Predicting pacing in the hills

In the microcosm that is our own private goal-setting, the question ‘what time will I run for this trail race?’ emerges regularly. I have used a particular way of estimating performances for some time and I will share it here.

It’s harder to estimate what times you can shoot for when you run on trails, hills, mountains, and fells. Some purists will, not without justification, say that it’s also unnecessary as you are racing the other competitors primarily rather than the clock and ever-changing conditions on the mountains and trail-surfaces make direct time comparisons from run to run and race to race nearly meaningless.

I see the truth somewhere in the middle: changing underfoot and weather conditions does make it difficult to compare ‘like for like’ what a time on a course will be one day versus another day, one year compared to the previous year. Still we keep records for most classical off-road routes and many runners seriously target them. Kenny Stuart’s magnificent records on Snowdon, Ben Nevis, and Skiddaw, for instance, are still prized, if elusive, possessions. While conditions on a given day can make a race a ‘slow day’, this matters much less when we look at results over 20 years because these 20 years will reflect both several ‘good’ and ‘bad’ days.



I begin by running over the course in training at a known controlled effort trying to keep it even throughout the run. If I cannot do this I look at historical gpx files for the run completed by myself or by other runners. For races like Snowdon this was particularly easy as the year’s progressed, as I had access to several recordings. This allowed me to pick splits consistent with my own strengths and weaknesses. The Snowdon race is a classical ‘up/down’ route. A weak climber will spend a greater proportion of the race on the uphill than the downhill and vice-versa for a strong climber but weak descender.  When running the route in advance – such as I could do for a local race like Leinster’s highest peak Lug na Coille – I could use the existing splits as they were as long as ‘ran true to myself’ during the training run.

Once I have secured a reasonable baseline run recording, I will copy and paste the splits into an Excel sheet. I then insert my own subjective estimated (or desired) target time next to the time of the baseline run. In the example below a baseline run of 67:37 is compared to a target time of 55:00. The difference between the two is calculated (in this case a difference of 18.7% is desired).

I then apply this difference to each individual lap split – that means I reduce the time for each split by 18.7%. The result can be shown below:


Using the information

How do you read this? So if you look at ‘lap 1’ which is the split that represents the first kilometre (uphill as you can see from the slow time), the training run time was 8:44 min/km. In order to, run 55 minutes with the same distribution of effort as in the training run, this needs to be reduced to 7 minutes 6 seconds. The second kilometre needs to be reduced from 8:31 to 6:56 and so on.

This allows me or my runner to have a rough goal in mind – we’d know in this case to try and get to the 2 km mark in just around 14 minutes. Any slower than that and we need to push the pace more. If we’re too far ahead we may be in danger of blowing up OR we may be able to hold back a bit more on the descent (by building a ‘cushion’ on the uphill).

I can also mark out ‘hot spots’ or splits that look too difficult to achieve.


A potential hot spot may be the 3:21 min/km on kilometre 6. From the recording I know that this is a downhill kilometre with a drop of 73 metres so 3:21 min/km is likely not unrealistic at full tilt. But it’s worth highlighting as runners often run downhill quite fast in training and then it becomes difficult to run them MUCH faster in the race situation. More time then needs to be clawed back on the uphill.


Breaking records

Let’s take a more exciting example: the Scarr mountain race (ok, maybe not exciting to me but I live on it). There’s a relatively new record on this course set since the course was amended from a straight up and down to a ‘looped’ course. It’s held by Des Kennedy in 36:43 and set in 2018 (remarkably as conditions were notoriously windy – showing you can run good times on ‘slow days’). Des Kennedy also holds the second-fastest run in the modern era (some old results have been lost to the mists of time) for the traditional course and thus we know the current record is from good pedigree.

By inserting the recording of the record run into an Excel sheet, we can calculate what it would take to run a record of say – 35 minutes:


A potential challenger could use this to gauge where to insert their main charge. They can see they need to go below 22 minutes at the top (which Des reached in 22:51) in order to break the record and descend in 13 minutes 15 seconds. No kilometre looks truly unachievable for someone talented and fit enough to even consider attacking the record but putting them together is obviously the challenge.


How to use the information

Personally, I used these markers to establish rough segment targets within the greater run to help me along. On Snowdon I’d know when I should reach the climb halfway point to be on track for a sub-1-hour ascent, for instance.

If we have access to the route we are trying to ‘attack’ with this level of detailed preparation, we can go out and run the key sections in training to see how far we are from the level required. Looking at Des Kennedy’s example from above we could try to ascend in a sub-22 minute time, for instance, or simply try an ‘all out climb’ and see how far we fall short. As the workout is relatively short (4 km up / 22ish minutes), it would not stress our bodies unduly as long as general conditioning has been done before.

When working with Jason Kehoe, we often used this approach even when we didn’t have daily access to the mountains he was attacking. We would know pretty much how long it would take him to be at the top of the various summits to be ‘competitive’ and we’d be familiar with the average gradient. So, if we knew he had to climb for 48 minutes up a 5 km slope of an 18% incline, we could design a simple progression of workouts:

  • 8x 6 minutes uphill rep (15-20% incline), 90 sec recovery
  • 6x 8 minutes uphill rep (15-20% incline), 90 sec recovery
  • 4x 12 minutes uphill rep (15-20% incline), 90 sec recovery
  • 45-50 minute uphill time trial (15-20% incline) at ½ to ¾ effort
  • 2x 24 minutes uphill tempo (15-20% incline), 90 sec recovery

This particular progression is just an example but illustrates the basic logic. The general fitness needed to do the above had to be completed before this type of training began. Most runners do not engage in this type of very specific preparation because mountain running is either a ‘side-show’, ‘2nd priority’ or simply ‘part of the conditioning for road and cross-country’ – so it’s only an approach I would suggest to people for whom the mountains is the be-all end-all.

What could I do?

For ‘mere mortals’ it may interesting simply to see what running part of a course at ‘record pace’ feels like. I remember once going out on Scarr to run a section at the estimated record pace. It was a humbling experience but gave an insight into the physical requirements.

Scientifically, we can go much further with this level of analysis by using correction for terrain and climb based on mathematical equations as well as input from the new Power meters for running (such as the Stryd to which I am affiliated). Power meters allow us to calculate what power output we currently can create for a certain duration (uphill, downhill or flat) and then estimate based on that what is required to generate the desired pace on uneven courses. I may dedicate a future post to exploring this if I sense an interest from my readers.