base running stuff I'm not too good with. There's no single stat you can really incorporate into a rating. The first instinct is to use SB and SB attempts, but we all know base running is a whole lot more than stealing bases.

[grappa]

Re: 2011 Projection-Based Rosters: Methods, Formulas, & Discussion ...

Thanks for the help. I'll check out the clutch rating.