I noticed that there's been some discussion about posters' contract proposals and their feasibility (or lack thereof). This happens every offseason, so I decided to create a quick heuristic that might help people determine if a proposed contract value is reasonable or not.
1) I took Fangraphs crowd sourcing data from the last three offseasons, comparing (for simplicity) projected versus actual total contract values. Obviously some nuance is lost by eliminating years and AAV from the equation, but I wanted to keep this very, very simple.
2) I took the % delta value for each contract, calculated as follows: (actual - projected) / actual
3) I plotted the comparison between the % delta and actual value for each offseason to ensure that the general relationship between actual contract value and crowdsourced error (% delta) for each of the offseasons is basically the same (it is). I then pooled all that data into one big chart.
4) This chart was noisy as hell (especially for the lower dollar contracts) so I binned the data into five groups:
Contracts <= 10M
Contracts > 10M and <= 20M
Contracts > 20M and <= 50M
Contracts > 50M and <= 100M
Contracts > 100M
The trend line for these bins was nice and smooth, and logarithmic, flattening out closer to zero. Better predictions for bigger contracts, in other words.
Is this scientific? Hell no. But think of it as a guideline for the following question: Is the contract that I'm proposing reasonable?. Start with the Fangraphs crowdsourcing, found here, and plug the crowdsourced total value into the following formula:
expected error = ln(crowdsourced value in millions) * 0.185 - 0.64
That will give you an expected error. To get the expected value, do the following:
adjusted value = crowdsourced value + (crowdsourced value * expected error)
As an example, use David Price:
Fangraphs says $190.8 million
Quick n'dirty says $254 million
I would expect somewhere between the two, but maybe his contract is even bigger? Who the hell knows. But what I *do* know is that it's probably not a safe guess to project a big money contract below the crowd projections, unless you have good reason to do so.
Heres the fast plot of my work. I've plotted the average error value (dark blue) with lines corresponding to +/- one standard deviation of expected error in each bin (light blue). As you can see, there's considerable variance, but the trend is solid.
X-axis is *actual* contract value in millions of dollars
Y-axis is expected error
p.s. I'm sure I've fucked this up six ways til' Sunday so if you'd like to improve it, that would be great.
1) I took Fangraphs crowd sourcing data from the last three offseasons, comparing (for simplicity) projected versus actual total contract values. Obviously some nuance is lost by eliminating years and AAV from the equation, but I wanted to keep this very, very simple.
2) I took the % delta value for each contract, calculated as follows: (actual - projected) / actual
3) I plotted the comparison between the % delta and actual value for each offseason to ensure that the general relationship between actual contract value and crowdsourced error (% delta) for each of the offseasons is basically the same (it is). I then pooled all that data into one big chart.
4) This chart was noisy as hell (especially for the lower dollar contracts) so I binned the data into five groups:
Contracts <= 10M
Contracts > 10M and <= 20M
Contracts > 20M and <= 50M
Contracts > 50M and <= 100M
Contracts > 100M
The trend line for these bins was nice and smooth, and logarithmic, flattening out closer to zero. Better predictions for bigger contracts, in other words.
Is this scientific? Hell no. But think of it as a guideline for the following question: Is the contract that I'm proposing reasonable?. Start with the Fangraphs crowdsourcing, found here, and plug the crowdsourced total value into the following formula:
expected error = ln(crowdsourced value in millions) * 0.185 - 0.64
That will give you an expected error. To get the expected value, do the following:
adjusted value = crowdsourced value + (crowdsourced value * expected error)
As an example, use David Price:
Fangraphs says $190.8 million
Quick n'dirty says $254 million
I would expect somewhere between the two, but maybe his contract is even bigger? Who the hell knows. But what I *do* know is that it's probably not a safe guess to project a big money contract below the crowd projections, unless you have good reason to do so.
Heres the fast plot of my work. I've plotted the average error value (dark blue) with lines corresponding to +/- one standard deviation of expected error in each bin (light blue). As you can see, there's considerable variance, but the trend is solid.
X-axis is *actual* contract value in millions of dollars
Y-axis is expected error
p.s. I'm sure I've fucked this up six ways til' Sunday so if you'd like to improve it, that would be great.
Last edited:
