What stats matter for dynasty prospects? An introduction to Correlation Scores

Introduction

I've had a few people ask me where to start when it comes to scouting draft prospects for dynasty rookie drafts and what my thought process is. This got me thinking of my process, and truthfully I couldn't give a concrete, systematic answer. And that bothered me. Now don't get me wrong, there are things I look for and patterns I've noticed combing through the years of data, but I didn't have specific justification for valuing those stats and trends over others. That brings me to today, where I finally crunched the numbers and ran the tests to find what pre-NBA statistics are most correlated with fantasy basketball success.

Testing:

The sample: drafted players between the 2008 and 2019 draft classes (I wanted players that have played at least 5 full seasons in the NBA)

The test: finding the correlation between various pre-NBA metrics and peak season 8-category fantasy value. I found the r value for these metrics for the entire sample, and then also split the sample by height as a proxy for position.

The results: 

Let's get right into the general takeaways:

• As always, age-adjusted production is king

• How highly a player is drafted determines opportunity. The lower a prospect goes in the draft the less chances they’ll get to show themselves. In rookie drafts, deviating too far from the consensus order (outside of extreme, Filipowski-like circumstances), not only lowers your team value but decreases your odds of success

• A lower or negative correlation doesn't mean that the metric is actually negatively correlated with success or is unimportant, per se. It just means that players have found success in fantasy all across the spectrum of performance for that metric. I'm thinking about it like this: performance in any metric can be a red or a green flag, but the more correlated with fantasy success a metric is, the redder or greener that flag is. Yes "redder" is a word, I googled it

• The r values for shooting are wonky because there's a sweet spot. You want prospects to shoot threes showing that they have an ability to do so, but you don't want them to shoot too many indicating they’re limited to a specific role

• I've always somewhat believed in this as a concept, but the results have given me more confidence: defensive rebounding is about opportunity, offensive rounding is about skill

• Steals are a proxy for athletic dominance and feel for the game. They're really important

• Net Rating, aka team performance when the prospect is on the court, is super valuable but requires a ton of context

  
  

Now let's get into some position specific takeaways, starting with the Guards (I’m going to refer to the 6’5 and lower subset as guards for convenience):

• College production is much less predictive of NBA success than it is at other positions. This means that evaluators should project skills and talent more at the guard position than they typically would. It's also the position where pedigree (HS Rating) matters the most

• Rim pressure and self-creation are the star traits

• Passing volume appears to be more important than passing efficiency.

• Low usage pre-NBA seasons are less debilitating for backcourt players

  
  

Wings

• Passing, passing, passing. Star wings and big guards are playmakers.

  
  

Bigs

• Production is everything. No excuses, either they filled up the stat-sheet or they didn't. Pedigree means nothing if they didn't perform

• The best bigs are the ones who had high usage and could create for themselves. They can still be a great asset if they don't, but the ceiling is capped

• Even if a player can get a bucket, they still have to do big man things. Rebounding, finishing at the rim, blocking shots, etc.

• So few bigs create for themselves from three, so the ones that do have a sky-high ceiling

Correlation Scores and Analysis

Note: If you are just here for the analysis and results of the 2024 and 2025 draft classes, skip to the bottom.

This was originally the end of my analysis, until I saw a tweet by @criggsNBA (a great follow for general NBA draft content). In his tweet, Colton created an all-in-one metric for prospects, with each individual input weighted by how correlated it is with VORP/year. This inspired me to do the same but with how correlated the inputs are with peak 8-category value. I then normalized the individual scores to a 0-100 scale, thus creating Correlation Scores. Ex. here are the Correlation Scores for the first round of the 2024 draft:

To test the true value Correlation Scores I had to find one more test to run: to see how predictive the new cumulative Correlation Scores were of peak 8-category value. My hope was for Correlation Scores to be more predictive of future fantasy value than NBA Draft order, making my metric more relevant to dynasty drafts than just linearly following how NBA teams draft. Some might view this a low bar, or even a prerequisite, for any process or metric to be taken seriously, but I think you’d be surprised at how important and useful draft capital is as an indicator of talent and NBA opportunity. 

Below are the r values for Correlation Score and peak fantasy value, as well as the values for NBA draft order, and the difference between the two.

Thankfully, Correlation Scores accomplished just what I hoped it would, a higher overall r value than the NBA draft order. Interestingly, as you go up in height, Correlation Score becomes a more valuable predictor of fantasy value than the NBA Draft. To me, the r values showing that draft position is a more valuable predictor than pre-NBA production for players 6’5 and under comes down to a few things:

• The shorter you are, the more adjustment time is required to handle the size and physicality of a new level. This means that some smaller players might struggle statistically in the relatively small sample of an NCAA season, but will start playing their true ability with time as they adjust. Not to mention that the point guard position is often considered the most mentally demanding position on the court, leading to longer acclimation periods that might negatively impact a prospect’s statistical production. This also explains why a prospect’s high school rating is much more correlated with success for shorter players, while barely a factor at all for taller prospects. Short guards have notoriously also had longer development curves in the NBA, with players like Darius Garland and De’Aaron Fox struggling mightily as rookies only to turn things around in subsequent years.

• On the flip side, because many shorter players are outmatched physically and it's difficult to overcome that physicality difference and master the mental aspects of the game from the sideline, development in the guard position often comes down to reps and playing time. This makes a guarantee of minutes and a long leash key for the growth of shorter players, and the players that are most likely going to get those early minutes are players that teams have invested high draft picks on. This contributes to a heightened importance of draft capital as we go down in height. Even if a taller player’s game is not as polished as it should be to typically warrant NBA minutes, their physical advantages caj allow them to see the court early on whereas shorter players might need artificial factors (such as draft capital incentives) to get those key development reps.

  
  

Even if Correlation Scores aren’t enough to justify selecting a guard with great production over a less impressive guard selected earlier in the draft, it can still be a useful differentiator between guards within the same draft capital tiers. To showcase this, let’s split and compare the 28 late lottery (picks 10-14) guards in the database in half, based only on their Correlation Scores:

	Top Half of CS	Bottom Half of CS
Top 50 Fantasy Seasons	16	4
Top 25 Fantasy Seasons	8	1
All-Star Appearances	8	2
All-NBA Selections	4	0
Avg. 9cat Ranking Y3	194	265

The trend continues among guards taken at the end of the first round (picks 26-30):

	Top Half of CS	Bottom Half of CS
Top 100 Fantasy Seasons	20	3
Top 50 Fantasy Seasons	8	0
All-Star Appearances	1	0
All-NBA Selections	0	0
Avg. 9cat Ranking Y3	242	261

When picking between guards of the same general range, Correlation Score can still play a meaningful role in getting that decision right.

As nice as it would be to end the analysis at the conclusion that Correlation Scores are more predictive than the NBA Draft order, there are issues with the data set that need to be addressed. The world of draft prospect data collection is moving more and more toward gatekeeping and pricing out the casual (thank you Synergy), and because of that I don’t have access to a large swath of advanced stats for non-NCAA prospects. This leads to a sort of input asymmetry, with players who played in international leagues or on the GLeague Ignite lacking some of the self creation and rim stats that NCAA players in the database have. To look at how this impacts the overall results and viability of Correlation Scores, I broke the sample down into players who played in the NCAA with all available stats, and players who are missing stats. The results with their correlations to fantasy value are below:

These results tell a pretty similar story to the overall correlations. While Correlation Scores are less predictive for non-NCAA prospects across the board, so is draft capital – and the difference between the two actually favors Correlation Scores more than it does for both the NCAA and overall samples. These results also show that it’s generally harder to predict the success of prospects who didn’t go to college, whether using statistics or the NBA draft order. What also transfers over from the overall analysis is the Correlation Scores’ inability to outperform the NBA when it comes to projecting guards. This reaffirms an emphasis on tools, traits, pedigree, and draft capital when projecting how smaller players will translate to the NBA.

To me, the proper way to evaluate international and GLeague prospects is to strictly compare them to each other. For example, Alperen Sengun’s Correlation Score of 81.0 might undersell just how good of a prospect he was, but when you contextualize that score with the scores of his international peers, it’ll highlight just how impressive that score is. Sengun’s 81.0 is the highest Correlation Score in the entire sample for non-NCAA bigs and is almost 5 points higher than the second highest score by Jusuf Nurkic.

Projecting Forward

Peak Ranking Projections

Now that we understand the relationship between Correlation Scores and peak fantasy value, we can project where a prospect might finish in their best fantasy season solely based on their Correlation Score using the below trendline:

Obviously, any prediction that only uses one metric to predict an outcome is the equivalent of a rough guess, and the chart’s r-squared leaves some to be desired, but I think it puts the scores better into context. To give an example, let’s bring back the 2024 first round, but this time with peak projected rank added into the mix:

The savvier readers out there will point out that some of these players (Kel’el Ware, Bub Carrington, etc.) have already passed their projected peak as rookie, but this projection is just that, a projection. This number serves as the average peak that someone with that Correlation Score could be projected to achieve. My biggest takeaway from the data is just how difficult it is for rookies to truly become top tier fantasy assets. This graph says that it’s unlikely, and maybe even unreasonable, to expect even one career top 100 8-category finish from a prospect who is in the top quartile of all prospects drafted over the last 17 years. Something to keep in mind as rookie draft season approaches, and offers of proven assets flood your inbox in exchange for the fun and exciting draft prospects.

Takeaways for the 2025 Class

Now what you’re all probably here for: the Correlation Scores for the 2025 draft class. Here are the scores for all 30 players taken in the first round in last week's draft, followed by some observations.

2025 Player Notes

- Cooper Flagg’s projected Correlation Score is a perfect 100. There have been 9 other players to have a correlation score of 95 or higher. 8 of the 9 made All-Star teams, 7 out of 9 became All-NBA players, and that group averaged 3.55 top 25 fantasy seasons. Pretty good company.

- Ace Bailey’s projected Correlation Score of 44.8 (38th percentile) is the lowest of any top 5 pick, almost 5 points below the next lowest score, Dragan Bender. Alex Len is the next lowest top 5 pick that played in the NCAA, with a Correlation Score of 53.5 (56th percentile).

- 22 players have been a teenager on draft night with a BPM of 10 or more (a common filter among those on Draft Twitter), Jase Richardson has the lowest projected Correlation Score of that group.

- Johni Broome has the highest Correlation Score of any player drafted outside the first round.

For people interested, the full database of correlation scores are available to look through here