Tanda Marathon Predictor - Finish Time from Training

Problem: a fast 10K promises a marathon your mileage cannot deliver

A 10K in 50:00 converts to a marathon of roughly 3:50 with the usual race-to-race formulas, and those formulas assume you have trained like a marathoner. Thirty miles a week does not get most people there. The last six miles, not the 10K speed, decide the finish time, and nothing in a short race result measures them.

Giovanni Tanda of the University of Genova looked at the problem from the other side. Instead of a race, he used what runners actually did in training: the average distance per week and the average pace of every run over the eight weeks before a marathon. Across 46 marathons run by 22 amateurs, those two numbers predicted the finish time with a standard error of about 4 minutes.

Solution: predict the marathon from the training itself

No race needed. The two inputs come straight from a running app: total distance and total time for the last eight weeks.

Volume counts. A runner at 35 miles a week and one at 45 with the same average pace get different predictions, because the weekly distance is one of the formula's two inputs.

It shows the trade-off. The result tells you what 10 more miles a week or a 10 s faster average pace would be worth on race day.

Filling it in from your watch history

How you enter training - "8-week totals" takes the total distance and time straight from Strava, Garmin Connect or a similar app; "weekly averages" suits a spreadsheet or a training log.
Miles or kilometers - the unit for both the distance and the pace. The prediction is the same either way.
Average weekly distance or total distance - every run of the eight weeks that end one week before the race, warm-ups, cool-downs and recovery jogs included. Cycling and swimming stay out.
Average training pace or total running time - the pace of all those runs together, not of your workouts. For totals type hours and minutes, for example 41 h 30 min.
Read the prediction - the finish time, the race pace per mile and per km, the band that held about two thirds of the study's races, and a table of the same pace at other weekly distances.

Weekly miles against average training pace, side by side

Find your row and column to see the marathon the formula expects. Moving one column left (a pace 1 minute per mile faster) is worth about 14-15 minutes; moving from 30 to 40 miles a week is worth about 6.

Per week	8:00 per mile	9:00	10:00	11:00
20 miles	3:30:24	3:44:49	3:59:15	4:13:40
25 miles	3:26:56	3:41:21	3:55:47	4:10:12
30 miles	3:23:37	3:38:02	3:52:27	4:06:53
35 miles	3:20:26	3:34:51	3:49:16	4:03:42
40 miles	3:17:23	3:31:48	3:46:13	4:00:39
50 miles	3:11:40	3:26:05	3:40:30	3:54:55

The study's runners averaged 25.1-68.8 miles a week with training paces of 6:48-8:52 per mile and finished in 2:47-3:36. The 20-mile row, the 10:00 and 11:00 columns and every time over 3:36 lie outside that data, so treat them as rough.

Six training logs and the times they point to

35 miles a week at 8:50 per mile. A typical first-marathon block. Prediction 3:32:27, race pace 8:06 per mile, band 3:28:23 to 3:36:30. Race pace is faster than the training average because the average holds every easy mile.

25 miles a week at 9:30. The lower edge of the study. Prediction 3:48:34 at 8:43 per mile, already past the 3:36 the data reached, so the calculator flags it.

50 miles a week at 7:45. A strong club runner. Prediction 3:08:03, pace 7:10 per mile; ten more miles a week would still be worth about 5 minutes.

40 miles a week at 9:00. Mostly easy running with one long run. Prediction 3:31:48, almost the same as the first log: more miles at a slower average land in the same place.

30 miles a week at 10:00. Prediction 3:52:27 at 8:52 per mile. Both inputs sit outside the study, and a 4-hour goal is the more honest target here.

72 km a week at 5:08 per km. The same tool in kilometers: prediction 3:18:23, pace 4:42 per km.

What one change in training buys

Start from 35 miles a week at an 8:50 average. Each row changes one thing and leaves the rest alone.

Change	Prediction	Gain
Starting point	3:32:27	-
5 more miles a week	3:29:24	3:03
10 more miles a week	3:26:29	5:58
15 more miles a week	3:23:41	8:46
Average pace 10 s faster	3:30:03	2:24
Average pace 20 s faster	3:27:39	4:48
10 more miles and 10 s faster	3:24:04	8:23

Each extra mile a week is worth less than the one before it, because the weekly distance enters the formula through an exponential that flattens out. The pace term does not flatten: every 10 s per mile of average pace is worth the same 2:24. That is the formula's view, not a training plan; pushing the average pace up usually means fewer easy miles, and both inputs move together in real training.

Where the formula stops working

The equation came from a specific group: 21 men and 1 woman aged 28 to 54, racing flat marathons at an even pace, each with one or more long runs of 30-35 km (19-22 miles) three to four weeks before race day. Their training was logged on tracks or flat routes with GPS. A hilly course, a hot day, a race run with a big positive split or a block without long runs all push the real time above the prediction. For runners with a BMI under 23 the paper reports a tighter standard error, about 3:35 instead of 4 minutes.

The eight weeks end one week before the race. Taper weeks are left out on purpose; including them lowers the weekly distance and makes the prediction slower than it should be.

Runners usually ask

Why is my predicted race pace faster than my training pace?

Because the training average includes warm-ups, cool-downs, recovery jogs and every easy run. At 35 miles a week and 8:50 per mile the formula expects 8:06 per mile on race day, a gap of 44 s, which matches how marathoners actually train.

How accurate is the Tanda prediction?

In the original data 69% of finish times landed within 4 minutes of the prediction and 50% within 2 minutes. About one runner in three missed by more than 4 minutes.

How many miles a week do I need for a sub-4 marathon?

It depends on the average pace. At 10:00 per mile, 25 miles a week gives 3:55:47 and 20 gives 3:59:15, both outside the study's data. At 11:00 the formula only dips under four hours from about 42 miles a week. A long run of 19-22 miles matters as much as the weekly total.

Should I use this or a race-based predictor?

Use both. A recent 10K or half marathon in the VDOT calculator shows what your speed allows; this tool shows what your training supports. When the two disagree, the slower number is usually the safer goal.

Do cross-training and walking count?

No. The study counted running only. Walk breaks inside a run are part of that run's pace; a separate walk or a bike ride stays out of both the distance and the time.

I missed two weeks to illness. What now?

Enter the eight weeks as they were, including the gap. The lower average is the honest picture of your preparation, and the prediction will be slower accordingly.

Related tools

VDOT Calculator

Training paces and a race-based fitness number from a recent result - See calculator

Half Marathon Splits Calculator

Mile-by-mile splits for a tune-up half before the marathon - See calculator

Running Pace Calculator

Pace and speed of any run, with race predictions - See calculator

Weather Pace Adjustment Calculator

How much a warm race day slows the predicted pace - See calculator

Training Heart Rate Calculator

Heart rate zones that keep easy miles easy - See calculator