With the recent discussion about RFAs and the number of teams that seem to have a need to move players with salary, I wanted to see if it was possible to create a way in which one could estimate the cost to trade a player with negative value or an unwanted contract.
To begin, I decided that all the components involved in a trade needed to be quantified in the same stat. This stat ended up being the WAR created by Evolving-Wild; while WAR isn’t perfect I do believe it offers a relative accurate estimation of a player’s overall value. To find the average WAR of a player that was being traded I used a basic weighted average of the player’s last 3 seasons. Their most recent season was worth 5 points, the 2nd most recent was worth 4 points, and the 3rd most recent season was worth 3 points. The total was then divided by 12 and the resulting value was their weight average across those 3 seasons (5*S1+4*S2+3*S3)/12.
The next step in the process was to try to create an expected WAR for each draft pick. To speed up the process I estimated the expected WAR for each pick by finding the WAR generated from 1st overall picks over their first 8 seasons in the NHL, some players had less than 8 seasons due to evolving-hockey only tracking WAR from the 07-08 season and onward. I averaged those values out to create an expected WAR per season for 1st overall picks. The average WAR that a 1st overall pick produced over the reviewed seasons was 2.075 per season.
The 2.075 value was then applied to the 2016 version of the draft pick valuation chart created by Michael Schuckers to give an expected WAR for each draft pick. While the chart is used to show the value of each pick I believe this valuation table also applies here. This is because I am just translating the values he found into WAR. Instead of pick 10 being worth 422 points it is worth (422/1000)*2.075 = 0.875 WAR, in both cases the 10th overall pick is worth 42.2% of the 1st overall pick. The main difference is that this translation allows us to compare picks to players and establish values in trades.
After completing an expected WAR for each pick, I was left with an individual value for all picks which I then averaged into the following groupings:
Table 1
With a value for picks and players the last step was to identify the trades that would be selected. The trades included were manually selected, so this review was very subjective in nature. To qualify there are 2 main criteria that had to be met:
Table 2
The trades can then further be broken down based on how much term was remaining on the player’s contract when the team acquired them:
Table 3
While I typically wouldn’t break a sample size of 12 into smaller groups there was enough of a difference in the "Average WAR Gained per 1% of Cap Increase" that I think it makes sense to group them separately.
The important thing to take away from the table above is the different rates of return per 1% increase cap increase, as well as the standard deviation. For example, a team that is taking on a negative value player with 1 year remaining on their contract could expect to receive about 0.055 WAR for every 1% their salary cap increases. The standard deviation allows us to evaluate if a trade is fair or not, through some testing I believe anything within 1 standard deviation is about fair value.
Lastly, the table does not include anything in the 4, 5, or 6-year range. This is because no such trade has ever been made. By making some assumptions about what the shape of the data for those periods I believe I’ve been able to create an estimation that would be accurate for contracts that are 4, 5, and 6 years in length.
Table 4
The values for 3, 4, 5, and 6 contracts are really just for fun as there are either no real-world examples or the data is limited. I thought it would be interesting to include them as people seem to want to throw those types of players into trades with some regularity.
To show how to properly use some of the values and rates I’ve created here is an example using a player that may be moved this offseason: Ryan Callahan.
Table 5
What’s important to note here is the Total Player Value and the standard deviation. The total player value is Callahan’s current worth including his contract and his recent performance put into the terms of WAR. The Total Player Value essentially tells us that in order to convince a team to take on his contract TB would need to include assets that are on average worth .404 WAR . The standard deviation is calculated by taking the 0.21 value from Table 3 and multiplying it by the "Total Player Value". The standard deviation provides us with both the low and high end of what we could consider to be a fair trade; so long as the value a team receives are between +/- 0.149 WAR they will have received fair compensation for taking on Callahan's contract.
Table 6
Using the estimated pick values from earlier these are 3 possible trade scenarios that I came up with; there is a low cost, medium cost, and a high cost. I don't believe any of those trades are completely outlandish and depending on the situation and what the trade market is like all 3 are reasonable outcomes.
Overall, I think the method discussed can provide a decent estimation of the type of assets a team would need to move to “dump” a player. Even when accounting for the small sample sizes, I think that by using the values from “AVG WAR Gained Per 1% of Cap Increase” in conjunction with the standard deviation there are very realistic trades created. By keeping the trade within 1 standard deviation you can effectively see an accurate estimation of the cost depending on the trade market.
I have never done this sort of analysis before so if anyone has any suggestions to improve my method or if they noticed some sort of critical error please let me know. Also, if you have any questions about anything that I talked about or the process in which I came up with the values let me know. I'm happy to discuss.
And of course a big thanks to CapFriendly as that is where I gathered all the salary figures; another huge shootout to Evolving-Wild and their WAR model, you should check their site out if you haven't already; and to Michael Schuckers for creating the draft valuation chart that I based my expected WAR of draft picks on. You can read into his most recent research on the topic here.
To begin, I decided that all the components involved in a trade needed to be quantified in the same stat. This stat ended up being the WAR created by Evolving-Wild; while WAR isn’t perfect I do believe it offers a relative accurate estimation of a player’s overall value. To find the average WAR of a player that was being traded I used a basic weighted average of the player’s last 3 seasons. Their most recent season was worth 5 points, the 2nd most recent was worth 4 points, and the 3rd most recent season was worth 3 points. The total was then divided by 12 and the resulting value was their weight average across those 3 seasons (5*S1+4*S2+3*S3)/12.
The next step in the process was to try to create an expected WAR for each draft pick. To speed up the process I estimated the expected WAR for each pick by finding the WAR generated from 1st overall picks over their first 8 seasons in the NHL, some players had less than 8 seasons due to evolving-hockey only tracking WAR from the 07-08 season and onward. I averaged those values out to create an expected WAR per season for 1st overall picks. The average WAR that a 1st overall pick produced over the reviewed seasons was 2.075 per season.
The 2.075 value was then applied to the 2016 version of the draft pick valuation chart created by Michael Schuckers to give an expected WAR for each draft pick. While the chart is used to show the value of each pick I believe this valuation table also applies here. This is because I am just translating the values he found into WAR. Instead of pick 10 being worth 422 points it is worth (422/1000)*2.075 = 0.875 WAR, in both cases the 10th overall pick is worth 42.2% of the 1st overall pick. The main difference is that this translation allows us to compare picks to players and establish values in trades.
After completing an expected WAR for each pick, I was left with an individual value for all picks which I then averaged into the following groupings:
Table 1
With a value for picks and players the last step was to identify the trades that would be selected. The trades included were manually selected, so this review was very subjective in nature. To qualify there are 2 main criteria that had to be met:
- There must have been at least a $2,000,000 difference in cap space received versus what was traded.
- There must have been a clear “Dump” of a player. For Example, The Boychuk trade was not considered as it is not what I would consider to be a salary dump. The Bruins did not need to incentivize the Islanders to trade for him, Boychuk had enough value to return assets on his own.
Table 2
The trades can then further be broken down based on how much term was remaining on the player’s contract when the team acquired them:
Table 3
While I typically wouldn’t break a sample size of 12 into smaller groups there was enough of a difference in the "Average WAR Gained per 1% of Cap Increase" that I think it makes sense to group them separately.
The important thing to take away from the table above is the different rates of return per 1% increase cap increase, as well as the standard deviation. For example, a team that is taking on a negative value player with 1 year remaining on their contract could expect to receive about 0.055 WAR for every 1% their salary cap increases. The standard deviation allows us to evaluate if a trade is fair or not, through some testing I believe anything within 1 standard deviation is about fair value.
Lastly, the table does not include anything in the 4, 5, or 6-year range. This is because no such trade has ever been made. By making some assumptions about what the shape of the data for those periods I believe I’ve been able to create an estimation that would be accurate for contracts that are 4, 5, and 6 years in length.
Table 4
The values for 3, 4, 5, and 6 contracts are really just for fun as there are either no real-world examples or the data is limited. I thought it would be interesting to include them as people seem to want to throw those types of players into trades with some regularity.
To show how to properly use some of the values and rates I’ve created here is an example using a player that may be moved this offseason: Ryan Callahan.
Table 5
What’s important to note here is the Total Player Value and the standard deviation. The total player value is Callahan’s current worth including his contract and his recent performance put into the terms of WAR. The Total Player Value essentially tells us that in order to convince a team to take on his contract TB would need to include assets that are on average worth .404 WAR . The standard deviation is calculated by taking the 0.21 value from Table 3 and multiplying it by the "Total Player Value". The standard deviation provides us with both the low and high end of what we could consider to be a fair trade; so long as the value a team receives are between +/- 0.149 WAR they will have received fair compensation for taking on Callahan's contract.
Table 6
Using the estimated pick values from earlier these are 3 possible trade scenarios that I came up with; there is a low cost, medium cost, and a high cost. I don't believe any of those trades are completely outlandish and depending on the situation and what the trade market is like all 3 are reasonable outcomes.
Overall, I think the method discussed can provide a decent estimation of the type of assets a team would need to move to “dump” a player. Even when accounting for the small sample sizes, I think that by using the values from “AVG WAR Gained Per 1% of Cap Increase” in conjunction with the standard deviation there are very realistic trades created. By keeping the trade within 1 standard deviation you can effectively see an accurate estimation of the cost depending on the trade market.
I have never done this sort of analysis before so if anyone has any suggestions to improve my method or if they noticed some sort of critical error please let me know. Also, if you have any questions about anything that I talked about or the process in which I came up with the values let me know. I'm happy to discuss.
And of course a big thanks to CapFriendly as that is where I gathered all the salary figures; another huge shootout to Evolving-Wild and their WAR model, you should check their site out if you haven't already; and to Michael Schuckers for creating the draft valuation chart that I based my expected WAR of draft picks on. You can read into his most recent research on the topic here.
