Prospect Surplus Value

by colintj on Jul 13, 2010 3:05 PM EDT

[Sky's note: Colin followed up on his prospect value question, creating these sweet tables. That's more than front page-worthy. I respect following-up a lot.]

After taking into account Matt's research (scroll down) to the best of my ability (sorta), I wanted to post a new table for prospect surplus value. Everything should be more or less self explanatory. $MM/yr is the dollar value of that prospect for a given year. Tot$MMsurplus is the total value for all 6 cost controlled seasons minus expected arbitration dollars paid. Hitters:

	WAB	fWAR	rWAR	WARP	$MM/yr	tot$MMsurplus
H1-10	1.40	1.75	1.51	2.19	10.94	47.44
H11-25	1.00	1.25	1.08	1.56	7.81	34.31
H26-50	0.80	1.00	0.86	1.25	6.25	27.75
H51-75	0.60	0.75	0.65	0.94	4.69	21.19
H76-100	0.50	0.63	0.54	0.78	3.91	17.91

And pitchers:

	WAB	fWAR	rWAR	WARP	$MM/yr	tot$MMsurplus
P1-10	0.60	0.75	0.65	0.94	4.69	21.19
P11-25	0.55	0.69	0.59	0.86	4.30	19.55
P26-50	0.50	0.63	0.54	0.78	3.91	17.91
P51-75	0.45	0.56	0.48	0.70	3.52	16.27
P76-100	0.40	0.50	0.43	0.63	3.13	14.63

When you use different replacement levels in a wins over replacement framework, you create a certain number of wins that can be bought in a given year. The higher the replacement level, the fewer the wins above replacement available. So whichever you're using (Wins Above Bench, fangraphs WAR, rally WAR and WARP) needs to be consistent or at least translated properly. The total wins for each look like:

WARP: 1250

fWAR: 1000

rWAR: 865

WAB: 810

After myriad adjustments, Matt found that WARP went at about $5MM per over the last few seasons ('08 to present). This figure will change over time, thanks to inflation, demand and/or CBA changes. Since the total sum paid for wins is constant, but the number of wins are different across systems, the dollars per win figure varies. The more wins, the fewer the dollars per. Here's dollars per win for each system, using $5MM per WARP as a baseline:

WARP: $5MM

fWAR: $6.25MM

rWAR: $7.23MM

WAB: $7.72MM

Tweet Comment 19 comments | 0 recs |

Do you like this story?

Comments

pardon the ugly tables

tableizer went awry on me.

Fireworks: Bang?

by colintj on Jul 13, 2010 3:07 PM EDT reply actions

This looks really good to me. Definitely makes sense. It should be interesting to consider discounting time and inflation to carry this forward, but these are really good tables to get the basics down. Good work.

by Matt Swartz on Jul 13, 2010 7:01 PM EDT reply actions

there is at least a discount built in

since I used VW’s discounted WAB. if that’s insufficient, let me know and i can edit. anyone who wants to c/p for themselves can check:

http://spreadsheets.google.com/pub?key=0AhBxILsMcvIMdFg5bnFhcGZjVHRqZ1JTdjREeks4QVE&authkey=CMqT3eoF&hl=en&output=html

Fireworks: Bang?

by colintj on Jul 13, 2010 8:25 PM EDT up reply actions

Oh, I didn’t remember Victor did that. I think that makes a lot of sense then. Very cool.

by Matt Swartz on Jul 14, 2010 9:46 AM EDT up reply actions

tangent: so draft pick compensation is already built into your MORP technique?

how did you do that?

Fireworks: Bang?

by colintj on Jul 14, 2010 9:05 PM EDT up reply actions

I went through all of the free agents who were Type A and figured out what draft pick that team gave up. If they were signed by their owned team, I counted that as an opportunity cost of the 25th pick and the 35th pick (as the supplemental pick) but ONLY if he would have been offered arbitration otherwise in my view (typically if he got a raise). For Type B signed by their own team, I figured out if they would have gotten a pick otherwise. A lot of those were judgment calls so I erred on the side of not assuming compensation.

Then I converted Sky Andrecheck’s rWAR Draft Pick Calculator, converted to WARP, multiplied by 2/3 to represent how much of that WARP is historically earned in the first six years, then by 0.7 to adjust for the fact that you do need to pay for arbitration.

Then I discounted them all by 8% over 6 years to get an amount to subtract from the WARP produced during the contract, and spread this out evenly across all years of the contract.

Then I got approximate bonuses for each draft round, ranging from $1.8MM for the 16th pick to $0.2MM for the 80th pick. Then I subtracted that from the AAV for the free agent, because the team saved that money.

So I had a net $ and a net WARP, and summed that across all players in each year and divided $ by WARP. For 2010, I approximated the number of WARP typically produced by players who had more than six years of service time and divided that (net of the draft pick WARP) into the net $ for the contract.

by Matt Swartz on Jul 14, 2010 9:31 PM EDT up reply actions

so the $/WARP number includes an average of those who get compensation

and those who did not? and this represents the fact that the free agent player pool from which a potential replacement would be acquired features some players whose acquisition includes the payment of compensation?

and that means i still need to include the value of the compensation picks in figuring, say, cliff lee’s total value?

also: are comp picks discounted properly? a pick in 2010 isn’t going to yield major league WAR at least until a year in—the point at which they can be traded for mlb ready talent—and more likely don’t “mature as assets”* until 2-3 years or so down the line. so the clock on a 2010 pick that makes it doesn’t start until 2012 or 2013 right? so you discount the two years or so in between plus the 6 years?

*i’m only pretty sure that’s the right jargon term

Fireworks: Bang?

by colintj on Jul 14, 2010 9:58 PM EDT up reply actions

Yeah, it includes players who did and did not get compensation. The six years came from the fact that the average draftee who reaches the major leagues does so 3 years after being drafted, and so typically would reach free agency 9-10 years later, and so their value would accrue about 6 years later on average. 8% was just to pick a number that made sense.

The Cliff Lee thing is much trickier because the $/WARP should be higher in season to reflect the relative value of of wins in season (check the Halladay and Fielder articles I did as examples). But for the draft pick compensation, that’s already put in to the $ value, but he also gets $13MM about in draft pick compensation if signed by his a team that surrenders its first round pick.

by Matt Swartz on Jul 14, 2010 10:11 PM EDT up reply actions

but you did do an average for each player's compensation

distributed across all players? how could a dollars per win figure itself include all the necessary compensation? if expected comp is above average, that has to be added in, no?

also, wrt mid-season $/WARP, there is somewhat more variance given the half sample size right? was that taken into account? and does the data support the varying cost of talent in or out of season in trades? i feel like a predictable rise in cost every summer would be taken into account in the preceding winter. it seems like paying for WPA rather than WPA/LI. you end up paying for circumstance.

…sorry for all the questions, but it’s a lot of material to digest.

Fireworks: Bang?

by colintj on Jul 15, 2010 12:45 PM EDT up reply actions

No problem with all the questions. I’ll do my best to answer.

It was literally a sum over a sum for each player’s compensation. So for each player I got the AAV of their contract that year if they had more than six years of service time (made some adjustments to only focus on the latter portion of the contract when free agent years were bought out while still under team control), and subtracted the “savings” on not spending on draft picks. Then I got the WARP and subtracted the draft pick compensation cost. For each player then, I had a net WARP and a net $. I summed net $ across all 350ish players who played or were paid major league salaries and had more than six years service time, and then summet net WARP for that group as well. Does that answer your question?

For the mid-season $/WARP, I just made a ballpark estimate of how much is being paid for reaching the playoffs and how much is being paid for the win itself, outside of context. I used some Nate Silver research to guide it, with some updated dollar figures. I also considered the impact on advancing in the playoffs and put some approximate numbers to it. The key insight came from looking at the effect on the probability of reaching a given win total when you add 5 wins in expected value to a 162 game season where the variance of final win totals is large, versus when you add 2.5 wins in expected value to a 81 game span where the variance in final win totals is small. (Note that the variance in expected win total increases with the number of games remaining, but the expected win percentage decreases with the number of games remaining).

Does that make sense?

by Matt Swartz on Jul 15, 2010 3:41 PM EDT up reply actions

ugh, there's something i'm still missing

i’m still chewing this over though. thanks for the replies.

i think i would need to see an example or something, since i’m still stuck on how you managed to get compensation accounted for with an average. unless you have individualized MORPs? blerg.

pitchfork weekend probably didn’t make my brain more capable of getting it.

Fireworks: Bang?

by colintj on Jul 23, 2010 12:29 AM EDT up reply actions

either way

i’m not sure it’s necessary to price in compensation like you have. though i do agree this reflects actual price paid better than the current set up, i think it makes it harder to use.

is there any way you could show me what Tango did and then what adjustments you did to get your figure? i definitely agree that the free agent bias screws with using projections, but past that i’m not sure if your conception is better.

the best would be a nice little demonstration with a decent sample size of ’spects-for-vets deadline trades.

Fireworks: Bang?

by colintj on Jul 23, 2010 12:33 AM EDT up reply actions

Outside of the free agent overprojection and the draft pick compensation, there is also the issue that using the first year of deals’ WAR will overstate the $/WAR because the WAR is lower later in deals.

I don’t have a whole spreadsheet for you, but consider this sample of what would be done.

Say there are 5 players in the league with more than six years of service time. Players A and B are on one-year deals at $4MM and $6MM, Player C is in the first year of a 3-year deal at $10MM and was a Type A FA who resigned with his old club, Player D is in the 2nd year of a 4-year deal and is making $12MM, and Player E is in the 4th year of a 4 year deal and is making $20MM and was a Type A FA who signed with a different team who surrendered a first round pick. Suppose Players A-E have WAR projections of 1.0, 2.0, 3.0, 4.0, and 5.0, but they actually produce 1.0, 1.5, 3.0, 2.0, and 3.5.

The traditional method would only look at the three guys in the first years of their deals, only at projected WAR and not look at draft pick comp. They will combine to produce 1.0 + 2.0+ 3.0 = 6.0 WAR, and will make $4m + $6m + $10m = $20 million. So that would say $20 million / 5 WAR = $4.0MM/WAR.

My method would call that Type A to be a $13MM tax on Player C, and say $8MM on Player E. So I would say that the real cost of Player A was $14.3MM/year since it was a 3-year deal. I would also count Player D and Player E who are in their decline phase and Player E will count as $22MM/year because of the $8MM cost spread cost 4 years plus his actual salary.

So, my way will say:

1.0 + 1.5 + 3.0 + 2.0 + 3.5 = 10.0 WAR, and $4m + $6m + $14.33m + $12m + $22m = $58.33 million. So I would get $5.83MM/WAR.

That’s a big difference.

by Matt Swartz on Jul 23, 2010 7:12 PM EDT up reply actions

One other question

Does Top 76-100 Hitter, for example, mean that this is a hitter ranked in the Top 76-100 players overall, or among the Top 76-100 Hitters ranked? In other words, how many players make up the groups “Top 1-10 pitchers” and “Top 1-10 hitters”, 10 or 20 people total?

by Matt Swartz on Jul 14, 2010 1:48 PM EDT reply actions

ah right

well i just used VW’s categories. i’m pretty sure that means “pitcher ranked in the top X” of all prospects. and given the findings, it definitely looks like we should be ranking hitters and pitchers separately.

Fireworks: Bang?

by colintj on Jul 14, 2010 2:34 PM EDT up reply actions