Arccos Caddie: does AI actually improve your game? An honest review after two years of data

I started using Arccos in April 2021. At that point my handicap was 11.2. After two full years, 180 tracked rounds, and approximately 2,400 shots logged, it sits at 8.8. I want to be honest about how much of that the system caused and how much was inevitable.

Arccos works like this: small sensors (£199 for a set of 14) screw into the grip end of each club. Your phone, carried in a pocket or on a clip, detects which club you've swung via Bluetooth and logs the GPS coordinates of each shot automatically. After the round, the app reconstructs your scorecard, shows you a shot plot of every hole, and builds a statistical profile of your game over time.

The data after two years is, genuinely, revelatory. Some things I now know about my game that I did not know before: I gain approximately 0.8 strokes per round from the tee. I lose approximately 1.2 strokes per round from 50–100 yards (my short game is worse than I thought). My 7-iron is reliable. My 4-iron is not (I have hit 14 greens from 170–185 yards in 180 rounds, a 19% GIR rate that would embarrass a 20-handicapper). I play significantly worse in wind, which is obvious but the data quantifies it: my scoring average in wind conditions above 15mph is 5.1 shots higher than in calm conditions.

The AI Caddie feature, which gives you club recommendations based on your personal distance data and the hazards on the hole, is genuinely useful but requires surrender of ego. It will tell you to hit 6-iron off the tee on a 350-yard par-4 if your driver is wayward enough that the expected score from a fairway 6-iron lie is better than the expected score from your driver. The first time it recommended a 6-iron off the tee at my home course I ignored it. I made bogey. It had said bogey was the expected result from my driver there. The data was right. I was wrong.

The weaknesses: the automatic shot detection, while impressively accurate, occasionally logs a practice swing or misses a shot (typically a short chip where the swing is too small for the sensor to reliably detect). Over time this introduces small inaccuracies into the stats — perhaps 3–5% error across a season. For a professional's analysis, this would be unacceptable. For an amateur trying to understand patterns, it is perfectly adequate.

Subscription cost: £99 per year after the first year (included with sensor purchase). This feels like the right price for what you get — it is approximately what you'd spend on a single round at a premium course.

My honest assessment: Arccos improved my game not primarily by giving me club recommendations, but by forcing me to confront data about my game that ego had previously obscured. I genuinely believed my short game was my strength. The data said otherwise. Addressing the short game is why the handicap dropped. Arccos didn't hit the shots. But it showed me which shots to work on.

Who should buy it: any golfer between 5 and 20 handicap who wants to understand their game properly. If you're below 5, the margins the system provides are probably insufficient for what you need. If you're above 20, the fundamentals of ball striking are likely more useful than statistical analysis.