Like a Scotch drinker, I’ve found my taste for baseball projections has matured over the years.  Where my initial taste was weaned on Dewar’s-quality projections like ESPN, Yahoo!, or some $4.95 magazine off the newsstand, I now hold out for premium, single-malt varieties like PECOTA/Baseball Prospectus and Ron Shandler.  I recommend buying both of their projections online as you can get their projections in spreadsheet form.

While peers suggest I try other high-quality (and free) projections like CHONE, ZiPs, etc., I’ve put projection experimenting on hold to tackle a greater quest – one that could benefit our site’s loyal readers and the fortunate souls who get redirected here by a search engine.

The challenge is answering the question “How do you convert player projections into rankings?”  As once you’ve settled on your projections, there are several key pre-draft considerations that need to take place to ensure success:

  1. Value of Player Based on Position Depth – e.g, how much does a player’s value increase/decrease based on the other available options for that position?
  2. Value of Players in Different Positions – e.g., how much do you sacrifice on a player’s total stats because they play 2B vs. 1B
  3. Value of a Player’s Stat Mix – e.g., how do you compare the value of 40/120/10 (HR/RBI/SB) vs. 15/75/40?
  4. Value of Hitter vs Pitcher Stats – e.g., how do you compare A-Rod vs. Santana?
  5. Value of a Player By You vs. Others – e.g., how long can you wait before picking a player?

(Note:  Risk and health are other key considerations but they ideally should be factored into the projections – i.e., Rich Harden shouldn’t be projected at 200 IP)

While a solution for the above factors appears complex, the concept behind how to do it is rather simple:  Convert all the statistics to the same metric (think money – it’s real easy to compare 10 dollars vs 15 Euro vs. 2000 Yen if you convert the Euro and Yen to dollars).  This is the underlying concept behind Bill James’ Win Shares.

So what metric makes the most sense for fantasy baseball?  Where real baseball success is measured in Wins, fantasy baseball success is measured in points.

Hence, “Point Shares”

Please click for our inaugural edition of Fantasy Baseball Point Shares for 2008.  I’m going to refrain from a drawn-out explanation of the methodology.  The important parts to understand are:

  1. Point shares represent the estimated impact on a team’s points by substituting a player for the average drafted player at his position on a team filled with average players.  So in a 10 team league, this team would otherwise earn 5.5 points per category (55 points).  Substituting one of those average pitchers (approx. Ian Snell) for Johan Santana would net an approximated gain of 7.8 points to 62.8.
  2. To account for a hitter’s value outside their position (The utility spot, the fact that a SS HR is worth the same as an OF HR), hitters receive 2/3 of points value based on their stats vs. the average drafted player in their position and 1/3 of points value based on the average drafted hitter.
  3. Since pitching positions can be filled with starters or relievers, player value was adjusted.  Starting pitcher values are 75% based on average drafted starting pitcher, 25% on average drafted pitcher.  Relievers are 40% on average drafted reliever, 60% on average drafted pitcher.
  4. Hitters are placed at their most valuable position where they are 20 games eligible.  Their rank/value at other positions they are eligible (down through expected eligibility like Ryan Braun in OF) is listed lower down in the spreadsheet.
  5. Two versions are included:  a 10 team, 5×5, MLB universe, C/1B/2B/SS/3B/CI/MI/5 OF/Util/9P and a 12 team, 5×5, MLB Universe with 2C/1B/2B/SS/3B/CI/MI/5 OF/Util/9P.

As with any player ratings system – especially one this ambitious – the standard question would be “How do you test this?”.  The beauty of this methodology is it was relatively easy to test.  I took 7 drafts off of Mock Draft Central and calculated the rankings based on the underlying projections (weighted model of PECOTA and Shandler) and the Point Shares.

After making a few adjustments, the results of the test were very promising – Point Shares predicted total team points within +/- 2 point for 45 of the 70 teams.  Another 18 were predicted within +/- 5 points.  Only 1 team fell outside of +/- 7 points.

On a category-by-category basis, the Point Shares correlate well with the total team stats.  For the hitting stats, the team Point Shares correlated at 97+% with the total stats.  For pitchers, Saves, ERA, and WHIP correlated at 90+% while Wins and Strikeouts were at 90% aside from one league where the projections tanked.  Why did the pitching stats not do as well as the hitting stats?  It is because of the random mix of starters and relievers who – unlike hitters – have vastly different counting stats.   ERA and WHIP proved most successful because they could be weighted by innings pitched.

Look out for future posts referencing these Point Shares and probably make some tweaks along the way – especially if we get revised player projections.

We also want to state clearly that this is NOT a recommended draft ordering.  The main reason is that it doesn’t factor in the 5th pre-draft consideration mentioned earlier – the “Value of a Player By You vs. Others”.  Yes, I believe Peavy is worthy of a top 5 pick but if you can get him in the 2nd round or possibly the 3rd round, by all means wait.  Average Draft Position stats are the one piece of valuable information you can get from Yahoo!, ESPN, etc.  If you’re playing in an advanced league, you may want to use those on Mock Draft Central (requires subscription).

Also note that some of the differences aren’t statistically relevant.  If you like Jose Reyes at 3.65 over Ryan Howard at 3.72, go with your gut b/c it’s a virtual pick’em anyway.

So use this as your rankings base then make adjustments based on your preferences and your feel for your fellow drafters.  And good luck…

 
  1. Lou Poulas

    Lou Poulas says:
    (link)

    You mentioned risk, but one thing that should also be considered is the general risk with any pitcher, which really precludes them from a 1st round pick in my mind. They go down to injuries more than any other position player. They also go done to severe injuries more frequently.

    Think of last few years where pitcher injuries killed some fantasy teams – Chris Carpenter, BJ Ryan, and to a lesser extent Chris Ray, Josh Johnson, Randy Johnson, and Freddy Garcia. This is in 2007 alone.

    To me, they are just too risky to draft high and draft high frequently.

  2. josh says:
    (link)

    i’m one of the fortunate souls who stumbled upon this site from a search engine. i pay for a couple of fantasy sites, but this is the first one i check each day.

    does this sheet officially make you a proponent of drafting pitchers in the first 8 or 10 rounds? or do you think the general inconsistency of pitchers should bring them down a lot? even so, this tells me that historically healthy and consistent pitchers like santana and webb are undervalued in drafts. would you agree?

    im going to try going after these guys in my next several mock drafts, using your list with bpro’s attrition rates

    thanks again for all your hard work. this is seriously awesome

  3. trajan says:
    (link)

    As a regular reader of your former “Fantasy Baseball Blog” I am wondering how much of the methodology involved in your Point Shares system incorporates previous research from your investigation into the validity of the ESPN Player Rater. I was highly intrigued by the three postings you did back in the fall and even downloaded your player rater spreadsheet. This looks very similar – is it based on the same principles?

    Also, you mention that this system is based on a composite of several projection systems. I am assuming that you used PECOTA and Shandler (since you mentioned them), but are any others included?

    Thanks for putting in such quality work!

  4. Hey Patrick –
    Agreed that projections can only be relied upon for so much – especially for younger players with less track record or players who have recently changed teams/leagues.

    So I definitely factor those things in when doing my personal rankings.

    My Point Shares methodology is projection-agnostic though. As long as I put in projections and determine the drafted universe, it’ll crank out estimates.

    The focus for my drafting this year is to avoid the drafting biases that naturally arise – for instance, I think most players overrate HR/SB and underrate R/RBI. Not saying the first two aren’t slightly more predictable, but drafting a leadoff hitter type at 70/20/70/30 appears more valuable than a 95/30/95/5 guy because of an SB bias…

    And don’t get me started about the biggest shortcoming in fantasy player analysis – how to rate pitchers vs. hitters….

    Guess it’s worth seeing how I do this year in fantasy leagues before investing too much pride in it… :)

  5. Luke says:
    (link)

    Hey, Rudy. I found this little gem you got here and I am using it as a big piece of my draft preparation, along with my ADP values and own personal rankings for a master spreadsheet.

    My problem is that I’m in a 13-team 5X5 league that also includes a DH and 8 pitchers instead of 9. So with those factors, I’m sure the point shares you have put in place would be slightly modified but I was curious if you could point me to a way to calculate these point shares with my own variables such as a 13-team league.

    I think these point shares are the best piece of information I’ve found on this great internet and I really appreciate your work here.

  6. Hey Luke –
    Thanks. Glad the Point Shares have been of help.

    Man, there are so many fantasy baseball variations out there. How do you end up with 13 teams, a DH (do they have to be DH-eligible?), and 8 pitchers? LOL. At least it isn’t like the H2H that Grey is in that considers 3B as valuable as HRs.

    The calculations are pretty complex and involve drafting a whole ‘drafted’ universe to understand league stat totals and then doing some testing to determine the stat increments for each point (e.g., how many HRs represent a point for a team).

    I’d use the 12 team Point Shares, discount catchers slightly (b/c it’s based on 2 C), and ignore the 8 vs. 9 pitcher difference. The ADP information is great – I use it too so I don’t draft guys earlier than i need to.

    Good luck!
    Rudy

  7. Luke says:
    (link)

    Excellent! Thanks for the tips. I’ll factor that in.

    The 13 team thing is a bit weird. This is a new league for me so I’m just kinda going along with things. So with the slight differences in each of my leagues, it causes a few little tweaks here and there to make sure I’m prepared.

    I’ve got this whole giant spreadsheet now full of numbers like I mentioned. For a while there, I had too much information and that was a bad thing. But, then I found your tool here and it really put everything I want into one place. So, by factoring in a nice ADP tool I found out there as well, it all works out to be some very powerful information.

    Hopefully all of these calculations and head-scratching will lead me on a path to fantasy baseball glory.

    Thanks again.

  8. Nick says:
    (link)

    Wow. I’ve been pondering this very question for some time now, and it’s good to see that others are on the same page. Very nice work. Question: if you have access to PECOTAs, you’re likely familiar with the PFM (BP’s draft ranking tool). Any idea on how it differs from your system? The PFM appears to work off of replacement level vs. average, and assigns dollar values as positive until you get to replacement level.

    Average is more interesting to me, however, as it can help to answer when starting a player is actually hurting you. Something doesn’t seem to add up with your numbers, though (or maybe I’m misunderstanding something). According to the totals, there are only 35 pitchers with positive values. My understanding is that that means that 35/90 pitchers are above average? That’s by definition incorrect, isn’t it? I suppose it could be that the top 35 pitchers contribute 50% of the value, statistically, which would make sense. So if that’s the case, then rather than picking a pitcher with a negative point share total, is it better to start no one in that active spot? That seems counterintuitive, but maybe it’s true. Any thoughts?

  9. Hey Nick –
    I’ve played with BP’s Player Forecast Manager and – as a snake drafter – it just doesn’t work for me. I’m not sure if it’s using VORP instead of average (each has its own issues) but I just don’t buy the results. $62 for Reyes in a 12 team/$260 league vs $47 for A-Rod?

    I don’t like being given one big $ figure because it doesn’t tell me the value of each of the statistics. I find this important since the purchase of Reyes makes Brian Roberts less valuable since there’s only so many SB points to be had. The Point Shares can adjust for that letting you see that Reyes 3.6 (in a ten team league) means that just having a league average SB crew for the rest of your squad nets you 9.1 points. Same rationale why

    As for pitchers, it’s true only 35 of the presumed 90 pitching slots in a 10 team MLB league provide ‘above average’ performance. I don’t know if I’d quite state it that they provided 50% of the value though. When you create a blended average, you could have any number of people on each side of the average. Only choosing a median value would guarantee a perfect 45/45 split.

    One fault of using the average is that it means negative players DO have value. The assumption here is that you need to fill in your whole roster. Almost every hitter provides some value as long as he gets ABs (vs an empty slot). Pitching has two ratios/averages so an empty slot may be preferable to having livan hernandez funkify your ERA/WHIP.

    So think of below average players as trying to acquire the lesser evil. Each one chips away at the above average performance of your top players. You’re just hoping they don’t drag you down too far.

    Why I’m feeling strong about Point Shares is that I’ve tested it a number of times. If you want to replicate it, just take your league’s draft (or a mock draft) and calculate the standings based on 2008 projections (PECOTA recommended). Then run it using Point Shares. My tests showed it came quite close to predicting the standings. So I feel confident that it does a good job at valuing each player contribution. I think it would do a better job than BP’s PFM $ because it couldn’t account for cases where you overinvested in one category vs. another.

    Hope this helps…

  10. Ignore ‘same rationale why’ in 2nd paragraph…

    One more note: If you got any value out of the spreadsheet, the post, the comments, etc., please click the Ballhype button under the article. If you haven’t been to Ballhype, it’s like a Digg for sports news. Great site and one that helps us spread the word about our little site. I also recommend using the site to find other great sports articles….

Comments are closed.