Which Stats Really Matter? How the Golden State Warriors Used Correlations to Set an NBA Record.

Estimated Read Time: 5 minutes

In sports today, it’s tempting to jump straight into machine learning models and flashy dashboards. But here’s the thing: before we try to predict anything, we need to know what’s worth paying attention to. So which stats actually matter?

One of the simplest, and most powerful tools to answer that, is correlation. Correlation is not a complicated method (although the maths might look it), it’s just a way to see how closely two things move together. A score close to +1 means they rise and fall together. A score close to -1 means they move in opposite directions. And a score near 0? No real link at all.

Think of it like a pair of dancers, and we’re scoring them based on how in sync they are. If they move perfectly together, in unison, we’ll give them a score of +1. If they mirror each other perfectly, so when one goes left the other goes right, they get a score of -1. If they both decide to have impromptu solos, that’s a score of 0.

Behind the scenes, the maths looks something like this:

1. Looks at how far each value is from its average (called “deviation”).

2. Multiplies those deviations together for each pair of values.

3. Adds them all up and divides by a scaling number to keep it fair.

Not too bad, right? 

But why does this matter for sports? Because it tells us which numbers are meaningful, and which are just noise. Today, we’ll focus on basketball. For example:

• Do explosive jumps actually lead to next-day soreness?

• Is RPE (Rate of Perceived Exertion) really a good workload gauge?

• Does better sleep actually show up in performance?

These are the types of questions correlation helps answer, without overcomplicating things.

A Week in the Life: Example Data

Let’s take a look at a stripped-down dataset from an imaginary basketball team. It mixes workload (like distance, jumps, heart rate) with recovery and wellbeing (sleep, soreness, hydration, mood).

Day	Total Distance (m)	Sprint Distance (m)	RPE	Soreness	Sleep Hours	Heart Rate Avg (bpm)	Jump Count	Hydration Level	Mood Score
Mon	6200	1100	7	3	6.5	155	120	3	4
Tue	5400	900	6	2	7.0	150	110	4	4
Wed	—	—	1	1	8.2	60	10	5	5
Thu	7000	1300	8	4	5.8	160	130	2	3
Fri	4600	800	5	2	7.5	145	100	4	4
Sat	8500	—	9	5	5.0	170	140	2	2
Sun	—	—	3	1	8.1	65	15	5

Now, we’re not going to do a full code breakdown in this blog, but if you do want to know more about the what’s under the hood, take a look at our public code repository, which has a line by line breakdown of the code and methods used. The repository can be found here: Statsmart code repository. 

Once we’ve ingested the data, we can carry out a correlation analysis on the numeric columns. This will tell us how much each attribute moves with another. Results from a correlation analysis can be presented in a number of ways, but we here at Statsmart enjoy a heat-map for this kind of activity, so take a look at our lovely graphic below (if we do say so ourselves) presenting the results of the analysis.

What Do the Numbers Tell Us?

It’s key to remember here what we mentioned earlier in the blog about understanding the results: A score close to +1 means they rise and fall together. A score close to -1 means they move in opposite directions. And a score near 0? No real link at all. So what are some of the key results we can see?

• High-Speed Distance & Soreness (0.98): Sprint-heavy sessions are almost perfectly aligned with next-day soreness. This isn’t just a hunch, it’s a near-lockstep relationship. If you’re planning high-speed drills, recovery better be part of the plan.
  

• RPE & Total Distance (1.00): A perfect correlation. The more distance players cover, the harder they say it feels. This is a strong validation that RPE is being reported consistently and reflects actual workload.
  

• Sleep & Everything (around -1.00): This is the big one. Sleep hours had a near-perfect negative correlation with soreness (-0.96), RPE (-1.00), heart rate (-1.00), and jump count (-1.00). Translation? More sleep = less fatigue, lower heart strain, and reduced explosive load. It’s the clearest signal in the dataset, sleep is the recovery lever. After all, sleep is the miracle performance enhancer that many ignore!
  

• Hydration & Soreness (-1.00): Another perfect negative correlation. Better hydration is strongly linked to reduced soreness. This might seem obvious, but the strength of the relationship suggests hydration tracking should be taken seriously, not just as a wellness checkbox.
  

• Hydration & Mood (0.87): Players who are well-hydrated also report better moods. This is a great example of a “soft” metric (mood) being backed by hard data. Want better energy and engagement? Start with the water bottle.

But why do these results matter and what can we do with them? Let’s take a look at the Golden State Warriors, and NBA team who during the 2015/16 season recorded an NBA regular season win-loss record of 73-9 (that’s an 89% win rate!), and one of the keys to their success was a correlation which most would have ignored.

From the NBA: Why Sleep Was Golden for the Warriors

During their record-breaking season, the Golden State Warriors weren’t just innovating on the court  they were transforming how elite teams think about recovery. One of their most impactful changes? Prioritising sleep.

The team began noticing a pattern: sluggish third quarters, marked by missed shots, slower defensive rotations, and mental lapses. Analysts correlated these dips with biometric and wellness data, especially poor sleep and elevated heart rate variability (HRV). That’s when they brought in Dr. Cheri Mah, a Stanford sleep scientist known for her work with elite athletes.

Dr. Mah helped the Warriors:

• Adjust travel schedules to reduce circadian disruption

• Implement individualised sleep tracking

• Modify warm-up and recovery protocols based on sleep and HRV data
  

The most compelling case came from Andre Iguodala, who began tracking his sleep consistently. The results were dramatic. Igoudala himself said that he would typically get less than 6 hours of sleep per night due to his night-time regime, but when he began to get at least 8 hours of shut-eye, his points per minute increased by 29%, his free throw percentage up by 9%, his turnovers down by 37%, and his fouls also down 45%. These figures were shared by Iguodala himself during a presentation at the Aspen Ideas Festival and as he put it, “Sleep good, feel good, play good.”

This wasn’t just about feeling rested though, it was about measurable performance gains. Iguodala’s improved sleep correlated with better shooting percentages, fewer mistakes, and more consistent play. His teammates took notice, and sleep became a core part of the Warriors’ performance culture. At this point, we'd also recommend diving into the work of Dr. Cheri Mah a bit further if your interest is piqued, there's some really interesting studies!

Now we're not saying and extra 2 hours of sleep made the Warriors the best team in the NBA, far from it, but in a high-performance environment, this correlation certainly hinted towards that marginal gain that they needed to be at the pinnacle of their powers.

What This Means for You

• Coaches: Use correlation to sanity-check what you think is happening in training.

• Analysts: Validate KPIs before building predictive models. Garbage in = garbage out.

• Athletes: Sleep and hydration don’t just “feel” important, they are important.

Correlation doesn’t prove cause and effect, but it helps you ask sharper questions. And in sport, the right question can be the difference between tired legs and fresh ones come game time.

So next time you’re looking to make some marginal gains or understand why a certain trend is happening, turn to correlations, it might just be your new best friend!

That’s blog one of our series: starting simple with correlation. Next time, we’ll dive into something a bit more spicy!

📲 Follow on @Statsmart_