PGA Tour vs DP World Tour: How Accurate Are Our Predictions Across Tours?

Last month we published an analysis of 220,000 PGA Tour matchups showing that our Elo model is calibrated to within 1.5 percentage points. The obvious follow-up question was: does it work on other tours?

So we ran the same analysis on the DP World Tour. Not over a small sample but 365,769 head-to-head matchups spanning 25 years and 1,058 tournaments, from the 2000 season through to today. Combined with the PGA Tour data, that's nearly 586,000 matchups tested against the same prediction model.

Quick recap: the PGA Tour baseline

If you haven't read the original post, here's the summary. We group predictions into buckets (28%, 33%, 38%, etc.) and check whether the actual win rate matches. The green line is our model; the dashed grey line is perfect calibration. If they overlap, the model is well-calibrated.

The model is slightly underconfident across the board — when it says 57%, the true win rate is closer to 58%. Average calibration error: ~1.5 percentage points. For the full breakdown, read the original post.

The DP World Tour result

Here's the same calibration chart for the DP World Tour — the model tuned per tour for the best possible accuracy:

The green line hugs the diagonal closely. The 67% bucket is essentially perfect (66.9% predicted vs 67.0% actual), and the 62% bucket is off by just 0.4pp. The blue bars confirm this is all backed by data with 30,000+ predictions in that bucket and over 100,000+ matchups in the 52% bucket alone, more than the PGA Tour's sample sizes.

Average calibration error: ~1.2 percentage points — tighter than the PGA Tour's ~1.5pp.

Why is the DPWT more predictable?

This was the most interesting finding. By every metric, the DP World Tour is slightly easier to predict than the PGA Tour:

• Brier score: 0.2195 (DPWT) vs 0.2241 (PGA Tour) — lower is better
• Predictive skill: 12.2% (DPWT) vs 10.4% (PGA Tour) — higher is better
• Average calibration error: ~1.2pp (DPWT) vs ~1.5pp (PGA Tour) — lower is better

Why? The answer is skill distribution. The PGA Tour is the most elite field in golf. Even the "weakest" player in a PGA Tour field is an extraordinary golfer. The skill gaps between players are relatively narrow, which means more matchups are close to coin flips.

The DP World Tour has a wider talent funnel. Challenge Tour graduates, aging veterans, emerging talents, and world-class Europeans who split time between tours all compete in the same events. This creates larger genuine skill gaps between players in the same field, and larger skill gaps give the Elo model more signal to work with.

One model, two tours

Here's what we find most encouraging: the same Elo model works across tours. We'll be adding more professional tours in the upcoming weeks and months and seeing how Elo predictions perform — because Elo is more than just a rating system. It's also a prediction model.

What this means for you

Whether you follow the PGA Tour, the DP World Tour, or both — our predictions are backed by nearly 586,000 tested matchups across both tours. When you see a win probability on an upcoming tee time, it's not a guess. It's the output of a model that has been validated across decades of professional golf.

With our Public league growing, every scorecard submitted makes the ratings and predictions more accurate. So next time you tee up with your group, submit a scorecard and start tracking your own Elo.

This is a follow-up to Can You Actually Predict Head-to-Head Golf? We Tested 220,000 Matchups. For more on how the Elo system works, see What Is Elo Rating in Golf?