Fantasy Value Above Replacement Spreadsheet

[Rufus T Firefly]

2011 Fantasy Baseball Values Google Document Link

After reading through Zach Sanders' three-part series on FanGraphs about Fantasy Value Above Replacement (Link), I was inspired to create an excel sheet with player values for the 2011 season. I would recommend reading through the linked article before looking at the spreadsheet as it can be a bit overwhelming at first glance.

The spreadsheet contains values for each 5x5 stat by deriving each player's z-score (standard deviations from the mean) for their position. The totals were added for each of the stats and sorted in descending order. This displays what each player is worth in comparison to the rest of the field at his position. These are available for each position by viewing the position tab on the bottom left.

This is where Sanders finished his article, but I wasn't satisfied with simply having all of the players' values in relation to their position. We all know that Joe Mauer is not a top-ten player, due to him just having a high mark comparatively for his position (he was after my first calculation and compiling of the player pool). To remedy this situation, I formulated z-scores for each player in comparison to the average for the entire player universe. This is to ensure that the total production of a player is not to be overlooked and that players from weaker positions do not have padded rankings. With the two values now computed (against position and against complete player pool), I simply averaged the two scores. This helped to get players to where I felt there real value lays. This total can be viewed on the “ALL” tab. It might be possible to further tweak the scores to more weight scarcity or total production, but I left it 50/50 and liked what I saw with the results.

To compute the dollar value for each player, I used Sanders’ formula available on the third part of the article. The computations listed on the spreadsheet are for my 20-team, 25 man roster league set to a $260 budget.

Changes can easily be changed by adjusting the formula to your league specifications: [(Team Budget - (1*no. of players per team)) / number of players per team] * (z-score above replacement / average z-score for above-replacement players) +1 = Dollar Value

While Sanders suggested multiplying the pitchers’ value by .8, I took it a step further and multiplied it by .7 - this helped to get the pitchers’ values closer to where I believe they should be, as well as more properly reflecting my fears and general avoidance of pitchers. I would be interested to hear anyone’s thoughts on this matter.

Under the “My Team” tab I tried to get a little experimental. I have my current team, which is three-quarters of the way through a slow draft, listed with their values. I was interested to see how my team broke down in comparison to the average dollar amount per player and per stat. As you can see, I have listed the value of each stat per player on the page with the overall value and dollar value. The totals below show the aggregate standard deviations I am above/below the average player.

This is somewhat flawed though as, even with this being a deep league, not every player will be selected. With 500 rostered players, I made the decision that players ranked 400 and below would be considered “replacement level or below”. I then averaged the top 400 players with each of the stats that are listed under: “Average Above Replacement Value Per Stat” and determined the value per stat for the average non-replaceable player (top 400 players). These values can be seen besides the LEAGUE designation. My players’ values per stat are rated underneath. My next order of business was to find a dollar value for each category, which I simply subtracted my value from the league average and multiplied the result by the dollar value multiplier (discussed above by Sanders’ formula – my league = 6.23). As you can see for an example, I have a $3.2 advantage over the average player in home runs, but trail $0.14 in stolen bases.

I’m interested to hear the board’s reception of this. While it may look complex, it didn’t take very long to complete and is easily adjustable. There is certain value in this, while it should not be used as a total reliance for drafts, it can be used to make more informed decisions. I plan on using this as somewhat of a “trade chart” to value incoming and outgoing players since my draft is near conclusion and most of the current available players rank as below replacement value.

In the future I plan to mess around with creating keeper league values. It would be interesting to see how you can rate players as they decline in age. With the standard rule of thumb being .5 WAR per year after 30, it would be fun to take .5 off the overall ranking each season to help value players into the future. Another fun idea I have had is to help value prospects using their potential multiplied by their bust rates (courtesy of Royals Review). Lowering players due to risk would also be interesting to see.

Notes:
- This system is only as valuable as its rankings. For this experiment I used KFFL'srankings because they had free excel downloads, which saved hours of data entry. They don't look horrible off of first glance, but they could be tweaked to personal preferences.

- To edit the projections, make the changes on the position tab. For example, to adjust Hanley Ramirez's stolen base total, go to the "SS" tab and edit from there. The rest of the pages will automatically update.

- The metric stats are multiplied by at bats or innings pitched to weight the higher total.

- SP and RP values were calculated together as a pool to more accurately show their worth.

- If anyone would like the excel document I can send it to them.

[joncarlos]

Looks very good. I like the way you handled position scarcity which isn't something I've seen before.

I've done projections using this method (with no adjustment for position) and there were 3 issues, one of which is derived from the other two.

-SBs get valued too highly because the standard deviation is so low. There are many, many players with 0 or 1 SB which means that the SD ends up being something like 6 or 7. That puts Michael Bourn something like 8 standard deviations above the mean. (Checking your sheet, you've got an average around 9 and an SD around 10) This works out OK for one-category guys like Bourn who end up with $50 in SB value and enough negative HR/RBI value to balance it out, but it puts a guy like Kemp through the roof. (Which maybe they should be, I guess)

-HRs end up getting a little overvalued too, just because they also count for runs and RBI. Again, for a guy like Dunn, it works out OK because he gets killed in AVG and SB.

-So the derived issue is that you have all of these things that usually cancel each other out, but then you get a guy like David Dejesus and he usually ends up a little too low because he doesn't have extreme highs and lows.

In the end it comes fairly close to working itself out right but the process of getting there doesn't always work. Your numbers look pretty reasonable though.

[Gregg]

Just reading your description is time consuming. I am afraid to open it.... I will have to wait until the family is in bed tonight. I am really looking forward to it.

[amcg]

*looksupfromreadingthenewspaperatthebar*

Replacement valuation eh? You'll be wanting to talk to a fella name o' Zola. He wanders in now and then to have a look around... don't buy much. Usually spends most of his time in another bar down the road... I'll let him know you called.

*backtonewspaper*

[Gregg]

I must be reading this wrong. It is showing James Loney at #11 in value. I am not sure Loney is 11th on his own team much less in the pool of 1B.

[joncarlos]

I must be reading this wrong. It is showing James Loney at #11 in value. I am not sure Loney is 11th on his own team much less in the pool of 1B.

.280-11-89-70r-9sb puts him one spot ahead of Konerko at 1B, .274-30-95-87r-1sb. You are reading that right.

His rank among 1Bs is skewed because he's 2 standard deviations above the mean for steals (zSB = 2.107). That's why doing position scarcity/values by positions can be tricky. He's properly valued as 95 overall ... maybe still too high, but below Konerko who is 70th overall ... when he is listed with everyone and he's not an outlier in steals any more.

[onejayhawk]

I saw Royals Review and thought of KS.

I have been playing around with it. There are some interesting things. For example, some multi position players cahnge significantly when put in their more valuable position. I have Cuddyer, so I put him in the 3B pool, and his value is higher, enough that he would climb a full round overall (if I am reading this right). Getting the composit sheet to work right, after the change, is non trivial.

Adding the stat line also does interesting things to the variances within the position. Taking Cuddyer out of OF causes barely a ripple, but his stats make a noticable shift in the 3B numbers. This is something to think on. For fantasy purposes, most players are properly valued,, not at their true position, but at their most difficult position. For 1B/OF types, I doubt it matters, but for the Jorge Posadas and mike Napolis, it is huge. Fortunately, both are rated as Cs already.

J

[Rufus T Firefly]

I must be reading this wrong. It is showing James Loney at #11 in value. I am not sure Loney is 11th on his own team much less in the pool of 1B.

I noticed that this morning and like joncarlos said it's because of the SB standard deviations. I might have to mess around with it, because something like that definitely is not right.

I have thought about devaluing stolen bases for the 1B and C position possibly by cutting in half each player's total. With the averages at these positions so low even a modest amount, like Loney's, can skew the rankings. I'll mess around with it and see if I can create a formula that has some logic to it and also devalues stolen bases. The way it current is working, Albert Pujols most valuable category is the stolen base. That certainly isn't right.

[axman]

I have played around with this a few times and find by the time I'm done tinkering around I drop the standard deviation stuff and end up weighing players on their category outputs (vs. a replacement level player). SB is an issue but can be solved with a formula change. Weighing what a player does in categories against players of the same position is never really going to work. Your wBA calculation doesn't value BA properly. A .240 hitter with 600ab's is going to hurt your team because the .240 is far less than what a replacement player or average player would offer. The wBA category needs to have the average weighed against something.

I have tinkered with this stuff for years and have come to the conclusion that sometimes simpler might just be better. The one thing I do like with these exercises is that while the values may not line up with "traditional" values they can sometimes show you were you might find value in your auction/draft.

[DMT]

I have tinkered with this stuff for years and have some to the conclusion that sometimes simpler my just be better.

That's the same conclusion I came to about 6 years ago and haven't bothered with z-scores, standard deviations, or running my own projections since.

[onejayhawk]

Z-score are very handy for something that has a symmetrical distrution, such as ERA or BA. Counting stats are different. They tend to mass at the left end, ie at 0. For large number stats, like Runs and RBI, the z-score still works pretty well, since the players with zero, tend to be the players who dont play. Its HR and SB where the problem lies.

Just as an experiment, try gifting everyone a (1) SB, and see what it does to the results.

J

[Erik]

That's the same conclusion I came to about 6 years ago and haven't bothered with z-scores, standard deviations, or running my own projections since.

I never did -- mainly because I can compete in my main league without any of that. Also because I'm not that good at advanced math.

[NL Only]

Hi Rufus - did you ever do anything more with this? And continue using it after 2011?

[chancellor]

That's the same conclusion I came to about 6 years ago and haven't bothered with z-scores, standard deviations, or running my own projections since.

Been interesting re-reading this thread. I do that work more for the process - gives me a feel for players who might have a skill I wasn't aware of or who might carry more risk or more risk/reward. I don't bring that work into my auction, but the process has helped me think through who I want to price enforce, go after, avoid, avoid at all cost and so on.

I'm not vain enough to think my projections are better than the folks who do this for income, like Todd Zola, Ron Shandler, or Paul Sporer, but the process does help for quick reactions during an auction or deep draft.