Pardle

The big picture

The setup — describing today's course

Tournament

The tool follows the currently active PGA Tour event. You'll see the tournament name populated automatically once the field roster loads. Once the event finishes, the tool switches to next week's.

Round to forecast

Pick which round you want a forecast for. R4 today, R1 for a Wednesday preview, R3 for a Saturday morning look at the day ahead.

Conditions

Same course, different day — how much has this week's play already revealed about the setup?

• Conditions like the most recent finished round (default for R3+): if yesterday played softer than typical because greens were receptive and pins were fair, the model expects today to play similarly.
• Based on R3 / R2 / R1: anchor on a specific round rather than the most recent one.
• Average of the week: blend every played round. Softer signal, less R3-heavy.
• Typical setup for this course: ignore this week's data entirely and use the historical baseline for this round.

Behind the scenes this becomes a level shift — a per-hole stroke adjustment carried over from the reference round(s). The measured softness of the reference round is spread evenly across today's 18 holes.

Yardage

Pardle's prediction (default): pulls the exact yardages the tour has posted for the round. Each hole plays at the length the setup team has actually set up, not some rounded value.

Manual delta from a prior round: when the tour hasn't posted yardages yet, you can say "the course will play 100 yards longer than R3" and the model spreads that delta evenly across the 18 holes. Useful for early-week previews.

The yardage impact per hole is derived from per-hole regression coefficients — every hole at this course has its own historical yardage sensitivity. Long par-5s scale differently than drivable par-4s.

Pins

Pardle's automated clusters (default): several years of pin-by-pin scoring at this course have been clustered by green zone (front-right, back-left, etc.). Each hole's pin position today is matched to its nearest historical cluster, and the model uses that cluster's residual scoring difficulty.

Manual scoring adjustment: skip the auto-match; enter a single total-round stroke adjustment (+0.5 = the setup plays 0.5 strokes harder than the historical average). Useful when the setup team has done something novel that the historical clusters can't capture — new pins outside any known zone, or an unusually fast green speed the sample doesn't reflect.

Wind

The forecast defaults to HRRR — a high-resolution NOAA weather model that updates hourly and provides wind speed + direction at the course location for every hour of the day. This lets the model give late tee times the wind they'll actually face at 3pm rather than a smoothed daily average.

GFS blend: a global forecast model, smoother in the mid-afternoon peaks but longer forecast range. Useful for Sunday when we're looking at Thursday.

Manual override: force a specific wind speed (mph) and direction (degrees FROM). Useful for what-if analysis or when you disagree with the forecast.

Wind translates to scoring via per-hole headwind coefficients. Each hole's historical sensitivity to headwind is baked into the regression — a downhill par-3 into the wind plays much harder than an easy short par-4 with the wind at your back.

Adding a player

Season SG

Strokes Gained (SG) is the standard measure of golf skill. It counts how many strokes better or worse than the tour average a player is, per round, after adjusting for the fields they've played.

• Roughly +3 SG: elite, top-of-the-world level
• Roughly +1.5 SG: top-50-in-the-world level
• Roughly +0.5 SG: solid tour regular
• Roughly −0.5 SG: below-average tour player, cut risk most weeks

The tool auto-fills this from Pardle's pre-tournament model when the player is in this week's field. You can override with your own number.

Tee time

When set, the model reads the wind at the specific hour this player will face each hole. It assumes ~15 minutes of walking per hole from tee-off, so a 1:30 PM starter is playing hole 9 around 3:30 PM, hole 18 around 5:00 PM.

This matters most when a late tee time faces a building afternoon wind. On a calm morning that gusts up by mid-afternoon, the day-average forecast blends the two regimes and understates the difficulty for the late group.

Rounds this week (auto-filled)

Every player who's completed at least one round shows up with per-round tiles: their vs-par score plus the four strokes-gained categories (OTT, APP, ARG, PUTT) and a persistence factor.

The form signal isn't the raw vs-par score. It's how much a player over- or under-performed their own expected score for that round — the round's actual field average adjusted for the player's season SG. An elite player is expected further under par than the field mean; an average tour player is expected right at it. An average tournament round for an elite player is a negative form signal because he under-performed his baseline. A tour- average player shooting the same round is performing exactly to expectation and gets no form bump either direction.

The persistence factor is the model's smartest trick. Not all strokes-gained categories persist equally from round to round. Once we know the over/under-performance, we scale it by which skills drove it — approach and driving carry forward reliably, putting mostly regresses to the mean.

The persistence factor shown on each round tile compares that round's skill mix to a category-balanced round. A balanced round sits at 1×; approach- or driving-heavy rounds tilt above 1× and count for more; putt-heavy rounds tilt below 1× and count for less.

Advanced knobs

Form weight

Slide left to ignore form entirely and project purely from season SG. Slide right to lean heavily on this week's scoring.

The default sits at Pardle's calibrated value — informed by the Connolly-Rendleman shrinkage literature and our own back-testing. Enough to catch a genuine hot streak, not so much that a few outlier rounds dominate.

The mechanics: the model measures how much a player has over- or under-performed their expectation across their played rounds this week, then applies a fraction of that average delta to the projection. Small fraction, real signal.

Skill compression

Some courses reward a specific set of skills less than SG:Total captures — a "bunching" course where everyone scores in a tight band regardless of talent.

Slide right for no compression — the player's season edge translates 1:1 into the projection. Slide left for aggressive compression at bunching-friendly venues, where an elite player's edge shrinks meaningfully.

The default is set per-course-type based on eight-plus years of scoring dispersion at each venue. At TPC Twin Cities the field bunches, so the default compresses the raw SG number by a notable but not-extreme amount.

Skew adjustment

Golf rounds don't follow a symmetric distribution. The occasional blow-up (a triple bogey, a lost ball) sits in a fat right tail and inflates the mean while leaving the median stable. The skew adjustment is how much wider we assume the mean is than the median.

• Elite players: tightest gap — blow-ups are rare, so mean and median sit close together.
• Mid-tier players: moderate gap.
• Below-average players: widest gap — wider right tail, more blow-ups.

The default gap is picked automatically from the player's season SG tier. You can override with the slider if you have a reason to.

Why this matters for betting: the median is your typical outcome. The mean is inflated by rare catastrophes that don't happen most rounds. When betting UNDER a line, the median is often the right number to compare against.

Reading the results

Field forecast (hero)

The projected total-strokes score for the average player in the field, teeing off with today's setup. This is the top-line answer — the vs-par number below tells you whether it's an easy day (green, under par) or a grinder (red, over par).

Model delta

How much easier or harder the model expects this round to play compared to the historical average for this specific round number at this course.

−0.5 means today should play half a stroke easier than a typical R4 here. +1.2 means it's shaping up as one of the harder R4s in the historical record.

Historical mean

The historical average total-strokes score for this specific round number at this course, with no adjustment for today's conditions. Useful as an anchor: the field forecast should feel roughly like historical mean + level shift + wind delta.

Wind

The wind figure shown is aggregated across every field member's actual tee time — so an 8am-to-3pm tee-time spread with rising wind produces a higher effective wind than the average of the morning and afternoon peaks alone.

Level shift

The per-hole stroke adjustment carried in from the reference round(s) selected in Conditions. Negative = course expected to play easier than historical because this week has been playing soft.

When you see an attenuated value next to the level shift, the Conditions preset has scaled the raw measurement down. That's a hedge against overnight drying or firming: when we're not fully confident yesterday's softness will persist unchanged, the model carries a fraction of it forward instead of the whole thing.

Player projections

For each player added:

• SG / Edge / Form: the three drivers. SG is season baseline; Edge is that baseline compressed for course type; Form is the Bayesian shrinkage bump from this week's persistence-weighted rounds.
• Mean: the average score you should expect if this player played the same round 1,000 times. Use for outright bets where blow-ups still count as losses.
• Median: the middle outcome. Use for round-score UNDER bets, where the fat right tail doesn't hurt you as much as it inflates the mean.

Both are colour-coded: emerald for under par, tang for over par. The Median box is accented emerald because it's the payoff number for most betting decisions.

Per-hole projection

Collapsed by default as an 18-block colour ribbon — emerald for under-par, tang for over-par, saturation scaled to how far from par. Click Expand to see each hole's projected score, par, yardage, and headwind.

This is where you spot the model's per-hole story: a drivable par-4 today (H16 at 301 yds instead of 411) shows up as a deep emerald tile with a big negative vs-par projection. A par-3 into a 15 mph headwind shows up as tang.