GPS Is a Receipt, Not a Target

I've sat in plenty of meetings where the GPS sheet gets walked through and the question is some version of "is the number going up?" If high-speed running is up on last week, the conditioning must be on track, and if it's down, something needs changing. I don't think that's a silly way to look at it, and to be fair I thought that way myself for a long time.

Where I've landed now is that match GPS is more like a receipt, it's descriptive, not prescriptive. It's a really good record of what the game asked of each player, but it doesn't tell you how well they played, and once you start turning those numbers into targets they get worse at telling you anything at all.

This piece is based on our GPS data from Kintetsu's league matches, 2019 to 2025.

When a number becomes the target

Goodhart's law says that as soon as you start using a number to control something, it stops behaving the way it used to. A better way of putting it might be "when a measure becomes a target, it ceases to be a good measure", and I couldn't agree with that more strongly.

You see this in rugby all the time. If a player knows his high-speed running is being watched, he'll find some, and it's usually jogging back into the line at a pace that looks great on the sheet and does nothing for the team. If a coach is told his session has to hit a number, the drill ends up being designed to produce metres rather than to look like rugby.

Many years ago, when we first really started analysing GPS, I got a little bit excited by metres per minute. We really wanted to see metres per minute high during certain drills, and unfortunately this pushed the running load in training so high that it became more of a fitness exercise than an actual rugby exercise. I remember discussing this with a coach, quite proud of what I'd achieved because the numbers were all going up and the graphs were going in the right direction. He simply said to me, "this is all too helter-skelter, there's no quality."

The number I see used like this more than any other is HMLD, which is the metres a player covers while running fast, accelerating or decelerating. As a record of what a session cost, I think it's genuinely useful, because I want to know whether Thursday was lighter than Tuesday. What bothers me is that it's now common to call a session "match intensity" because it hit a certain HMLD per minute, or to call a week good because the total went up. Sitting underneath that is the idea that more running means better rugby, and when I went looking in our data, I couldn't find it.

Does good rugby come with more running?

The fairest test I could think of was to split possessions by how they ended. Our analyst had tagged how every possession finished in 8 of the matches, and which ones had a line break in 11, so I could compare the running with what actually happened, allowing for how long each possession lasted.

In attack, possessions that ended in a try were run at about the same pace as ones that ended in a kick. When we made a line break, our backs ran about 20 metres per minute more in that possession, which makes sense, because the break is what created the running rather than the other way round.

Defence was the part I found most interesting. When the opposition broke our line, our defenders ran about 20 metres per minute more, because everyone was chasing back. When we forced an error or a turnover, we actually ran less, about 26 metres per minute less for the forwards and 40 for the backs. So our best defensive possessions had less running in them than our worst ones, and if we'd had a defensive HMLD target, the possessions where we got broken would have been the ones that looked best on paper.

The match results tell a similar story. Across Kintetsu's 2024/25 season, how much we ran per minute only had a weak link with the final margin, and when I built a model to explain the margin from everything on the match sheet, not one of the running numbers made it in. The things that did were execution and discipline. The wider league helps explain why. Across 194 Division 1 matches, the losing team made about 185 tackles and the winning team about 161, because if you haven't got the ball you're defending, and if you're defending you're running. Judging a team on its defensive running is a bit like judging a shop on how busy the returns desk is.

I should say I nearly built this whole piece on the wrong number. In my first go at it, attacking sprint distance in the second half lined up with points scored at +0.70, and I wrote that the signal was in attack. When I checked defensive sprint distance against points scored, it came out at +0.69, almost identical. What I'd actually found was how fast and open the game was, showing up on both sides of the ball, so I've taken it out.

The thing that did hold up sits further upstream, in fitness. Across 27 players, the fitter ones on the Bronco kept a higher floor of running late in games, and the fittest third of the squad was running more in the second half than the least fit third was running in the first. I can't prove from one squad that the training is the reason, but if I'm going to chase something, I'd rather chase the fitness than the match number.

What does a match actually ask for?

If match GPS shouldn't be the target, I think its best use is describing the game properly, so the training can be built around the right kind of work. This is what a starter in each position did in a typical league match for us.

The totals are what people usually quote, but for training I think two other things matter more: how the running is spread through the match, and how hard it is while the ball is actually in play.

The ball was only in play for about 34 of the 80 minutes, and it came in fairly short bursts. The middle passage of play lasted about 30 seconds, about four in five were over inside a minute, and the longest passage in a typical match was a bit over two minutes.

That's also why the ball-in-play column in the table is worth a look. Every position ran roughly a third harder while the ball was live than its whole-match number suggests, so a prop's 70 metres per minute is really about 92 when the game is actually on. Long passages weren't run much slower than short ones either, so what makes a long passage hard is that the pace doesn't drop off, rather than it getting any faster.

What the metres miss near the try line

Martin Buchheit has written recently about what he calls "GPS 3.0" in football, and one of his points is that speed-zone numbers miss a lot of the real work because they ignore changes of direction. In rugby I think the obvious gap is contact, and it's at its biggest near the try line.

The number I'd look at for this is acceleration density index. Catapult works it out as how much a player accelerates and decelerates for every 10 metres they cover, so it gives you a feel for whether the running is long and smooth or short and stop-start.

When our forwards were defending their own line, they covered about a third fewer metres per minute than when they were pressing high up the field, but they were accelerating and decelerating almost as much, so their acceleration density index went from about 3.1 to 4.6. It was higher on the goal line in all 11 matches that had field position tagged. Attacking inside the opposition 22 looked much the same, with nearly a quarter fewer metres and a higher index in 10 of the 11, and the backs showed the same pattern.

So the work doesn't go away near the try line, it just gets shorter and more stop-start, and if you only looked at metres you'd think goal-line defence was a fairly light few minutes. One thing to be clear on is that acceleration density still comes from the GPS, so it's picking up the accelerations and decelerations rather than the collisions themselves. This works on any Catapult setup where possessions are tagged, and in a game where more of the contest happens close to the line, it's the first thing I'd want to check.

If you know a position's worst case, should you train to it?

This is a question I keep coming back to. Say you know the worst case for a position is some number, X. Wouldn't you want your players to have done X in training before they meet it in a match?

A few years after that helter-skelter conversation, worst-case scenarios became really popular in rugby, the idea being that you find the hardest stretch of a match and train for it. For us, realistically, the worst case was about three minutes of nonstop play, so for a while I thought we had to be training for three minutes all the time. But why would we train for something like that when it only makes up about 3% of the actual play? Wouldn't we want to train for the other 97% and get better at that? That's what made me rethink it and dig a lot more into the statistics of the game.

The first thing the data showed me is that X isn't really one number. Each position has a fairly clear typical worst case, with a back-rower's hardest three minutes around 115 metres per minute and a scrum-half's around 127. But across the 47 players with five or more games, the same player's worst three minutes usually moved by about 9% from one game to the next. Novak and colleagues found the same thing in football in 2021.

The second thing is how rarely the true worst case turns up. If you go by the clock, passages over a minute made up about 40% of the time the ball was in play, so they definitely matter, but passages over two minutes made up only 8% of it, and anything around three minutes was more like 1 to 3%. The longest single passage I have on record is just over four minutes, and that only happened once. Most of a match is short, hard efforts with short breaks in between.

So I don't want to spend 90% of the training on something that happens 10% of the time or less. I'd want most of the conditioning to look like the common passage, so short, intense and repeated, with contact in it for the forwards. The long worst-case passage still gets done now and then, so it's never new to anyone, but it doesn't get to run the programme.

The one exception for me is top speed. Hamstrings need regular exposure to near-maximum speed, so there I would deliberately hit match numbers or higher most weeks. That's about getting the tissue ready rather than copying the game.

So how do you prepare players without chasing the number?

I do think we have to condition players to the demands of the game, whatever they turn out to be, because the match can't be the first time they meet them. For me the difference is what the number is being used for. If it helps you design the work, and you check afterwards that the session landed roughly where you meant it to, it's a guide. If players are judged on it, or get extra running because they missed it, it's become a target. A simple test is who gets to see the number and what happens when it's missed.

It's rugby first and quality first for me, not GPS first. A short, sharp, accurate session can be exactly the right session whatever the GPS says afterwards, and I'd much rather have that than a session that hit its metres by being sloppy.

So the demands shape the work rather than grade it. Most efforts should be somewhere around 20 to 50 seconds, which is where half of all passages fell. They should be at the pace the game is played when the ball is live, which for us was about 90 to 100 metres per minute for the forwards and 103 to 113 for the backs, with a bit more rest than work. The accelerations, decelerations and contact need to be in there too, especially for the forwards. The targets themselves go on things we can actually control and test, like the Bronco, top speed and strength, because the only way to game a Bronco is to get fitter.

Top-ups are where I'd lean on the numbers most, and the point is that bench players never drift below their own normal in the first place, rather than hitting a total. A player who's sat on the bench for three weeks still needs to be ready to play at any point, so we keep him topped up as we go, and a player who's just played the full 80 probably doesn't need anything. Buchheit makes the same point in football, that how much topping up a player needs depends on the minutes he's played and the turnaround to the next game, and that the top-up should look like the game rather than just adding metres to fill in the dashboard.

The extra conditioning in a top-up needs to be as close to the game as possible, which means thinking about change of direction, the metabolic demand and contact conditioning. If a player's HMLD is low and the answer is just to get him running and accelerating in a straight line, it's missing the point.

I had a big forward who was a key player for us. He'd win lineouts, break the line and dominate the contact, and his HMLD was always well below everyone else's, because he was 135 kilos. Does that mean he's unfit, or that he's a poor player? No. So stop judging him on his HMLD, stop trying to top it up to get him level with everyone else, and start conditioning him the way he actually needs for the demands of his job.

Then the check on whether it's working isn't the match GPS going up, it's the fitness markers moving: the Bronco times, and how much running each player can still produce late in games.

What this can still get wrong

This is one club's matches, and the result tests are based on 8 to 16 matches depending on what was tagged, so it tells you what happened for us rather than what's true everywhere. Only nine of the tagged opposition possessions ended in a try, which is why that row in the first figure is so wide.

The ball-in-play windows were tagged by our analyst, and the tagging changed a bit between seasons. The worst-case numbers only include players who stayed on for an hour, so props who came off early aren't in there. And none of these numbers sees contact directly, which for a forward is most of what makes a hard passage hard.

If I had to put one thing on the whiteboard on Monday morning, it would be that match GPS is for learning what the game asks of your players, and the targets should go on the things that make them able to do it.