Talent Distribution Curve Improves on Ordinal Ranks
Major league talent drop is not linear and it shouldn’t be treated as so in fantasy baseball but it does for many owners because we love our simple ordinal rankings. The talent drop from the best shortstop to the 10th is more than the drop from the 10th to the 20th. As soon as an owner moves away from using just rankings and goes to an overall production value they will gain a leg up on the competition. The whole idea can be explained by the talent distribution curve.
The talent distribution curve can be created with auction dollars. Other values can be used such as raw Standings Gain Points. But setting our auction calculator to the standard 12-team roto settings. Here is the talent distribution curve:
Note: While only 276 players are needed for this example, I extended the data out for deeper leagues. For those in 12-team -Only leagues, they will be picking at the 552 mark where the talent really begins to drop.
I’m not even sure if it has a shape. Maybe a sideways ‘S’. There is an extreme drop for about 100 picks and then starts to level off. Owners want to acquire as many players in this first group. In this part of the draft, talent needs to be prioritized over a position since talent differences exist more at this level.
Put another way, half the picks are in the $37 drop from Trout’s $46 value to the $9 rank. In the last $8 form the $9 to the $1 players are the other half.
This curve has always existed. There has never been a linear distribution of talent in real baseball and it extends to fantasy also. There way more 1 WAR players than 6 WAR or even 3 WAR players. In a draft, secure as much of the top end talent. In auctions, the strategy can differ since owners don’t have to try to pick the best player at a given time. They can pick and choose how and may bunch their talent together.
There are two flattenings in the talent curve. To show them, here are just the 276 picks for the 12-team league.
The first leveling happens around pick 50 and goes to pick 150. After pick 150, it extends for the next 550 picks. Here are those last 550 picks viewed graphically.
An even flatter segment could be created if I cut off the first and last few picks. Instead, they help to show when the drop is no longer linear but starts to turn up and down.
Now, the drop around pick 500 is important in deep leagues. Owners don’t want to be picking up their talent once it transitions into the Quad-A or bench player talent pool.
The non-linear drop is even more important when just using position rankings, like with shortstop. Carlos Correa is a near $20 value as the #3 shortstop (previous ranking discussion with Lindor and him). Going down to the #8 spot is Jean Segura at $7. If moving down another 5 spots is Didi Gregorious at #13 and $4. The first five-player drop was $13. The second one was $3.
Some owners like to use several tiers and that approach can be a good compromise. The one issue with tiers is to make sure the values jive from one position to the next.
In today’s fantasy game where little to no position scarcity exists besides at catcher, the top five first basemen may be in a tier with only two shortstops. In the next tier, no first basemen may meet that level’s criteria while four shortstops may be in it. Don’t just assume the same number of players from each position will be in each tier.
The talent distribution curve is a simple concept. The curve as a logarithmic shape to start but later levels off. The problem is that fantasy owners are given players in ranks. These ranks change linearly but the talent doesn’t. It’s better to use a system with gives players a talent value (like dollars) and use it to determine who to pick.
Jeff, one of the authors of the fantasy baseball guide,The Process, writes for RotoGraphs, The Hardball Times, Rotowire, Baseball America, and BaseballHQ. He has been nominated for two SABR Analytics Research Award for Contemporary Analysis and won it in 2013 in tandem with Bill Petti. He has won four FSWA Awards including on for his Mining the News series. He's won Tout Wars three times, LABR twice, and got his first NFBC Main Event win in 2021. Follow him on Twitter @jeffwzimmerman.
So consolidating talent is almost always a good idea?
I just made two trades, I got Albies for Stroman, Verdugo and a 9 pick swap in the second round of our fyp draft.
And I got Bregman for 1.01 (Ohtani) and 3.01 pick 40ish
Keep forever no contracts.
Rule of thumb in even trades has seemed to be ‘get the best player in the trade’. Looks like this validates the conventional wisdom
I think this means that if you screw up the in most of the first few rounds, expect to lose.
I asked Rudy Gamble who has finished 2nd, 2nd, and 1st in the Tout Wars draft and his secret was his first three picks each year were healthy and productive.
Broken link in the 4th-to-last paragraph. Otherwise good article!
This does not change anything from a snake draft standpoint. Just get the best guys you can
Well I disagree that that because I see that this points to value based drafting and positional scarcity. You have to weigh the drop in talent/stats within positions and across positions. “The best” guy you can take is not necessarily the best guy available, but the best guy when compared to the talent drop of the next guy(s) when accounting for positional scarcity.
What position scarcity (besides with catcher).
Positional scarcity at any given tier. Lets say SP is fairly linear drop from this point in the draft, lets say round 5 or 6 of a 12 teamer. Lets also say there is only 1 more good 2B before a severe drop.
Do you get the ‘fair’ or ‘best guy’ SP pick or reach a couple spots for the 2b?
the reason that there is no positional scarcity is because of the point of this article. There isn’t really such a thing as “last 2b left before a big drop,” unless you’re already talking about taking a 2b who shouldn’t be starting for you in the first place.
In today’s fantasy game where little to no position scarcity exists besides at catcher…
Your particular league rules may render this moot, but in general THIS year it looks like positional scarcity isn’t a thing they way it has been in the past. Yes, particular drop-offs do occur at various points in each position. No, middle infielders shouldn’t automatically get a bump just because historically SS can’t hit very well.
I am speaking from a perspective of a deep league who starts 2 players at each position (4 OF) and has some unique 7×7 categories so my perspective is not the norm.
I do understand that there isn’t the same scarcity today that there used to be, but my point isn’t necessarily about a lack of any value in certain positions, rather the value of the drop that exists across the players who will be drafted within a position. The drop in stats in some positions will be greater than in others.
So would you say the best way to valuate players in fantasy is to go in the auction calc, set it up with your team settings and go off of those $ values? Even in non auction leagues?
It’s one way. I use a little more convoluted method in which I didn’t want to go into. The main reason I use my method is that I can manually change a projection if I don’t agree (mainly playing time).
Another complication for players in auction leagues: the value curve and the cost curve (ie, $ spent at auction) have two different shapes! I found last year that a correction factor can be applied to more accurately predict auction values: https://www.getbigboard.com/harper-wallbanger/2017/2/27/calculating-better-auction-prices-with-pre-inflation
So happy I saw this post. I ran into the same things when running my auction values- not enough $1 players and the value of top guys was a little low. I was thinking of screwing with the projections of each guy, but this method is much easier!
If the inflation exists to this level, abuse those who pay the inflation (I wrote a full article on it for The Fantasy Baseball Guide). Owners have $12 to spend on each player. I just wait for about the $20 point to jump in and load up (these are around 5th to 6th round players). I’ll still play the cheap endgame but just limit the loss from the high-end inflation. I call it Lesser Stars and Scrubs.
I’ve observed very similar inflation in both my Baseball and Football auctions, so it’s a behavior that drafters exhibit. I think there’s another draft strategy you can employ based on this data, which is to go Stars and Perceived Scrubs by identifying guys on the value curve that have an auction value cheaper than the actual value they’re projected to return. I realize that’s the point of any pick – to get surplus value based expected return of draft slot or auction cost, but, if you can identify those guys that’ll cost you $3-5 and return $10 of value it can offset the cost of getting 2-3 high end talents.
Jeff- if you see this behavior happening year over year, but owners employing that strategy are winning, do yo consider that maybe you should do the same? I guess the question is- how do you validate your projections and methodology are correct? Obviously winning the league can help, but if you are coming close to winning but finishing 2-4, how much do you look to adjust what you are doing vs. attributing it to normal variance or in season management?
I’ve employed the strategy in Tout Wars this year and won by a record amount with a nearly perfect pitching score. The hardest part is to sit out the first half of the auction and own no one:
Auction:
https://docs.google.com/spreadsheets/d/1XXvdV_obvo4pk4xIMZBtyAXWRHu7ZTHKBKUMObxKugY/edit#gid=5
Results:
http://baseball7.onroto.com/baseball/webnew/display_live_scoring.pl?toutmixed+6+stand&session_id=w6ubaq5MowL5nONPMk8D71yOTxgkUrna7h
I’ve always wondered if by including the average auction value in a Yahoo! (or any other site) draft, does it anchor people to that value? If there’s data comparing auction results from a site showing avg. values vs. no values I’d be interested in seeing that.
Fun Fact: assuming Bregman will be drafted as a SS, the largest drop in positional value I can find is at 3B — $25.6 Donaldson to $13.5/11.6 Gallo/Rendon. (Fangraphs Auction Calculator with a basic 1C, 3OF, 2UT setup was used.)
This got me thinking… are we accounting for such precipitous drops in the current projection rankings?
With the exception of catcher, all current positional replacement values are roughly the same. But, imagine if all OF’s except Trout are worth OF replacement level. This does nothing to change Trout’s current value after positional adjustment, but does this hypothetical (albeit comical) large drop not make him worth more?
Not really because smart owners would know the OF are at the draft’s end and are picking the other positions until then.
Jeff, thanks for the response! In practice, I agree that large drops tend not to matter.
In theory, however, I am not convinced that my silly hypothetical is as simple to analyze. Assuming the situation above, the team that gets Trout has a ~$45 OF, while all others are at $0. That’s a difference that no non-Trout team can make up for. I think this makes Trout WAY more valuable than his price tag states
But of course, this is not due to the drop, but from the fact that Trout is worth literally ALL of the OF value… this is a whole new can of worms I’m opening 😛
I have felt this for a while, but I haven’t seen it in data. You can just tell when you are at a tier of a certain position where you are no longer in a hurry to draft someone at that position. I am currently in a slow draft and have SS to fill. I am looking at Anderson, Baez, Semien, Polanco, Marwin, Arcia. I do have preferences at this tier, but the reality is Matt Olson, Smoak, Andrew Miller, Jon Gray and Domingo Santana are still on board so why on earth would I fill my last remaining starting position now?
And I should have written some on reaching. It’s better to reach for a scarce position when the talent curve flattens out so it is a $1 or $2 reach instead of a $5 reach.