How Undervalued Are SBs?

[zoobird]

For the past few years (basically, ever since I got really serious about fantasy baseball), I've been playing primarily in points/salary cap leagues. This year I'm going to focus on more traditional formats, so for the first time, I tried to calculate player values for hitters. I'm using CHONE projections, and did values for both 'raw' projections, and for projections that normalized all players to 600 plate appearances.

In both cases, it appeared that stolen bases are undervalued. I don't mean a little undervalued, but shockingly, unbelievably undervalued. My system was basically this:

-I'm going to play in one of the large, 'national' contests, so I took the stat thresholds that I would need to reach the 85th percentile in each category. If I can achieve that, I'll be in contention for the overall prize.

-For counting stats, I simply divided a constant by the target amount. So if the 85th percentile was 1200 RBI and 300 HR, then I valued each HR as 12/300 and each RBI as 12/1200.

-For batting average, I estimated how many AB would be normal, and calculated the target number of hits (based on the 85th percentile batting average). I think used the same approach for hits as other stats (ie if he target is 2200 hits, then hits are worth 12/2200). I calculated target # of outs, and then used that as a negative...ie if target is 5000 outs, then outs are each worth -12/5000.

As I said, SBs appeared to be horribly undervalued. Guy like Michael Bourn and Eric Young came out as potential top ten players. Ramirez, Reyes, and Ellsbury were FAR more valuable than Pujols.

What am I doing wrong? I realize that this system doesn't take into account the 'spread' within categories...is that the entire problem? And is it a more serious problem than systems like SGP, which ONLY look at the spread...not the overall totals needed to be competitive?

[Todd Zola]

Without actually seeing your spreadsheet, I can't comment, but I honestly feel 100% confident in saying there is a flaw in your system.

Plus I don't really like the manner BA is done, but I don't think that is the source of the error.

While I don't have as much free time as I have had the past few seasons, if you want to send me your spreadsheet, I can take a look.

Put my first name before the @ and the name of the site afterwards, and you have my e-mail addy.

[rotodog]

Actually, dont you mean Sbs are overvalued? If EY jr is a top ten player in your system, Sbs must be overvalued instead of undervalued.

If I had to take a quick guess using HR and Sbs as an example, all teams should have more HR than Sbs and your calculation of dividing 12 by the category target to find a value is a bit flawed. The sbs multiplier you come out with will be much higher than say HR. A guy that will swipe 30 bases will be disproportional to a 30 HR guy by a wide margin....if you dont negatively value any sup par production in other categories.

That being said, in a 12 mixed team league, there will be guys that can swipe 20 bags that wont even be drafted. You probably wont see many 20 HR hitters sitting on the FA wire though.

And the big flaw in the system I think is that you are giving every player a positive value for his contribution in any stat category...This is wrong as you need to also penalize a below avg contribution also...Not just reward every single sb, rbi, run and HR...

I am not going to go thru replacement value here or discuss SGP versus PVM.. But you need to somehow penalize a guy with a stat contribution that is subpar somehow..

Think of this. a guy that hits .260 with 0 HR , 20 rbi and 45 runs but swipes 35 bags... You gave value to the Sbs, but you need to discount his RBIs and Runs and his AVG negatively somehow because they are terrible contributions in the grand scheme of any league.... Negative value in those 4 categories added to any positive value the sbs will contribute will most likely have a negative value even though he swipes 35 bags....but those 35 bags come at a price... Terrible subpar production in the others cats...
Your system gives every single unit a positive value....

Thats what I see...there is more to it than that and your ratio categories will need adjustment, but only after you fix the bigger flaw I think...

[Todd Zola]

Actually, this type of system does not require a replacement consideration, there is something else going on.

[zoobird]

I won't have access to my spreadsheet until tonight, but I'll definitely be sending it to Todd to take a look.

I'm actually thinking there may NOT be a major problem. If I'm remembering right, the 85th percentile in this format was something like 300 HRs and 190 SBs. If that's the case, it would make sense that each SB is worth substantially more than each home run. There's also a wider gap in the projected number of steals between the top base stealers vs. the top HR hitters than there is in the projected number of homers. So the sluggers need to be a lot better in the other three categories to make up for their shortfall in combined value from SB+HR. For many of them, the batting average and runs are basically a wash, and the gap in RBIs isn't worth enough to make up the difference.

That said, the values for some of these guys were so shocking that I definitely won't trust it until Todd takes a look.

[viper]

but consider the average player gets 20 HRs and 10SBs [give or take one]. I think you will find the incremental increase over average to move up category points is almost the same, it is just the starting point is offset by 10. This is basically the SGP concept. I used to use it but changed a few years back after reading all of Todd's replacement player theory and realizing it was better - although the way I used SGP, the numbers were comparable. The problem was it was very hard to run all the waht if scenarios I like to runs plus category weighting was real hard.

To get a 12 in the NFBC in HRs the past four years it took
2009 :271
2008: 269
2007: 263
2006: 292

For SBs it took:
2009: 169
2008: 156
2007: 170
2006: 157

Average HR scores [8] were:
2009: 253
2008: 246
2007: 244
2006: 268

Average SB scores were:
2009: 145
2008: 135
2007: 144
2006: 136

First Caveat: I got these number by looking at the year-end overall results for all teams. The average score is the dead middle result in each category. The "12" results was moving an appropriate number of slots towards the top.

Second Caveat: I personally feel last years results are nice pieces of histroy and nothing else. You are not drafting against last years results but against your personal lists. Add all the HRs from you personal top 210 and divide by 15. That is your average. I would bump that by some historical percent and that would be my target.

[Milnertime]

Hi all! I'm new here and I'm really enjoying the stuff I've read so far.

I was planning on posting a thread similar to this just to see what some of the more experienced guys thought about my system, but I guess this is as good a place as any.

Rotodog points out that in the OP's system, he's not penalizing sub-par performance. What I do is create a pool of what I consider "draftable" players based on some kind of projection system (I've used CHONE, PECOTA, and MB) in excel. Then, I take the average of each statistic for all of the draftable players followed by the standard deviation of that stat. I then enter the formula (Player A's HR-Average HR)/Standard Deviation of HR.

So, if Player A's HR total was less than the average of the draftable players, he'd get a negative value for his HR total. Then, I add up all of the values for the stats the particular league uses to find the overall value of that player compared to the rest of the draftable players.....or at least I think that's what it's doing.

This formula is really what I had questions about, although the results seem to work pretty well in terms of placing players in what seems to be pretty reasonable order. Stolen bases tend to be overvalued by doing it this way, since league average is so low compared to where the best base stealers are. It's pretty easily fixed with some adjustment to the final addition of the individual values.

I find that this also allows me to pretty easily put players into tiers because it gives me a number that can be ranked with other formulas and such.

I'm just wondering if this is the right kind of approach or if I'm totally off base...and hopefully it helps the original poster with his question, as well.

[Todd Zola]

With respect to the original question, we worked on a couple of things and some manner of replacement is necessary, he has some ideas how he wants to do it.

With respect to the SD method, a few years ago I may have gone on a diatribe about how much I don't like the system, but over the years I have really softened my philosophy and now believe anything that gives you a relative rank is more than sufficient, so long as you realize it is a relative rank and therefore do not use these static values as be all end all to drafting or an auction and instead focus on other elements of game theory.

Don't get me wrong, I still believe the system we have developed over the years is the finest out there and does the best job at providing a foundation of values. But I no longer believe the "better" values we provide are a huge edge when it comes to drafting. There are other facets of game theory that trump whether a guy is $23 or $26. It took me almost 10 years of nose-to-my-spreadsheet to learn this. I don't expect others to believe me just because I am typing it on a message forum under the guise of being a so-called expert. It is something you literally have to experience.

With all that said, any system that accounts for categorical replacement level by positions will suffice in producing a relative rank.

[viper]

I was in the process of responding when I saw the Todd had already responded.

I used a similar system for years but with an SGP twist. I took averages of draftable players and had players with negative and positive value. Each category summed to zero for draftable players. It was precise. Still I needed to make some type of adjustment for catchers but I figured out a way to do it. And my lists were almost the same as Todd's CVRC methods.

The downside was time it took to create a set of rankings based on projections. Each time I changed the league setup [15 teams to 10 teams, AL to NL to Mixed], I had to redo the statistics. I was spending more time in calculations than I was in analysis of the results. No more. CVRC gives you all the flexibility you need. If you feel a category, like SBs, to too influential, you can reduce its weight. Now I spend more time examining the resultant tiers looking for places of strength and weakness.

[Milnertime]

Thanks for the responses. I've only prepared for fantasy baseball season in a more advanced manner for a couple of years, now. I am liking the CVRC projections, but the problem I'm having is that one of my leauges, in particular, uses weird categories that the CVRC doesn't account for. Stuff like H, XBH, and both AVG and OBP. For the record, it was not my idea to use all those stats.

As for your comment about relative rank, Todd, I don't really hold the values that come out of this as anything other than useful for creating tiers within each position and categorical value over replacement.

I appreciate the replies. I've learned a lot from just reading these forums and I think it's going to help me out a ton this season.

[Todd Zola]

but the problem I'm having is that one of my leauges, in particular, uses weird categories that the CVRC doesn't account for. Stuff like H, XBH, and both AVG and OBP. For the record, it was not my idea to use all those stats.

Maybe this is my fault for not spreading the word, but I routinely do customized valuation for subscribers every season. People usually PM me the specs, donate some money to my favorite charity and I give them customized values. Actually, teh donation part is optional

Jokes aside, I am more than willing to provide customized values of that nature.

[kjduke]

I won't have access to my spreadsheet until tonight, but I'll definitely be sending it to Todd to take a look.

I'm actually thinking there may NOT be a major problem. If I'm remembering right, the 85th percentile in this format was something like 300 HRs and 190 SBs. If that's the case, it would make sense that each SB is worth substantially more than each home run. There's also a wider gap in the projected number of steals between the top base stealers vs. the top HR hitters than there is in the projected number of homers. So the sluggers need to be a lot better in the other three categories to make up for their shortfall in combined value from SB+HR. For many of them, the batting average and runs are basically a wash, and the gap in RBIs isn't worth enough to make up the difference.

That said, the values for some of these guys were so shocking that I definitely won't trust it until Todd takes a look.

Your problem is twofold. First, the problem is that marginal value does not equal x/target needed, so SBs are not worth proportionately more (not even close) than HRs even though they are more scarce (at least in the NFBC which my model is built on). Secondly, the expected baseline stat of a replacement player for your particular league also is important, not an afterthough - less so if just for relative rank in a draft league, more so for auction value. The bulk of the error, however, is in the marginal value concept. I recall certain experts always overvaluing SBs early on in the NFBC, thought it was pretty funny that they swore by and drafted by a model which anyone could see was incredibly flawed had they just stepped back from their religion for a moment and done real world scenario analysis.