LIV Golf Predictions: What 7,000 Matchups Reveal — And The Challenges With Comparing Across Tours

While Wyndham Clark was winning the 2026 US Open at Shinnecock Hills, we've been busy analysing LIV Golf data. When Majors come around, it's an interesting time in the Elo space as many tours come together and cross-pollinate. To fully leverage this cross-pollination we've added LIV as a standalone league now on Elo.golf. You can see the LIV rankings here.

From an Elo point of view the LIV data is quite different to the PGA and DP World Tour. LIV has a fixed roster of ~48 players. No cuts. Shotgun starts. Team formats alongside individual competition. So the obvious question is: does Elo still work here?

Quick recap: how we measure predictions

If you've read our previous posts on the PGA Tour and DP World Tour, you know the drill. We group predictions into buckets and check whether the actual win rate matches. The green line is our model; the dashed grey line is perfect calibration. The closer the two lines are, the more calibrated the predictions.

Here's the PGA Tour baseline — 220,000+ matchups:

And the DP World Tour — 365,000+ matchups, slightly tighter calibration:

The LIV Golf result

Here's the calibration chart for LIV Golf:

A few things jump out immediately.

First, the good news: some buckets are well-calibrated. The 43% and 48% buckets are nearly perfect — our model says 47.7% and the actual win rate is 48.0%. That's essentially spot-on. The 57% bucket is extremely close as well.

Second, the not-so-good news: the edges are noisy. The 33% bucket is off by nearly 5 percentage points, and the 67% bucket is off by 4.5pp. But look at the blue bars — those buckets contain just 69 and 86 matchups respectively, compared to 50,000+ in the PGA Tour's middle buckets. With samples that small, statistical noise is always going to be a thing.

Third, notice the chart only runs from ~33% to ~67%. There are no predictions in the 28% or 72% buckets — LIV simply doesn't produce enough extreme matchups to populate those ranges. With only ~48 players, the skill gaps are narrower than you might expect. That plus the fact that LIV pairs the leaders together from round 2 onwards means you get less variance in matchups.

Why is LIV noisier?

There are a few reasons but the main one is: data. LIV Golf launched in June 2022. We have about 4 years of matchups. Compare that to:

The PGA Tour has 32x more matchups than LIV. The DP World Tour has 53x more. More data means tighter calibration — there's no shortcut around that. As LIV plays more events, the edge buckets will fill in and the calibration will tighten, just as it has for the PGA Tour and DP World Tour over the years.

Despite the smaller dataset, the model's core accuracy holds. The Brier score for LIV is 0.2210 — very close to the PGA Tour's 0.2199 and the DP World Tour's 0.2195. The Elo system still works across all three tours but more data will make it better in LIV's case.

Three tours, three completely different shapes

Another interesting factor for those that want to go a bit more in-depth into Elo. Below is the Elo rating distribution for each tour — how spread out the ratings are within each league. All three tours start every new player at 1500, and ratings evolve from there based on results.

The PGA Tour and DP World Tour have nearly identical shapes — similar bell curves centered around 1500, with standard deviations of 72 and 64 respectively. They look like what you'd expect from a well-behaved rating system.

LIV Golf looks completely different. The curve is much wider and flatter, with a standard deviation of 99 — nearly 50% wider than the PGA Tour. At first glance, you might think this means LIV has bigger skill gaps between players. It doesn't. The shape difference is structural, not skill-related, and it comes down to three factors.

1. Pool size

We've talked about this already but this is the biggest factor. Elo is a zero-sum system — when one player gains points, another loses them. In a pool of 2,754 players (PGA) or 6,442 players (DPWT), there's a huge population to redistribute rating points across. An elite player can accumulate a high rating by beating many different opponents.

In a pool of ~48 active players (LIV), there's far less rating "mass" to go around. When Dustin Johnson beats Jon Rahm, the points come from Rahm's rating. But there are only so many opponents in the pool, and the same players face each other repeatedly. This creates a pool-size ceiling effect — the ratings spread out more because wins and losses are concentrated among a small group, not because the skill range is actually wider.

2. A closed field

The PGA Tour and DP World Tour have heaps of new players entering each season through qualifying, sponsor exemptions, and tour school. These new entrants start near 1500 and cluster in the middle of the distribution, creating a peaked center and adding more Elo points for the taking.

LIV's roster rarely changes. The same ~48 players compete week after week, giving the ratings time to fully separate. Without a constant influx of 1500-rated newcomers, the middle of the distribution thins out.

The challenge of comparing across tours

This brings us to the harder question. Every Elo system runs independently per league. Each pool starts at 1500, whether you're a PGA pro or an amateur in our public league. Ratings evolve based on performance within that pool. A "1550 PGA" and a "1550 LIV" and a "1550 DPWT" are three different statements about a player's ability — and comparing them directly isn't a fair comparison.

A LIV player with a 1550 rating has beaten a very different set of opponents than a PGA Tour player with the same number. The rating tells you where a player sits within their tour — but naively comparing the raw numbers across tours misses the context.

This is a well-known challenge in rating systems. Chess faces it across federations, tennis faces it between the ATP and WTA. Whenever separate pools evolve independently, the raw numbers aren't directly comparable without some form of cross-pool calibration.

So what do the ratings tell us?

Within each tour, the ratings work exactly as intended. If a LIV player has a 1580 rating and their opponent has a 1480 rating, our model converts that 100-point gap into a win probability — and as we showed above, those probabilities are well-calibrated against actual outcomes.

The LIV Golf rankings on Elo Golf show you where every player stands relative to their LIV peers. The upcoming matches page shows you real-time predictions for every tee time group. All of that is backed by the same Elo model that powers our PGA Tour and DP World Tour predictions.

What the ratings don't do — yet — is place a LIV player and a PGA Tour player on a single unified scale. That's a harder problem, and it's one we're actively working on. The good news is that we have events like the US Open that provides an opportunity for players from a variety of tours to play against each other. More on this in a future post.

Where to from here?

LIV Golf is the third professional tour in our system, and it validates something important: the Elo model is tour-agnostic. The same algorithm, the same rating update formula, applied to three very different competitive formats — 72-hole stroke play with cuts, 54-hole no-cut events, shotgun starts — and it works on all of them.

With 593,674 total matchups across all three professional tours, Elo Golf now tracks the broadest set of professional golf head-to-head predictions anywhere. And with our Public league growing, every scorecard you submit makes the ratings more accurate for amateur golf too.

Check out the LIV Golf rankings, see upcoming predictions, or submit a scorecard to track your own Elo.

This is the third post in our prediction accuracy series. Read the first: Can You Actually Predict Head-to-Head Golf? We Tested 220,000 Matchups, and the second: PGA Tour vs DP World Tour: How Accurate Are Our Predictions Across Tours?.