With the ISFL draft approaching and a huge number of mock drafts being posted, I had to weigh my options as to what approach I would take to generate a successful mock. Being a new player to the league, would it really be a smart strategy to try to understand the complex interplay between team needs, positional quality, player intangibles, etc.? At the same time, how could I really decide whose mock draft to copy without the experience to know who knew what they were talking about?
While a fellow CB prospect decided to create a model of these complex dynamics to generate their mock draft for them, I decided to take the opposite approach and simply copy everyone else at once! With the idea that the wisdom of crowds often outperforms any one individual forecaster in the long run, I decided to generate a consensus mock draft by aggregating each mock draft posted and ranking each player by the average position that they appear across all mock drafts.
To do this, I recorded every unique mock draft posted in the point task thread up until a cutoff date of 10am EST on draft day (so this article would have time to get read before mocks are due). I focused only on unique mock drafts because the vast majority of people posting duplicate mock drafts would be copying solely to get the TPE credit without putting much thought into it. This approach limits the consensus only to those who have demonstrably put at least some thought into creating a unique mock draft of their own, and hopefully improves the forecast accuracy. I then compared the draft positions of every player included in 2 or more mock drafts to come up with a consensus ranking across ISFL rankers.
The results are found at this link, and the consensus top 12 is:
Methods
One major consideration in any mock draft situation is the importance of positional needs. If positional needs are considered, then the average valuation of a player may not reflect that player's final draft position. For example, if the only teams willing to pick WRs in the first round are at pick 4 and pick 8, and the top 2 WR are considered equal by everyone posting mock drafts, then the average position for both would be around pick 6, but the team actually picking at pick 6 would never take them. This raises a concern as to whether this approach is valid for generating a mock draft, and is definitely the advantage of taking an approach like @Beefstu409's. I chose to ignore consideration of positional needs for this exercise because the ISFL has a unique feature that reduces the importance of this: positional switches. While not every player may be willing to change positions for the team that drafts them, the possibility of doing so can have a potentially huge impact because teams could be willing to draft a WR even when they don't need one if that WR then changes to a position they do actually need. In addition, the fact that we get credit for the pick even if a different team uses that pick via trade further increases the relative importance of accurate valuation rather than team fit. Those factors aside, I believe this approach could definitely be improved by finding a way to incorporate team fit.
Posted mock drafts being limited to only the 1st round also complicates the analysis, as I had no information as to whether posters thought that any specific player they didn't have in the first round would be drafted at, say, the 13th pick or the 50th. To solve this problem, I based the consensus rankings on each player's combined "win rate" against all other players in the draft. For each pair of players, I look across all mock drafts and keep score of which player is ranked higher in each mock draft. If one of the players is included in a mock draft but the other isn't, then the one that is in gets a win since it is clear that the ranker preferred the one that's in the mock draft. However, if neither player is included, then I assign neither a win since it's impossible to know which the ranker thinks will be drafter higher. I then calculate a combined win-loss record for each player by totaling their records in each individual matchup, and rank the players by their win%.
Beyond the Win% metric that I use to ultimately rank players for the consensus mock draft, I additionally calculated:
Highlights
While a fellow CB prospect decided to create a model of these complex dynamics to generate their mock draft for them, I decided to take the opposite approach and simply copy everyone else at once! With the idea that the wisdom of crowds often outperforms any one individual forecaster in the long run, I decided to generate a consensus mock draft by aggregating each mock draft posted and ranking each player by the average position that they appear across all mock drafts.
To do this, I recorded every unique mock draft posted in the point task thread up until a cutoff date of 10am EST on draft day (so this article would have time to get read before mocks are due). I focused only on unique mock drafts because the vast majority of people posting duplicate mock drafts would be copying solely to get the TPE credit without putting much thought into it. This approach limits the consensus only to those who have demonstrably put at least some thought into creating a unique mock draft of their own, and hopefully improves the forecast accuracy. I then compared the draft positions of every player included in 2 or more mock drafts to come up with a consensus ranking across ISFL rankers.
The results are found at this link, and the consensus top 12 is:
- 1.
Juan Domine
2.Zoe Watts
3.Alejandro Chainbreaker
4.Asher Montain
5.Joel Drake
6.Joshua Campbell
7.Dorothy Zbornak
8.Harrison Andrews
9.Quentin Button
10.Busch Light
11.Gunner Thorbjornsson
12.Mai Fukushu[/li]
Methods
One major consideration in any mock draft situation is the importance of positional needs. If positional needs are considered, then the average valuation of a player may not reflect that player's final draft position. For example, if the only teams willing to pick WRs in the first round are at pick 4 and pick 8, and the top 2 WR are considered equal by everyone posting mock drafts, then the average position for both would be around pick 6, but the team actually picking at pick 6 would never take them. This raises a concern as to whether this approach is valid for generating a mock draft, and is definitely the advantage of taking an approach like @Beefstu409's. I chose to ignore consideration of positional needs for this exercise because the ISFL has a unique feature that reduces the importance of this: positional switches. While not every player may be willing to change positions for the team that drafts them, the possibility of doing so can have a potentially huge impact because teams could be willing to draft a WR even when they don't need one if that WR then changes to a position they do actually need. In addition, the fact that we get credit for the pick even if a different team uses that pick via trade further increases the relative importance of accurate valuation rather than team fit. Those factors aside, I believe this approach could definitely be improved by finding a way to incorporate team fit.
Posted mock drafts being limited to only the 1st round also complicates the analysis, as I had no information as to whether posters thought that any specific player they didn't have in the first round would be drafted at, say, the 13th pick or the 50th. To solve this problem, I based the consensus rankings on each player's combined "win rate" against all other players in the draft. For each pair of players, I look across all mock drafts and keep score of which player is ranked higher in each mock draft. If one of the players is included in a mock draft but the other isn't, then the one that is in gets a win since it is clear that the ranker preferred the one that's in the mock draft. However, if neither player is included, then I assign neither a win since it's impossible to know which the ranker thinks will be drafter higher. I then calculate a combined win-loss record for each player by totaling their records in each individual matchup, and rank the players by their win%.
Beyond the Win% metric that I use to ultimately rank players for the consensus mock draft, I additionally calculated:
- The total number and proportion of mocks that included that player (#Mocks / %Mocks)
- A raw average draft position among mocks that included that player (Raw Avg)
- An adjusted average draft position (Adj Avg), where any player not included in a specific mock is treated as if they would be picked 13th, 14th, 15th, ... based on the highest ranked players by Win% not included in that mock.
- The standard deviation (Std Dev) of each player's adjusted rankings (i.e. those including the 13th, 14th, 15th, etc. calculated above)
- The number of times each prospect was mocked to each specific pick slot, including inferred 13th, 14th, etc. ranks (found on the Pick Frequency sheet). I think there could definitely be a way of using this to incorporate consensus positional fit metrics as well.
Highlights
- In addition to being the #1 consensus pick, Juan Domine is also picked in the most mock drafts with only 1 failing to include him.
- The most controversial prospect is Joel Drake, who was ranked in the top 3 28 times in addition to being left completely out of 15% of mock drafts.
- Mai Fukushu was the 2nd most controversial, and was in the fewest mock drafts of any players in the consensus top 12
- The first three players left out of the top 12 by consensus, in order, are Maverick Bowie, Brach Thomaslacher, and Captain Rogers.
- Despite being left out of 65% of mocks, Maverick Bowie had the 7th highest average pick position of any player when only looking at drafts that did include him.
![[Image: 67893_s.png]](http://signavatar.com/67893_s.png)