Draft Manifesto

If the Sports Guy can release his playoff gambling manifesto and not worry about it being used against him, perhaps I can do the same with my fantasy baseball draft manifesto. Here goes.

1) Don’t draft a Utility guy like Travis Hafner or David Ortiz. Not only do these guys tend to be older and have ever-worsening contact skills and playing time, but they also keep you from taking advantage of some draft bargains. How many times have you been in a draft where a player is hanging around for 20 or 30 picks longer than you anticipated? I’m not talking the consensus 10th pick still around in the 4th round, but maybe a guy around 90 falling to 120 or 125. And you pass him by because you already have his position slot and your Utility slot is full? The Utility slot should be reserved for this type of occasion. In addition, if you’re not attached to your Utility player, you can keep churning through free agents and waiver wire pickups until someone drops a player they really shouldn’t have. Then, no matter the position he plays, you have a spot on your starting lineup for him.

2) Don’t draft a catcher in the top 100. These elite catchers just never seem to perform to their level. If Brian McCann is ranked 33rd preseason, he could have a solid season and still only be the 66th best player by the end of the year. Picking him 33rd is a terribly inefficient pick. It’s a brutal position on a player’s knees, and I’d bet catchers break down at almost the same rate as starting pitchers. I tend to go for catchers ranked somewhere between 100 and 200. Take a gamble on a catcher, don’t go for an established star. Let someone pick McCann, Joe Mauer, Russell Martin, and Victor Martinez 30 spots higher than their eventual end of season rank.

3) Don’t pick Carl Crawford. In the preseason, he always seems to be ranked in the top 10 or 12, and always seems to end up ranked 20th by the end of the year. I have no proof of this, it just seems like it.

4) Spread out your steals. Don’t count on a guy like Eric Byrnes or Brian Roberts for 50-plus, because a slight injury or change in philosophy could cut that number in half. Instead, find a bunch of guys that get 10 or 12 steals. A lot of times, opposing drafters won’t consider those steals as part of the value of a player, but having them will reduce the variance in your team’s output, and they can certainly add up.

5) Don’t worry so much about batting average and ERA in head-to-head leagues. As I showed in a previous column, being good in those categories doesn’t give you much of an advantage each week. On the other hand, if your team is truly excellent in runs, you’ll be rewarded with consistent head-to-head wins.

6) Use two basic principles to your advantage: regression to the mean, and aging. Almost every fantasy owner who mis-values a player will do so due to forgetting about those two concepts. Young players tend to get better, old players tend to get worse. Sounds simple when you say it, but how many people are going to draft Manny Ramirez as though his 2008 age 36 season is exactly what they should expect for 2009? And along a similar vein … don’t forget about regression to the mean. You have the tools on the stats section of this site to take a closer look at fantasy statistics. Be a critic, be skeptical, and stick to the null hypothesis: a career season from a player in his late 20s or anywhere in his 30s does not indicate a new true talent level, unless you have peripheral stats to back it up.


23 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Donald Trump
15 years ago

Dude, pardon my french, but you are really fucking my #### right now.  Please let me explain…
I clicked the link to your previous post, and found that it was excellent (your post was made during my December dark period for keeping up with baseball, so I missed it when you posted it).
I have actually done a similar analysis in my league for the past three seasons.  I used a different methodology than yours, and would like to discuss this with you.  I hope none of my leaguemates are reading this…
I will summarize how I came up with my category weightings.  I have taken everyones year end ranking in every category, and sorted the teams from highest to lowest (Thus, the top team shows 903 home runs on the season, the second highest shows 875, etc).  Then I calculate everyones winning percentage in that category over the season, using the formula Yahoo uses to calculate overall team winning percentage (thus, the team with the highest HR total actually had a 80% winning percentage in HR).  I then calculate the correlation between these two categories.  The correlation between these two data sets is what I use.  I have this data for three years, and use an average of all three years to try to get a smooth number.  This has given me a large advantage over my league mates when assessing player value.  I was somewhat displeased to see someone else writing about this, but now that you have would like to discuss our methods and results.
Here is how your numbers compare to mine, with your numbers first:
Runs: 82…83
HR: 87…76
RBI:75…66
SB:93…89
Ave:68…70
Three of the categories are almost exactly the same, R, SB, Ave.  Concerning HR, My data for 2006 and 2007 average to 81, still below your number, but my correlation for 2008 was only 66%.  Concerning RBI, my correlation averages to 70% for 2007 and 2008, but 2006 showed only 60%.  If I squint my eyes enough, I can see that we both rank SB the most consistent, with R and HR the next two, and RBI and Ave as the least consistent.
For pitching…
W:86…76
K:91….81
Sv:93…91
Era:66…81
Whip:76…74
Only two of our categories match up, Sv and Whip.  For wins, my correlations are 76, 60, 93 for 2006/07/08.  For K, I have 85,82,76, so all of my numbers here are lower than yours.  Era…86,86,71, so I show consistently more consistency in era than you do.  If I squint my eyes, I can see that we both have Sv as the most consistent, and K as the next most.

I an curious to know your thoughts about my methodology and results.

Also, I would like to discuss a more time consuming method which I have not done.  It would be interesting to see what everyones winning percentage would be if they had to play everyone every week, instead of just playing your single head-to-head opponent in a given week, and being subject to the luck of their week.  By this, I mean not gauging weather your record in HR in week 1 was 1-0 or 0-1, but what would your record in HR in week 1 have been if you had played every team that week.  In this analysis, while your actual record in HR in week 1 might have been 0-1, your record in home runs in week 1 when matched up against every team in the league might have actually been 7-4.  I thought about doing an analysis like this for every stat category, for all teams over the entire 22 week schedule, but this was a little too cumbersome.  I do think that this would give a clearer picture, as managers can change the appearance or strength of their teams over the course of the season, and this would more fully capture that.

Donald Trump
15 years ago

Excuse me, but my eyes glazed over when you were discussing Z score, as I have no idea what that is (shame on me).  I see that we actually had very similar methods aside from that.

Donald Trump
15 years ago

OK, I don’t really want to monopolize the comment space here, but need to address one of your points in this current post.  You say “Don’t draft a catcher in the top 100. These elite catchers just never seem to perform to their level. If Brian McCann is ranked 33rd preseason, he could have a solid season and still only be the 66th best player by the end of the year. Picking him 33rd is a terribly inefficient pick.”
I disagree.  It is all about value over replacement player, I think.  For an exaggeration to make the point, if you pick McCann 33rd, and he ends up the 66th best player over the course of the year, he can actually have been a great pick if the next most valuable catcher was ranked 787th over the course of the year.

Jeff Akston
15 years ago

I’ve been saying that about Carl Crawford for 5 years.  Absolutely! 

Your other points are also very valid.  I wish you would build this piece out.  It could easily be 2 or 3x as long.

Mike
15 years ago

Hi Donald, glad to see you entering the fantasy baseball field.  My methodology was very similar to yours, except instead of using home run totals (225, 213, etc), I used z-scores.  If I am remembering college stats class correctly, a z-score tells you how many standard deviations from the mean you are.  So if the mean for team HR is 175 with a standard deviation of 25, a team that hit 225 would have a z-score of 2.0.  150 HR would equate to -1.0 for a z-score.  I think your way of simply using the totals works just as well.

I excluded were totals and winning percentages from myself and from teams that mailed it in by the end of the season (I operationalized this in terms of the number of moves they made, to keep it from being subjective).  I used three seasons of stats as well, all for 12-team leagues.  I’m trying to get my hands on much, much more data from someone inside Yahoo, so fingers crossed!

As for your more extensive method, I think if we had enough leagues in our sample, the luck of who a team plays each week would wash out amongst the large data set.  Rather than spend the time to try to do that for a league, I’d suggest trying to get data from a friend or acquaintance’s league, to beef up the data set we’re using.

As for why some categories come out on top… I think SV and SB come out near the top because a few teams in each league willingly give up on those stats by midseason.  Consequently, from the All-Star break to the end of the year, the guy who is last in Saves (with 0) will of course have a .000 winning percentage in that stat… a perfect correlation that drags up the average across all teams in that stat.

As for catchers… I disagree.  If you pick McCann 33rd and he ends up 66th, and the next best catcher is 787th… you’re only doing well if someone drafted that guy well before the 787th pick.  But if he picked that guy in the right spot, right around 787th, and used his 34th pick on the guy who actually ended up 34th, you’re behind.

Jeff, thanks.  I’d love to make it longer, but was a little worried that readers wouldn’t be happy with the depth of analysis.  If folks enjoyed it though, I can write a follow-up and share some other rules, like why I never plan to draft a starting pitcher in the first 10 rounds, why I intentionally overrate closers, etc.

Mike
15 years ago

Oh – I should add, with respect to the z-scores versus totals…  I think if you’re comparing multiple leagues (especially with different settings), you should use z-scores as opposed to totals.  Imagine you were running this study across your league for 20 seasons.  The Winning Percentage associated with 200 team home runs in 1991 would be expected to be drastically different from that of 1999.  So there’s that seasonal impact, plus the impact of owner personalities.  Some owners like to eschew SB and focus on power stats, so if you have a lot of folks like that in a league one particular season, you need to adjust for that when comparing across other leagues.

Donald Trump
15 years ago

Mike, if you get that Yahoo! data, please do report on it.
Also, in your reply, when you say “As for why some categories come out on top”, it seems as though you are now talking about a roto league, whereas I thought your data was from a head to head league.

Donald Trump
15 years ago

to avoid any confusion related to my original post, when I said the top team got 903 HR, i meant runs, not HR.

Paul
15 years ago

I think this was the first time I’ve ever seen the word “operationalized” outside of a journal article.

Also, yea if you’re going across leagues and especially across time, Z-scores would be the way to go.

Donald Trump
15 years ago

I think there has been some confusion.  When you say “The Winning Percentage associated with 200 team home runs in 1991 would be expected to be drastically different from that of 1999”, this is not what I have done.  I created three separate analyses, one from each year.  Then I averaged the correlation of home runs to winning percentage in 2006 with that of 2007 and 2008.

Mike
15 years ago

I was referring to those categories being “on top” of the list of categories in order of highest correlation.

Gotcha, on averaging the three correlations.  You made that clear in your first comment, and for some reason I just didn’t process it right in my head.  That sounds like a fine way to go.

Paul, my college psych stats professor will be glad to know she had a lasting impact!

And believe me… if I ever get this Yahoo data, I’ll end up writing so many articles from it, people will be begging me to stop.  Or not.  Hopefully.

Andrew
15 years ago

I’m sorry, but you’re wrong here about catchers. Merely pointing out that their end-of-season ranks typically fall below where they were drafted does not tell the whole story. Scarcity is at play when it comes to the catcher position every single season. By that I mean, every team in your league must have a catcher (or 2), yet these players would not otherwise have been drafted if positional requirements did not exist. Once your run the numbers and compare the production of the elite catchers to that of a replacement-level catcher, they’re certainly worthy of a 4th or 5th round selection.

Jeff
15 years ago

I think the most important point when drafting is to look for value.  I don’t really understand how player rankings are generated in terms of fantasy VORP.  But I can see the point that if in terms of VORP a player ranks 60th and is being picked 30th (i.e. top catchers in Mike’s example), he is not a good value.  But,  if that player falls to you at a discount, you should probably take him.

I had Brian McCann “fall” to me at the end of round 4 in a 10-team NL-only league.  That was certainly a great “value” pick. 

On the other hand, if McCann and the other top catchers were the bizarro consensus top 1 -5 picks, anyone who bucked the norm and picked a true 1-5 player would come out ahead.

Jeff
15 years ago

So I guess the argument is whether in most leagues the top catchers are overrated relative to their fantasy VORP.  Mike is arguing yes based on the Yahoo Player Rater.

Derek Carty
15 years ago

Michael,
As one of the people who disagrees on that point, I think the disagreement from most (or at least from me) is on the point that the top catcher will only be ranked 66th at the end of the season even with position scarcity taken into consideration.  I think this is very far off.  The top catcher, using my evaluation methods, ranks in the top 15 or 20 of hitters.  Including pitchers, top 30 is probably about right.  I really don’t see how a catcher would only be 66th.  If you’re looking at raw stats, absolutely, but taking position into account, I just can’t see it.

Donald Trump
15 years ago

Mike:
referring to catchers, you say “I’m operating on the assumption that Yahoo’s Rank and o-rank listings take position into account”.
I disagree with this.  Like you, I have not been able to reverse engineer this, but probably like you, I have run the regressions on stats and yahoo rank, and have a decent idea if how yahoo ranks players.  Although now that I think of it, I did not take position into account, and only ran a regression on stats.  Hmmmm.  I was going to say that yahoo does not take position into account, but i will need to do another regression with position taken into account to determine that, and yahoo does not have the player page accessible in last years fantasy baseball leagues.  If you have any definitive info regarding yahoo taking position into account, please let us know.

Michael Lerra
15 years ago

Well, two points.  First, just because someone ends up 15th or 20th, doesn’t mean you know which one it will be.  I can tell you the league leader will hit around 45 HR next year, but I don’t know if it’ll be Ortiz, ARod, Manny, Pujols, Dunn, Howard, etc.

Second, I mean, this all hinges on your evaluation methods, right?  So we’re kind of arguing apples and oranges.  My evaluation method in this instance is Yahoo’s “Rank”, which I’ve been unable to reverse-engineer, but I’ve nonetheless found to be a very accurate measure of a player’s contribution to a team – position included.  And consistently, I do not see top-tier catchers ending the season with ranks anywhere near where they were drafted.

Mike
15 years ago

Jeff, precisely.  The other aspect of my argument, which I think holds when looking at players in terms of absolute production (regardless of position) is that there are enough catchers with production in the top 250 for each team to get one.  People seem to think that the replacement level for catchers is late-80’s Rich Gedman, but it’s not.  There’s plenty of serviceable guys.  Now if each team carries and starts 2 catchers, that could change things.  I’ve never been in such a league though, so I’ve never really thought about it.

Mr Trump… I actually never tried to run a regression.  What I did was look for players with very little playing time, who had similar stats.  For example, two guys batting .500, one with an RBI, one with a Run.  Whichever was ranked higher would tell you whether an RBI or a Run was worth more (according to Yahoo).  I *think* I saw at some point, a 2B or C with the same exact stats as a 1B (or something like that), but ranked higher.  I could totally be making this up in my head (“illusory correlation”, I think may be the term), but that’s what I remember. And it’s annoying that the Yahoo ranks aren’t published anymore.

Almost everything in the actual article above is purely anecdotal… which is in the spirit of the Sport’s Guy’s manifesto(s), but that doesn’t really cut it for the THT community.  So while I know I’m totally just arguing theory here, I do believe it – I’m not just trying to get a rise out of people.  And I am absolutely willing to be wrong about it, though I should say I haven’t been wrong about something since the fall of 1996.

Derek Carty
15 years ago

That’ll do it, Mike. My premise is based on two catcher leagues. I completely forgot that Yahoo default leagues do one catcher. All the leagues I play in do two. I guess we’re both right smile

Donald Trump
15 years ago

Mike:
In the 3 leagues that you studied, how many pitchers are active each week?  I would think that the more pitchers you have active each week, the higher the correlations would be.  I just want to see how applicable your numbers are to my leagues.

Mike
15 years ago

Hah, OK then!  I’ve played in 6 leagues I think, all of which have been 1-catcher (Yahoo-default) leagues.  I haven’t even thought about what a 2-catcher system would be like.  I think I would puke if I had to draft a guy who’s splitting time with a Molina or soemthing.

In my league, there were 7 active pitchers: 2 SP, 2 RP, 3 P.  Most teams carried one bench hitter and 4 bench pitchers, all of whom were SP, and rotated them into the SP and P slots to maximize their Ks and Wins.

Derek Carty
15 years ago

Yeah, I generally play in 2-catcher leagues with 9 or 10 pitchers.  It’s not quite as bad as a guy splitting time with a Molina, but the drop-off from #1 to #24 is pretty severe, which is why I like getting at least one early.

Michael Lerra
15 years ago

Yeah, I still disagree.  I should say, I’m operating on the assumption that Yahoo’s Rank and o-rank listings take position into account, and realize that 40 HR from a catcher is worth more than 40 HR from a 1B.

I tried to avoid hypotheticals, but at this point maybe I should see if it helps express myself.  I’ve heard a fair bit of disagreement about this point.

Imagine that all drafters but yourself are slaves to Yahoo’s preseason rank.  You, being an astute THT reader, know the “true” end of season rankings of players.  Brian McCann is ranked 33rd going into the draft, but you know he’s really only the 66th best player, even after accounting for the production he gives you at catcher.  Why would you take him at 33rd, and someone like an Ellsbury at 150th (his draft rank, and his ultimate final standing)?  You’d be using a 33 pick and a 150 pick to get a 66 production and a 150 production.  I’d say instead, use the 33rd pick on Adam Dunn, and the 150th pick on Geovany Soto, and you’ll actually get 33rd and 150th production out of each.

If the ranks of catchers were 33, 40, 55, 60, 700, 702, 735, 780, 900, 925, etc… then it’s a different story.  The 5th best catcher is not within the number of players drafted.  So if you last-pick Varitek with your 240th pick, but he gives you the 700th best production, you’re being inefficient and should maybe have looked to grab one of the top-tier guys.  But in a 12-team league, with 20 to 25 draft rounds, I think there are 12 catchers who fall within those draft slots.  So why be inefficient with your picks?