How do you turn stats into ratings? - Fantasy Baseball Cafe 2014 Fantasy Baseball Cafe
100% Deposit Bonus for Cafe Members!

Return to Beginner's Forum

How do you turn stats into ratings?

How do you turn stats into ratings?

Postby st3v3k4hn » Sun Apr 15, 2012 8:41 pm

This is my 1st season, I've done lots of reading but cant seem to find anything talking about how to rank players based on stats. I did my draft relying on others people's cheat sheets, but looking at the stats / rankings I can't always say I agree with the rankings. And looking at stats for guys now that the season has started, I'm a little lost on how to rank based on the stats. (Apologies if this has been discussed before, if so pls just point me to the info).

As an example, looking at available players in my Yahoo league, the top ranked guy (season so far) has:
8 runs, 1 HR, 2 RBIs, 5 SBs, and a .286 Avg. The next best guy is:
5 runs, 3 HRs, 9 RBIs, 0 SBs, and .379 Avg

Obviously these are very different type players, and if you had a clear need for runs and SBs you would choose player 1; if you need RBIs and AVG you would pick player 2. But in a pure sense, who will add more to a team? (I know you shouldn't make too many decisions based on so few games, but again this is just to probe the question.) I'd like to find a simple way of identifying the better player. My 1st thought was just to add the totals (I'm going to ignore Avg for now to make this simpler):
Player 1: 8+1+2+5 = 16
Player 2: 5+3+9= 17

So by that very simple method player 1 is "better." But then looking at the stats, I started thinking maybe HRs and SBs should be worth more since they are scarcer. In fact, since Runs and RBIs generally are about 3x HRs and SBs (think a guy with 90 Runs, 90 RBIs, 30 HRs and 30 SBs) I thought maybe I should multiply the HRs and SBs by 3 to make them more important. Altho that really seems like too much. Then I thought well maybe thats backwards anyway; since you need so many more Runs and RBIs maybe those are the most valuable? So, as you can see I've totally confused myself. Can anyone help a brother out? Thanks!
st3v3k4hn
Softball Supervisor
Softball Supervisor


Posts: 68
(Past Year: 1)
Joined: 26 Mar 2012
Home Cafe: Baseball

Re: How do you turn stats into ratings?

Postby bigh0rt » Mon Apr 16, 2012 12:57 pm

The short answer is there is no direct formula that anybody can agree on with respect to ranking a player, so there's a lot of subjectivity that goes into it. The good news is that once you get into the season, the only thing you need to focus on is acquiring the stats that your team requires, which you indicated early on in your post. The thing about RBI and R is that all we can do is predict opportunity based on role and ability to project how many times a player will score or drive a teammate in -- Some of the guys who have scored a bunch or driven a bunch of Runs in these past two weeks may fall off a cliff due to their role on their team or their ability over a long season. So, if your team is deficient in Runs, you're going to look at a player in the best position to score them -- a player batting low in a weak lineup is going to struggle to score runs even if they're playing pretty well; it's just the nature of the beast. Now, if suddenly their manager recognizes their production and bumps them higher in the order (you'll see guys move from 7th to 1st/2nd cause of an injury to a teammate or just playing their way there) and suddenly you've got a recipe for potential big return. But as far as weighing players against one another, the brunt of that work is done pre-draft. For the rest of this season it's a matter of filling in the statistical gaps in your team and increasing the positive ones.
Image
bigh0rt
Hall of Fame Hero
Hall of Fame Hero

User avatar
CafeholicCafe WriterGraphics ExpertMock(ing) DrafterEagle EyeInnovative MemberCafe MusketeerWeb SupporterPick 3 Weekly Winner
Posts: 24818
(Past Year: 348)
Joined: 3 Jun 2005
Home Cafe: Baseball
Location: Crowding The Plate

Re: How do you turn stats into ratings?

Postby st3v3k4hn » Mon Apr 16, 2012 2:54 pm

Thanks for your reply. Its helpful, and mostly true I know. Still, I would like to know how people (or ESPN or Yahoo or whoever) rank based on stats. And anyway, right now I am not trying to boost particular stats; I am trying to replace Ellsbury. Impossible, but you gotta try. At the moment the best guys available in my league (AL only) are (using ESPN projections and ratings among OFs):

Alex Rios (49): 73 Runs, 16 HRs, 63 RBIs, 20 SBs, 304 OBP
Ryan Rayburn (56): 53 Runs, 15 HRs, 52 RBIs, 2 SBs, 303 OBP
Austin Jackson (62): 90 Runs, 8 HRs, 47 RBIs, 23 SBs, 313 OBP
Brennan Boesch (64): 66 Runs, 17 HRs, 61 RBIs, 7 SBs, 321 OBP

Just eyeballing this list, its pretty hard to say who would add the most to a team, altho I think its pretty clear that Rayburn would add the least (fewest runs and SBs and near bottom for RBIs). And how is Rayburn rated higher than Boesch, who beats him in every category???

Steve
st3v3k4hn
Softball Supervisor
Softball Supervisor


Posts: 68
(Past Year: 1)
Joined: 26 Mar 2012
Home Cafe: Baseball

Re: How do you turn stats into ratings?

Postby jefferey13 » Mon Apr 16, 2012 9:15 pm

I just looked around for a second, I didn't delve in too deep into how the ratings are calculated, but here is an ESPN article for basketball that talks about how they calculate PR. The same ideas should apply to baseball.

http://sports.espn.go.com/fantasy/baske ... arater1101
Image
jefferey13
Minor League Mentor
Minor League Mentor

User avatar

Posts: 778
Joined: 13 Mar 2012
Home Cafe: Baseball

Re: How do you turn stats into ratings?

Postby StateStreet » Mon Apr 16, 2012 11:44 pm

This isn't really answering your basic question, but of those available players, I'd definitely go with Austin Jackson. He has been red hot to start the season and has had much better looking at-bats than he did last year. He has a ton of potential to score a lot of runs hitting at the top of that Detroit lineup.
StateStreet
Softball Supervisor
Softball Supervisor

User avatar

Posts: 32
Joined: 11 Apr 2012
Home Cafe: Baseball

Re: How do you turn stats into ratings?

Postby st3v3k4hn » Tue Apr 17, 2012 3:07 pm

jefferey13 wrote:I just looked around for a second, I didn't delve in too deep into how the ratings are calculated, but here is an ESPN article for basketball that talks about how they calculate PR. The same ideas should apply to baseball.
http://sports.espn.go.com/fantasy/baske ... arater1101


Thanks. I looked at that and then searched for more info about player rater and found a bunch of helpful stuff. It looks like the basic idea of "normalizing" the stats is actually ok. But most people that do it use exact formulas in spreadsheets. I am looking for quick and dirty (I play a lot of poker, poker players have lots of shortcuts to quickly get approximate odds, I am looking for something like that).

Re the standard 5 hitting stats, its convenient because Runs and RBIs are fairly easy to come by and HRs and SBs are fairly scarce; that makes HRs and SBs more valuable. You can argue about how much more, and which one should be weighted heavier, but for the purposes of quick and dirty I think you can weigh them the same. I'm proposing a simple method like the following:

(Runs + RBIs) + n*(Hrs + SBs)

Where n is the muliplier. One thing thats really nice about this is that each sum includes a "power stat" (RBIs, HRs) and a "speed stat" (Runs, SBs) so you don't end up overvaluing one or the other. The problem is how to set n. Guys like Miggy need an n of >6 (32 Hrs/SBs, 216 Runs/RBIs). Guys like Bourn and Crisp are closer to n=2. It looks like 3 works pretty well for the majority of guys. So looking back at the guys I posted earlier:

Alex Rios (49): 73 Runs, 16 HRs, 63 RBIs, 20 SBs, 304 OBP ----------> (73+63) + 3(16+20) = 136+108 = 244
Ryan Rayburn (56): 53 Runs, 15 HRs, 52 RBIs, 2 SBs, 303 OBP -------> (53+52) + 3(15+2) = 105 + 51 = 156
Austin Jackson (62): 90 Runs, 8 HRs, 47 RBIs, 23 SBs, 313 OBP -----> (90+47) + 3(8+23) = 137 + 93 = 230
Brennan Boesch (64): 66 Runs, 17 HRs, 61 RBIs, 7 SBs, 321 OBP ----> (66+61) + 3(17+7) = 127 + 72 = 199

Which rates Rios highest, then Jackson, Boesch and Rayburn. Thats pretty much how I would rate it.

Admittedly, this is a pretty trivial case, because the real value comes in cases where the guys have a lot of HRs/SBs compared to Runs/RBIs. (In fact, in this case the multiplier makes no difference at all - just adding the stats gives you the same results. But the multiplier will make a big difference when evaluating guys like Crisp or Jay Bruce.) Anyone think there is any value to this? I'm thinking this would be handy anytime you need to make a decision pretty quickly, ie in a draft when you have to pick a a given position player but you're past all the guys you really liked. The idea is just to be able to make quick decisions based on stats rather than on Yahoo or ESPN rankings. I realize I ignored BA/OBP, thats because I haven't found a quick and dirty way to estimate its impact yet, and I don't want to waste my time if people think this has no value. Thanks to all...
st3v3k4hn
Softball Supervisor
Softball Supervisor


Posts: 68
(Past Year: 1)
Joined: 26 Mar 2012
Home Cafe: Baseball

Re: How do you turn stats into ratings?

Postby Fenris-77 » Wed Apr 18, 2012 10:50 am

Quick and dirty is fine, but it needs a smidge of context too. For example, right now I'd take Jackson over Rios regardless of what the stats say because Jackson is in such a sweet spot hitting lead off in front of the Bash Brothers v2.0. Rios is a solid bounce back candidate this year, but I'm not rostering him over Jackson at this point.

From a stats standpoint I think you'd better off adjusting n for each separate stat, rather than lumping them into two groups like you have. Site like BBM that value various stats in order to rank players are mostly using Z scores. That's not so quick and dirty of course. The problem with your model is that HRs and SBs aren't created equal. HRs are pretty directly tied to another scoring cat (RBIs) while SB exist in more of a vacuum. Well, not exactly since Runs can be linked to SB opportunities, but that's another matter. The point here is that you can find guys who's sole skill is padding your SB totals (Rajai Davis anyone?) while you don't find HR guys like that at all.

I'm not sure I agree with your methodology in general, but I'll play along. I think I'd rather do something like this: (Run + 2(SB)) + (RBI + 3(HR)) to account for my valuing HRs a little more than SBs in a straight player ranking. Although that still stinks a little because I'm still picking multipliers at random. Another way to go about it would be to add HRs and SBs and then multiply by OPS. That's not as directly tied to RBI and Run totals as your method, but it still takes into account the ability to get on base and hit when it matters.

Rios (16+20)*.711 = 25.6
Radburn (15+2)*.742 = 12.6
Jackson (8+23)*.674 = 21.6
Boesch (17+7)*.765 = 18.4

Interestingly that ranks them the same as your system, but without having to pick an n score at more or less random. This has some of the same conceptual issues as I have with your original system of course, but in either case a little common sense to finish the process off solves a lot of problems.
Last edited by Fenris-77 on Wed Apr 18, 2012 2:43 pm, edited 1 time in total.
Fenris-77
Major League Manager
Major League Manager

User avatar

Posts: 1688
(Past Year: 51)
Joined: 10 Jul 2008
Home Cafe: Basketball
Location: SAGNOF

Re: How do you turn stats into ratings?

Postby st3v3k4hn » Wed Apr 18, 2012 2:39 pm

Fenris-77 wrote:Quick and dirty is fine, but it needs a smidge of context too. For example, right now I'd take Jackson over Rios regardless of what the stats say because Jackson is in such a sweet spot hitting lead off in front of the Bash Brothers v2.0. Rios is a solid bounce back candidate this year, but I'm not rostering him over Jackson at this point.

Yeah, i actually had added Jackson earlier and just dropped Crisp and added Nolan Reimold (trying to ride his streak). I figure no one available can actually replace Ells so I'm gonna try riding hot players. I hope I don't regret dropping Crisp but he's been a huge bust so far and I already have a base stealer (Dee Gordon) and I don't think I can afford to carry 2 fast guys who don't add too much more. And I never really liked him much...
From a stats standpoint I think you'd better off adjusting n for each separate stat, rather than lumping them into two groups like you have. Site like BBM that value various stats in order to rank players are mostly using [urlhttp://statistics-help-for-students.com/What_are_Z_scores.htm]Z scores[/url]. That's not so quick and dirty of course. The problem with your model is that HRs and SBs aren't created equal. HRs are pretty directly tied to another scoring cat (RBIs) while SB exist in more of a vacuum. Well, not exactly since Runs can be linked to SB opportunities, but that's another matter. The point here is that you can find guys who's sole skill is padding your SB totals (Rajai Davis anyone?) while you don't find HR guys like that at all.

Yes I've seen the standard deviation / z score stuff and it looks great but obviously not quick and dirty as you said. Re HRs and SBs being not created equal, thats tru-ish. I actually started with the dominance factor article I saw on Bleacher Report (http://bleacherreport.com/articles/1092 ... to-hitters). He choose his dominating stats as .320, 40HR, 120RBI, 120R, 50SB. Which is factor of 3 for HRs and slightly less for SBs. But a more realistic goal might be 90/90 and 30/30, which is straight 3x.
I'd also like to disagree with the "you can find speed but nor power late" thing. Again, in our league, Soriano and Vernon Wells are both available and both should give you mid-20s HRs. There are more guys who will give you mid-20s steals or better (Davis and Pagan), but both options are out there (at least for now).
I'm not sure I agree with your methodology in general, but I'll play along. I think I'd rather do something like this: (Run + 2(SB)) + (RBI + 3(HR)) to account for my valuing HRs a little more than SBs in a straight player ranking. Although that still stinks a little because I'm still picking multipliers at random.

Thank you for playing along! As I explained above it wasn't really random, but I think you're right its still not exactly scientific, is it? :) I like counting them equal. I think that the idea that HRs are more important because they generate more other stats only counts if you dont count the other stats in your formula - otherwise you are counting that twice.
Another way to go about it would be to add HRs and SBs and then multiply by OPS. That's not as directly tied to RBI and Run totals as your method, but it still takes into account the ability to get on base and hit when it matters.

Rios (16+20)*.711 = 25.6
Radburn (15+2)*.742 = 12.6
Jackson (8+23)*.674 = 21.6
Boesch (17+7)*.765 = 18.4

Interestingly that ranks them the same as your system, but without having to pick an n score at more or less random. This has some of the same conceptual issues as I have with your original system of course, but in either case a little common sense to finish the process off solves a lot of problems.

Thats a really interesting approach. Did you make it up? The only problem here is that OPS isn't always a stat thats easy to get (ie its not listed during the draft I think?). Anyway, thanks for the feedback, I'm new to all this and just thinking out loud and trying to learn so all help is greatly appreciate...
st3v3k4hn
Softball Supervisor
Softball Supervisor


Posts: 68
(Past Year: 1)
Joined: 26 Mar 2012
Home Cafe: Baseball

Re: How do you turn stats into ratings?

Postby Fenris-77 » Wed Apr 18, 2012 2:51 pm

Yeah, I just made that up, but it seemed about right conceptually, at least compared to your method (both as quick and dirty). I'm not completely in love with ESPNs projections, but they do project OPS, so you could use that as a resource. I really like OPS a general player eval stat, which is what made me think of using it here.

I'm not going to get hung up about the exact weighting for the counting stats either. If you like a straight three then role with it. Personally, I don't think that matches actual MLB production as well as I'd like, but whatevs. There are a lot of guys who give you SB and nothing else though, while those HR guys will generally give you at least RBIs as well. That true in general and wasn't really meant to be a comment on your league specific waiver wire. It's also generally the case that the SB base guys are more often on waivers too. If a guy like Wells keeps hitting, he won't be on waivers very long, but a guy like Davis will be on and off waivers all year.
Fenris-77
Major League Manager
Major League Manager

User avatar

Posts: 1688
(Past Year: 51)
Joined: 10 Jul 2008
Home Cafe: Basketball
Location: SAGNOF

Re: How do you turn stats into ratings?

Postby hot4tx » Fri Apr 20, 2012 10:26 am

This doesn't answer your question as to how Yahoo or ESPN rank players, but here's a couple equations I've used to rank players in 5x5 (based on forecasted stats, not just on the first 10 games stats of course) that could be used for the purposes you're talking about...

Hitters (higher number is better)
Tom Tango eq: HR + SB + R/3 + RBI/3 + (hits - 0.27 x at bats)
wTango eq: HR + SB/1.5 + R/3.5 + RBI/3.5 + 70% x (hits - 0.27 x at bats)
J35J eq: (at bats/9) + (batting ave x 700) + (HR x 1.5) + R/1.5 + RBI/1.5 + (SB x 1.2)

Pitchers (lower number is better)
J35J eq (starters): (ERA x 60) + (WHIP x 100) + (player age) - [ (IP/5) + (W x 2) + (K/2) ]
J35J eq (relievers): (ERA x 60) + (WHIP x 100) - [ (SV x 3) + (K/2) ]
wJ35J eq (starters): (ERA x 60) + (WHIP x 100) + (player age) - [ (IP/3) + (W x 2) + (K/2) ]
wJ35J (relievers): (ERA x 60) + (WHIP x 100) - [ (SV x 2) + (K) ]

I created the wTango and wJ35J eqs to tailor them more to what I care about in HTH leagues.

You could use these to rank players stats so far for the current year. I use them to value players in the preseason based on projected stats. In order to account for positional scarcity differences in player rank (a C or a SS that has the same stats as a 1B should be ranked higher) I rank all players in each position, sort them, and then come up with a dollar value for each (which gets complicated), then I rank all players based on dollar value and that's my personal rankings. When drafting I usually do some combination of each players rank in my personal rankings and in the hosts rankings (yahoo, ESPN, etc) or rankings that I think most of my competitors will use.
hot4tx
Major League Manager
Major League Manager

User avatar

Posts: 1933
(Past Year: 135)
Joined: 16 Feb 2007
Home Cafe: Baseball

Next

Return to Beginner's Forum

Who is online

Users browsing this forum: No registered users and 0 guests

Forums Articles & Tips Sleepers Rankings Leagues


Today's Games
Wednesday, Oct. 22
(All times are EST, weather icons show forecast for game time)

San Francisco at Kansas City
(8:07 pm)

  • Fantasy Baseball
  • Article Submissions
  • Privacy Statement
  • Site Survey 
  • Contact