A Way-Too-Long .com Article (and a Way-Too-Long Title) in Which I Analyze My Mock Draft in Context

[Gustav 6,406]

 

I like articles too! And my favorite kind of article is a mock draft, where I write a lot and get some positive reactions and also probably piss off a few people. This season, I did my first mock since becoming a GM, and I later decided to find out how well I did, specifically in comparison to my other mock drafts. Those being...

 

• The original Gustav's Mock Draft, written by me as a rookie and still the article I'm most proud of. It managed to correctly predict over 40% of all picks in the first four rounds!
• A giant collaborative project for S68. Not my idea and actually that of @McWolf , and an experience I'd gladly have again.
• Another group mock in S69 where I took over the organization after the rest of the S68 group decided that was a one-and-done. Different crew, same thing, still a lot of fun.
• A S70 mock for VSN. Fun fact: I was briefly part of VSN, but this was the only VSN article I ever put out (and this one was a group project as well). I don't know why I never really did anything for VSN, but I guess I just prefer to freelance to the extent that I want complete ownership of everything I put out (and many of my articles have a community focus anyway rather than a sim results one). 

 

Anyway, that's all I wrote before making it up and getting an inside look at who was drafting who. But how does my S82 mock hold up to the others?

 

The picks I predicted correctly in S82 were:

 Bubbles Utonium at 2nd overall

 Perry Laperriere at 11th overall

 Maxwell Mathias at 15th overall

 Calix Pearce at 21st overall

 Phoebe Bridgers at 25th overall...

 

...although the portal gave me a bad list of which teams were drafting where and some teams are off. I've chosen to consider a "correct" pick one that went at the same overall spot as was projected, even though that makes picks like Mikhail Kovalchuk to Seattle technically incorrect even though I certainly would have projected him there regardless (it also takes away Eldon Escobar to Moscow, though the GM player factor isn't there for that pick. However, I still have 5 picks correct if we consider those players correct and yeet the same-overall-different-team picks in Mathias and Bridgers, so same difference). 

 

This was the worst out of every mock I've written in both number and percentage:

S66: 17 correct (34.0% of all picks)

S68: 11 correct (28.9% of all picks)

S70: 10 correct (41.7% of all picks)

S69: 6 correct (16.7% of all picks)

S82: 5 correct (15.6% of all picks)

 

Per this list, S70 was arguably the best in terms of percentage (helped massively that I projected the first 7 picks accurately) but I don't remember how much insight I had about that draft as I was asked last-minute to pick for Malmo (and was about to be hired in Davos). It was also not a very deep class, so there were fewer options at each pick. S66 was at the time historically deep, and I was still new, so I'd argue that it's still the best one here.

 

To avoid judging a mock by just looking at correct picks, let's look at each by evaluating the picks predicted within one pick of their actual selection:

S66: 22 picks within one (44.0% of all picks)

S68: 16 picks within one (42.1% of all picks)

S70: 14 picks within one (58.3% of all picks)

S69: 13 picks within one (36.9% of all picks)

S82: 10 picks within one (31.3% of all picks)

 

But what about consistency? If every incorrect pick in S70 was wildly off, correct guesses don't tell the entire story. I went a bit further and ranked every draft by the average distance each guess was from the correct one, as well as the standard deviation of that average. Though this isn't exactly true, a good way to interpret this is that if the average is greater than the standard deviation, the mock was more precise than accurate--guesses were relatively consistent, but might not necessarily have hit the target. Conversely, if the standard deviation is greater than the average, guesses were more accurate than precise--the correct pick could be expected to be close to any given guess, but the picks were spread out enough that seeing some outliers is nothing out of the ordinary. Ideally, both numbers should be as low as possible.

 

S70: Mean distance of 1.63 picks from actual selection (standard deviation of 1.93)

S69: Mean distance of 2.69 picks from actual selection (standard deviation of 2.27)

S66: Mean distance of 3.48 picks from actual selection (standard deviation of 4.91)

S82: Mean distance of 3.81 picks from actual selection (standard deviation of 3.13)

S68: Mean distance of 4.32 picks from actual selection (standard deviation of 4.77)

 

This one is interesting because S68, which looks great on the first two lists, doesn't hold up at all here, while S69, which doesn't, ends up being fairly accurate in general despite not directly hitting on many picks. It also demonstrates the power of outliers--one pick projected in S66 was 25 picks away from what actually happened (lower-level goalers are a lot of guesswork to project), and a couple others were similarly way higher or lower than I thought. With those taken out, S66 is probably fighting for second here.

 

It's fairly conclusive that S82 was my worst mock ever. Why? There are a few things which may have come into play. One, I used the wrong draft order and went by what the portal told me, so some team needs were likely neglected in their correct places. If I'd even been able to call Campbell to Seattle correct, I would have had 6 correct with a greater percentage than in S69. Two, there was a lot of guessing with the GM players. I'd heard ahead of the draft that every GM in the draft was attempting to tank their own stock, an assumption that was very likely incorrect in at least one case when Vasile Lamb went first overall. So, a false rumor messed with the status of one pick, while it was practically impossible to make an educated guess about the placement of DeGroot. And finally, I guess I'm just a bit rusty. I love writing mock drafts and knew exactly what I was doing by S70, but a few things probably wore off.

 

Until next season!

[McWolf 3,115]

1 hour ago, GustavMattias said:

A giant collaborative project for S68. Not my idea and actually that of @McWolf , and an experience I'd gladly have again.

that was a lot of fun. Whenever I step down, I'd gladly join a project like that again

[Shindigs 1,771]

I'm pretty sure @Spartan is the most happy he's ever been. Having read a .com with both standard deviation and mean of multiple things in it.

 

You also forgot that this draft had the inherent mystery of #TeamClue going in it. So you could just blame the three of us for all being picked in the wrong order and completely throwing off the top 10.