Calculating H2H Value for Trades/Draft

[jcairns]

Hi everyone,

I'm trying to figure out a way to accurately weight categories in head to head leagues to determine value in trades and drafts. The problem a lot of people have understanding is that:
1) Each category is worth the same amount. ie. winning Hits is worth as much as winning goals. But....
2) Goals and Assists are harder to come by and totals are lower

Therefore, what I've done is made a spreadsheet using dobber's projections, and normalized all categories with goals and assists. How I did this was averaged the totals in each category my league counts (ie. Goals, assists, shots, etc).

This gave the following averages for all forwards (numbers are rounded):

G: 15
A: 21
Hits: 78
Shots: 125
Blocks: 28

Then what I did was took the average of goals and assists (18), and divided the other categories (hits, shots, blocks) by numbers that made them the same (eg. Hits: 78/4.33 = 18; Shots: 125/6.94 = 18; Blocks: 28/1.55 = 18). So, from there you can simply divide all players' stats by the corresponding factors above (ie. divide hits by 4.33, shots by 6.94, etc). Now, the appropriate value for each stat contribution should be equivalent to goals and assists. Therefore, you can them SUM all of the point categories to get the player's actual value/worth.

With that said, this is of course not necessarily correct, and I'm looking for feedback to see if there is anything I am doing wrong, if this is worthwhile, how it can be improved, etc. Basically, I am just trying to figure out a way to appropriately determine how valuable each player is when you are in a H2H league.

[LawMan]

first off check out fantasyhockeygeek.com they basically do near what you have done but with a few improvements for any league. Once you input your league scoring, setup and roster size they combine that with the projections you select (one option is Dobber's projections) and create a ranking of all players at each position.

I think you're better off using a "team goal" for stats then an average. IF you are using an average make sure you're only averaging the number of skaters who will be drafted, for example my league will draft around 220 skaters but Dobber's projections go MUCH deeper, if I were to average the entire projections the numbers would be skewed. FHG will estimate the number of G, A, Hits etc. you need or you can include your own from previous years, is this way you build a top team without over-focusing on any one stat.

Also check out this thread for a lengthy debate and explanation about how best to use FHG for preparing for your draft and (potential) pitfalls in their rankings: http://forums.dobbersports.com/showthread.php?t=124889

[leaftodd]

I usually use an average of what it took to win the cat across the last 2 years as my bench mark. Then i try and tailor my team to be over that line. But i also only ever put money into 5 of the 6 cat's. I drop one from day one. (4 of the 5 in your case)

[Pengwin7]

I have been using "category score", which essentially totals up the amount of stats for a "pool quantity" of players. Average that. And give a player a score relative to the weight they contribute.

I've found that the top 144 skaters contribute:
48% of all NHL goals
45% of all NHL assists
26% of all NHL PIM
60% of all NHL PPP
39% of all NHL SOG

In other words, in order of stats, somebody should focus on PPP first, G second, etc...
As we know, PIM... Hits... Blocks, these have real value... true... but they can also be plucked off the waiver wire with ease.

See attached.

[jcairns]

I have been using "category score", which essentially totals up the amount of stats for a "pool quantity" of players. Average that. And give a player a score relative to the weight they contribute.

I've found that the top 144 skaters contribute:
48% of all NHL goals
45% of all NHL assists
26% of all NHL PIM
60% of all NHL PPP
39% of all NHL SOG

In other words, in order of stats, somebody should focus on PPP first, G second, etc...
As we know, PIM... Hits... Blocks, these have real value... true... but they can also be plucked off the waiver wire with ease.

See attached.

That's excellent! Thanks for the pdf to visually demonstrate. I'll give it a whirl. I understand how you did it, my only question is how did you jump from category totals (ie. Goals = 3821) to your weight (ie. G = 1.55 multiplier)?

[Pengwin7]

That's excellent! Thanks for the pdf to visually demonstrate. I'll give it a whirl. I understand how you did it, my only question is how did you jump from category totals (ie. Goals = 3821) to your weight (ie. G = 1.55 multiplier)?

For a long time, I wanted to see if I could weight the categories so that I created a "score" that, when sorted, would match the Yahoo! "rank".

I started with making a +1 = 1pt.
-1 = -1pt.
A player that was +30 has a +/- category score of 30.
A player that was -16 has a +/- category score of -16.

From there, I looked at a few players with minimal stats.
Like this: a player with 0G, 0A, -1, 0 PIM, 0 PPP, 2 SOG is still ranked as worse than a non-playing player. That told me that a SOG was worth less than +0.5pts. (Player would still be negative).

Step 1: I started attributing some "weights" to each category.
This gave me some totals for each player.

Step 2: I compared a players score to the player ranked after him. (A difference). I summed these difference for all 900+ skaters in the Yahoo database.

Step 3: I tinkered with the weighting so that this "sum of differences" started shrinking. I'd slowly step through the weights of all six categories, then repeat stepping through each one.

After a few hours, I had the sum of differences down to ZERO.
Essentially, the scoring with those "weights" will rank players in the EXACT position that a Yahoo! "Rank" sort will place them.

Neat, eh.

(My most prized achievement of the off-season!!!)

[Pengwin7]

For those of us mathies... the negative players with very few stats are a great way to guestimate some weightings!!!

See attached.

[jcairns]

Very impressive! I don't suppose you happen to have figured out the weights for Hits and Blocks and care to share them?