The Data-Based Drafting Thread (what players would a Potato pick?)

[Melvin]

Where did I say that? I've already told you I've been using a similar methodology since the 80s, but it's easy to see that junior leagues scoring leaders are very often 19 and 20 year olds, it's just almost apples to oranges, for example did you run your numbers to include just undrafted when you included 19 year olds? Maybe try it with all 19 year olds and see where he falls?

See where he falls compared to what? The point is to rank him with other draft-eligibles. His numbers are discounted to account for the fact that he is a second-time eligible.

You seem to believe they are not being discounted enough? How much do you think they should be discounted?

 

[Chuck Feathers]

See where he falls compared to what? The point is to rank him with other draft-eligibles. His numbers are discounted to account for the fact that he is a second-time eligible.

You seem to believe they are not being discounted enough? How much do you think they should be discounted?

What I am saying is, run the numbers from this season for all 19 year olds drafted or not and see where he falls, my guess is he will still be well down in the rankings despite the step forward he took this year and that to me would be a strong indicator... likewise if he is in fact well up on that list, then he may well be worth the chance this year but even then I doubt very much teams would take a chance selecting a player with a top 10 pick who wasn't even good enough to be a 7th rounder in his draft year.

Take it a step further and see how many undrafted 20 year olds would show up in your rankings... you say his numbers are discounted for his extra year but I just wonder how much data there is on undrafted 19 year olds to accurately adjust for that? That's why I would run it for all 19 year olds to get a sense of just how much he may have moved up..

 

[Melvin]

What I am saying is, run the numbers from this season for all 19 year olds drafted or not and see where he falls, my guess is he will still be well down in the rankings despite the step forward he took this year and that to me would be a strong indicator... likewise if he is in fact well up on that list, then he may well be worth the chance this year but even then I doubt very much teams would take a chance selecting a player with a top 10 pick who wasn't even good enough to be a 7th rounder in his draft year.

Take it a step further and see how many undrafted 20 year olds would show up in your rankings... you say his numbers are discounted for his extra year but I just wonder how much data there is on undrafted 19 year olds to accurately adjust for that? That's why I would run it for all 19 year olds to get a sense of just how much he may have moved up..

But now you're really comparing apples to oranges because why would I care about how he rates as an undrafted 19 year old compared to drafted 19 year olds? How would you calculate, from that, the appropriate % to discount his numbers? Like, this is a math-based system, I'm not moving players up and down based on my gut.

 

[Chuck Feathers]

But now you're really comparing apples to oranges because why would I care about how he rates as an undrafted 19 year old compared to drafted 19 year olds? How would you calculate, from that, the appropriate % to discount his numbers? Like, this is a math-based system, I'm not moving players up and down based on my gut.

Why would you care? Because it's 19 y/o vs 19 y/o...it's a much better indicator than 18 v 19.

Math based is what I am getting at... if you include all 19 year olds you can quantify how much that entire age group improved statistically since their 18 y/o year... if you don't do that how are determining how much to discount the rare undrafted 19 y/o that shows up high in your rankings?

 

[Melvin]

Why would you care? Because it's 19 y/o vs 19 y/o...it's a much better indicator than 18 v 19.

Math based is what I am getting at... if you include all 19 year olds you can quantify how much that entire age group improved statistically since their 18 y/o year... if you don't do that how are determining how much to discount the rare undrafted 19 y/o that shows up high in your rankings?

Because I can compare how well a 19 y/o putting up certain numbers projects compared to 18 y/o's putting up the same numbers.

I have some 1,195 repeat-eligibles in my database. You are right that there might be something to get out of including already-drafted players, but I feel that the very fact that they've been drafted is extra information that would bias things significantly.

But you know what? It's something for me to think about. So thank you.

 

[Chuck Feathers]

Because I can compare how well a 19 y/o putting up certain numbers projects compared to 18 y/o's putting up the same numbers.

I have some 1,195 repeat-eligibles in my database. You are right that there might be something to get out of including already-drafted players, but I feel that the very fact that they've been drafted is extra information that would bias things significantly.

But you know what? It's something for me to think about. So thank you.

Absolutely, I appreciate that you've put all that work into something that I never got around to doing even once I got good at database development. As I said, I used to do the same thing only manually with the NHL OG&RB way back in the day, so while we may not agree on some things, it's cool to see someone else came to the same conclusions about how to establish an objective baseline ranking system that actually can help result in improved selection.

 

[DomY]

@Melvin I was thinking more about your drafting system, and while I do not totally subscribe to the theory that Benning is a below average drafter, I am starting to come around to the belief that it would be better to have a potato be in charge of your drafting than having a full time amateur scouting staff. It's the same as passive investing thru index funds rather than actively managed portfolios - something like 90% of them fail to outperform the index.

Have you ever evaluated, using your potato method, the value of trading down in the first round? Obviously we would have to play around with the formula. To standardize it, I would suggest the following:

Picks in 1-2 range - do not trade
Picks in 3-10 range - 3 position drop, plus a 2nd and 3rd round pick from the same draft of the new position. For example, if you traded the 3rd OA pick, you'd get back the 6th OA, 37th OA, and 68th OA.
Picks in 11-20 range - 5 position drop, plus a 2nd pick from the same draft
Picks in the 21-31 range - 5 position drop, plus a 3rd pick from the same draft

I would be curious to see if the overall score would improve if this happens; alternatively, if your system could discover the "break even point" whether trading down would be overall beneficial.

 

[vancityluongo]

It's the same as passive investing thru index funds rather than actively managed portfolios - something like 90% of them fail to outperform the index.

It is exactly the same.

Just as it would be insane to make stock picks based on gut feelings or by assessing the character of the management team, it is also insane to think that you can just extract certain "trends" or numbers and buy that way. The best option truly is to diversify a portfolio and passively manage it to minimize fees.

Similarly, the best drafting strategies will be around making lots of picks (diversified portfolio) and not overthinking it. If a player (company) has really good numbers (cashflow), chances are they will succeed. Not a guarantee, but a decent enough chance.

 

[DomY]

I always joke to my friends who are killer at fantasy sports but are daunted by the stock market that they are essentially the same skillset.

With that in mind, we should be increasing our cashflow (acquiring more picks and trading down) unless there is a break even point where it makes sense to keep the pick or otherwise trade up.

Using my own advice, I guess the Canucks shouldn't try to trade up for Byram given the negative balance they have on the books.

 

[Melvin]

@Melvin I was thinking more about your drafting system, and while I do not totally subscribe to the theory that Benning is a below average drafter, I am starting to come around to the belief that it would be better to have a potato be in charge of your drafting than having a full time amateur scouting staff. It's the same as passive investing thru index funds rather than actively managed portfolios - something like 90% of them fail to outperform the index.

Have you ever evaluated, using your potato method, the value of trading down in the first round? Obviously we would have to play around with the formula. To standardize it, I would suggest the following:

Picks in 1-2 range - do not trade
Picks in 3-10 range - 3 position drop, plus a 2nd and 3rd round pick from the same draft of the new position. For example, if you traded the 3rd OA pick, you'd get back the 6th OA, 37th OA, and 68th OA.
Picks in 11-20 range - 5 position drop, plus a 2nd pick from the same draft
Picks in the 21-31 range - 5 position drop, plus a 3rd pick from the same draft

I would be curious to see if the overall score would improve if this happens; alternatively, if your system could discover the "break even point" whether trading down would be overall beneficial.

The money management analogy is a good one and one I've made before,but it's important that I never said Benning was bad at drafting. I just don't believe "drafting" is a skill that exists in a way that we can meaningfully measure it. No gm has proven to be consistently better at picking players than my simplistic, stupid model which should serve as the bare minimum that any team should be able to beat. Whether or not Benning is "good" or "bad" relative to other gms is essentially gibberish to me.

I would be interested in doing what you propose. I've been thinking about this a lot, the relative strategy of reading up and down. My hunch is that teams should trade down for virtually every pick. The hard part is figuring out exactly what picks can be realistically traded for which picks which might require a lot of historical digging.

Ideally id like to have a formula to evaluate pickX = pickX + y + pickX + z

 

[Melvin]

Btw, congrats on the big contract given to Esa Lindell, picked by the potato in 2012 over Brendan Gaunce.

 

[Jyrki21]

Btw, congrats on the big contract given to Esa Lindell, picked by the potato in 2012 over Brendan Gaunce.

I wanted to make this remark when he was re-signed.

Here’s the thing – you can hold the potato up as a model of mindless simplicity, but the reality is that if humans aren’t meeting that bar, and it therefore remains the best thing out there, perhaps we should also (or instead) start thinking of it as a reasonable model for informed decision making.

 

[Jyrki21]

@MS, check your PMs!

 

[DomY]

I would be interested in doing what you propose. I've been thinking about this a lot, the relative strategy of reading up and down. My hunch is that teams should trade down for virtually every pick. The hard part is figuring out exactly what picks can be realistically traded for which picks which might require a lot of historical digging.

Ideally id like to have a formula to evaluate pickX = pickX + y + pickX + z

I’ve proposed a rough formula for trading down in the first round in my first post. I used the 2008 draft as a blueprint as there were a huge amount of traded picks across the entire first round. 5 (Schenn) was traded for 7 a second and a third all in the same round.

I think I would tweak my formula slightly to try to be able to add a 2/3/4/5 round pick in the next draft as well. This could be randomized as usually teams are willing to add future year picks as a kicker based on how badly they want the player they’re moving up to select to; it’s less to do with the nominal value of the future pick.

It’s also something I’ve picked up on that the Toronto Blue Jays do in every one of their trades - they always acquire a young starting pitcher prospect at least four years away in any trade they make. These lottery picks are starting to make an impact on the farm system even more than the main prospect acquired in the sell off of major league assets.

Let me know if you need any help with this. I’d be happy to grind thru some simulations with you.

 

[604]

It is exactly the same.

Just as it would be insane to make stock picks based on gut feelings or by assessing the character of the management team, it is also insane to think that you can just extract certain "trends" or numbers and buy that way. The best option truly is to diversify a portfolio and passively manage it to minimize fees.

Similarly, the best drafting strategies will be around making lots of picks (diversified portfolio) and not overthinking it. If a player (company) has really good numbers (cashflow), chances are they will succeed. Not a guarantee, but a decent enough chance.

That's pretty simple thinking that is generally sold to everyone.

The best strategy in my opinion is setting conservative exit points and entering and existing the market, on a leveraged basis based on general market trends. Can't be afraid to exit the market at the first sign of a downturn and re-enter as the market recovers.

 

[racerjoe]

I think the potato does a great job of giving you a baseline. Showing where some players are undervalued. Once you know this you can start to move people around a little bit depending on what you see from a prospect. I love seeing this type of stuff, it shows where the people are really undervalued. I think it is more accurate then most other lists we get that come out too.

 

[MS]

The money management analogy is a good one and one I've made before,but it's important that I never said Benning was bad at drafting. I just don't believe "drafting" is a skill that exists in a way that we can meaningfully measure it. No gm has proven to be consistently better at picking players than my simplistic, stupid model which should serve as the bare minimum that any team should be able to beat. Whether or not Benning is "good" or "bad" relative to other gms is essentially gibberish to me.

I would be interested in doing what you propose. I've been thinking about this a lot, the relative strategy of reading up and down. My hunch is that teams should trade down for virtually every pick. The hard part is figuring out exactly what picks can be realistically traded for which picks which might require a lot of historical digging.

Ideally id like to have a formula to evaluate pickX = pickX + y + pickX + z

Trading down with a top-5 pick is usually a bad decision, I’m guessing.

Past that, yes, I’m pretty sure trading down is almost always the right decision assuming you can get a decent return (ie pick 25 for pick 35 and pick 185 probably isn’t good) and assuming there isn’t a blindingly obvious potato-pick that has taken a huge drop to your draft spot.

 

[Frankie Blueberries]

Great post, and Canucks would undoubtedly be in a better position following your algorithm.

Sorry if it’s already been mentioned, but what stood out for Makar, Heiskanen, and Hughes not making the lists?

 

[The Poacher]

I’ve proposed a rough formula for trading down in the first round in my first post. I used the 2008 draft as a blueprint as there were a huge amount of traded picks across the entire first round. 5 (Schenn) was traded for 7 a second and a third all in the same round.

I think I would tweak my formula slightly to try to be able to add a 2/3/4/5 round pick in the next draft as well. This could be randomized as usually teams are willing to add future year picks as a kicker based on how badly they want the player they’re moving up to select to; it’s less to do with the nominal value of the future pick.

It’s also something I’ve picked up on that the Toronto Blue Jays do in every one of their trades - they always acquire a young starting pitcher prospect at least four years away in any trade they make. These lottery picks are starting to make an impact on the farm system even more than the main prospect acquired in the sell off of major league assets.

Let me know if you need any help with this. I’d be happy to grind thru some simulations with you.

I was also wondering if a team can add value by trading down to add more picks but not targeting current year picks and try to get picks from future drafts. A 2nd round pick in 2019 should be worth more than a second round pick in 2020 which should be worth more than a second round pick in 2021. My question is instead of looking to add picks in the current years draft can you add more value by looking for picks that are always 2 years away.

Is a 2021 2nd round pick worth less than a 2019 3rd round pick to any given team?

 

[Dissonance Jr]

I was also wondering if a team can add value by trading down to add more picks but not targeting current year picks and try to get picks from future drafts. A 2nd round pick in 2019 should be worth more than a second round pick in 2020 which should be worth more than a second round pick in 2021. My question is instead of looking to add picks in the current years draft can you add more value by looking for picks that are always 2 years away.

Is a 2021 2nd round pick worth less than a 2019 3rd round pick to any given team?

This is an interesting question. Another way to put it: what sort of discount rate should we apply to future draft picks, and do NHL GMs irrationally use too high of a discount rate?

Like, in finance, getting $100 today is worth more than getting $100 a year from now because of inflation and the opportunity cost of investing that $100 for a year. So you might reasonably ask for a discount rate of (say) 7% or so—that is, getting $100 today is roughly worth getting $107 a year from now.

But does it make sense to apply a high discount rate for draft picks? There’s no inflation here and it’s unclear how big the opportunity costs of waiting a year actually are.

NHL GMs pretty clearly discount future draft picks heavily because they know their tenure as GM is limited—so a 3rd round pick today is more valuable to them personally than a 2nd round pick two years from now (the further off the pick, the less chance there is that the current GM will reap any benefits from it). But that’s not necessarily rational from the standpoint of the franchise, which will likely be around for decades.

So I dunno, there’s possibly a big arbitrage opportunity here but good luck convincing any GMs or owners to take advantage of it.

 

[DomY]

I was also wondering if a team can add value by trading down to add more picks but not targeting current year picks and try to get picks from future drafts. A 2nd round pick in 2019 should be worth more than a second round pick in 2020 which should be worth more than a second round pick in 2021. My question is instead of looking to add picks in the current years draft can you add more value by looking for picks that are always 2 years away.

Is a 2021 2nd round pick worth less than a 2019 3rd round pick to any given team?

To use another personal finance analogy, when you’re saving for retirement your time horizon is usually 30-40 years when you start saving for retirement.

GMs at the draft have 1-5 years to make an impact due to their own shelf life, and this is something I believe has never been properly exploited in the NHL. In the NBA you see teams trade picks from 4 years out. I don’t even know what the rule is in the NHL but I’m guessing that GMs have access to picks at least four years away (Offer Sheet Compensation). I’ve only seen them trade picks two years away though.

Also to answer your question, if a team trying to trade up in the first round and offers you a lower first and the option of a 3rd in the current year or a 2nd in the next year, or a 2+3 in 2 years, I’m not entirely sure which to take. On the face the 2+3 has the most nominal value, but the 2nd next year might be a mid 2nd while the one in two years might be a bottom 4 pick if the team makes the conference finals. I would have to evaluate the historical weighting of a mid 2nd vs a bottom 2nd. My gut tells me you should take the future picks though.

If I were Jim Benning and I didn’t fear for my job security, I would trade the 10OA two or 3 times and try to acquire an array of future picks then repeat the exercise next year too.

Using the example of the 1999 TB lightning, they turned the 1OA pick (in probably the weakest draft in history into the following:
1OA for <#4> #75 #88 in 1999, and #4 was flipped for Dan Cloutier, Patrick Sundstrom, #8 and #74 in 2000

 

[DomY]

Using another example, in 1998 the SJS were gifted the 1OA pick in a trade made during the season with the Panthers for Viktor Kozlov.

The Sharks #1OA (Vinny Lecavalier) for #2OA (David Legwand) and eventually traded down to #3OA (Brad Stuart). For their trouble they received Bryan Marchment, David Shaw (?), and 29OA (Jonathan Cheechoo) while also surrendering Andre Nazarov and 85 OA.

Lost:
Nazarov: 74pts and 402 GP (in career after leaving SJS)

Traded:
Lecavalier: 949pts and 1212 GP
Legwand: 618pts and 1136 GP

Received:
Marchment: 66pts 322 GP (in career after joining SJS)
Stuart: 335pts 1056 GP
Cheechoo: 305pts 501 GP
Total: 706pts 1879 GP

Net gain for Sharks = (706-74 = 632pts) (1879-402 = 1477 GP)

Lecavalier was not a bust of a 1st overall and was extremely highly touted coming out of junior. Unless there is more to the story that I am unaware of, Sharks would have been better off keeping 1st OA.

Of course, Cheechoo was eventually traded in a package for Heatley and Stuart in a package for Thornton, but those were value proposition type of trades, and Cheechoo and Stuart adequately stored trade value at the time of their trades, not because they were sought after by the other teams.

 

[MS]

Both of those trades turned out horribly for the team trading down.

I agree that trading down is usually a good call, but trading down from a top-3 pick is pretty much always an atrocious call.

 

[drax0s]

Update: I did.

Congrats dude. Glad to see it.

 

[drax0s]

I would be interested in doing what you propose. I've been thinking about this a lot, the relative strategy of reading up and down. My hunch is that teams should trade down for virtually every pick. The hard part is figuring out exactly what picks can be realistically traded for which picks which might require a lot of historical digging.

Ideally id like to have a formula to evaluate pickX = pickX + y + pickX + z

I think we were talking about something related a while back in another draft thread I think. Figuring out sort of a "relative value" for a pick based on previous years comparables to see if you're getting good value for a pick year X vs year Y. Not all top 10 picks have the same value.

That said, I can easily see coming in as an analytics guy being of beholden to things "going right" for you to stick around. Picking the right statistical selection and whiffing on players a few years in a row is totally possible and doesn't disprove the method. Extra picks would definitely help to soften out the impact of randomness.