Jim2's Running Page

Predicting A Marathon Time

Home
About Me
Favorite Links
Contact Me

7/22/03 

There have been a couple of recent threads on this subject. I thought I would add a few comments....OK, more than just a few comments :-)....and suggest a couple of methods to make marathon time prediction more realistic. This is going to be really long. So, if you are pressed for time or not particularly interested in the subject, then I suggest that you move on to another post.

I think that predicting a race time is one of the more interesting and challenging cerebral elements of marathoning simply because there are so many variables involved in running a marathon. Some of them are specific to "short term" race preparation, such as the intensity of a current training cycle, and race day conditions. Others are functions of longer term experience and conditioning which can span years of running. Finally, however, actual race day performance is dependent on honest goal setting and a corresponding "smart" race plan/execution. It all starts with a reasonably accurate prediction of the time that you are currently prepared to run.

A precisely accurate race goal and perfect race plan are most desired, of course. However, a race goal and associated plan that are somewhat conservative can still produce a very good race. OTOH, a race goal that is optimistic....even by just a little bit....will probably result in an overly aggressive race plan and a big disappointment in race performance. In the case of a conservative goal/plan, there is still opportunity to make up for some of the conservatism in the latter phase of the race and at least reach a satisfying, if not ideal, result. However, the cost of an overly aggressive race is often slamming hard into the proverbial wall and not coming anywhere close to realizing your potential in the race. It's all dependent on an accurate assessment of your current marathon performance potential....or "prediction" of a realistic marathon time. That's why your "baseline" criteria that is used for predicting a marathon time is so important.

The most popular method of setting a marathon goal is to simply plug the results of a race of a shorter distance, usually 5k-half marathon, into one of the many race calculators or tables that are available today to "predict" a marathon time. However, the accuracy of such a prediction is often suspect because the calculators don't or can't consider several factors that affect race performance. More thought and analysis is often needed for the marathon time prediction to be reasonably accurate. Let's look at some of the factors involved.

Choosing the race....I call it the "baseline race"....that you use to predict a marathon time is critical. The timing and length of it directly affect the accuracy of the resultant marathon prediction.

Let's start by looking at the timing of the baseline race. Using a 5k, 10k or half marathon PR time that was run sometime in the past....even just a couple of months ago....might give you an idea of your marathon potential, but it won't necessarily tell you what you are really prepared to run on race day. You might not have trained "adequately" to support your predicted marathon potential, or your marathon training program might have taken you to another level which would make a prediction based on an "old" race unnecessarily conservative. The baseline race should reflect your readiness to run a marathon now, not someday or yesterday. For it to do that, it should be as close to marathon race day as practicable. Run it 3-4 weeks before the marathon, which is when you should be just about peaked from your current marathon training cycle. That isn't to say that races shouldn't be run earlier in, even before, a marathon training cycle. They certainly should! However, they should be used primarily to calibrate training progress and to establish training paces, which should always be based on CURRENT, not goal, race paces.

A minor taper should precede the baseline race. It shouldn't be a long taper of a week or more. After all, 3-4 weeks before marathon day is peak training time, which you certainly don't want to lose. Depending on how long the baseline race will be, a short taper of 2-4 days will leave you fresh enough for the race to be a meaningful indicator of what you can expect on marathon day. And the taper shouldn't be terribly deep. Keeping runs short and easy should be adequate. If you are "tapering" for 4 days to run a longer race, like half marathon, it might be a good idea to run a few strides a couple of days before the race to pep yourself up, then take off the day before the race.

Just as important as the timing of and prep for the baseline race is the length of it. A general rule of thumb that many marathoners follow is that the longer the race, the greater the credibility of the predicted marathon result. For instance, a 10k race is much better than a 5k for projecting a marathon time. (More detail later.) A race of 15k to half marathon might be even better....even optimum, in the opinion of many runners. A metric marathon (26.2k or 16.3 miles) might be best of all simply because it is the longest "standard" race distance that is short of a marathon.

(A sidebar for mid-Atlantic runners.....One limitation with using a metric marathon as a baseline race is that they are rare. However, two Maryland running clubs....Annapolis Striders and Howard County Striders....schedule metric marathons in early October, which is about a month before several major fall marathons, such as MCM, NYC and Chicago. The scheduling of the Annapolis Striders metric marathon on the first weekend of October goes back over 20 years to when MCM, NYC, Chicago and the old Maryland Marathon, all of which were popular with mid-Atlantic runners, were run on the same day....the first Sunday of November....which was 4-5 weeks after the Annapolis Metric Marathon. Both of these metric marathons make for excellent baseline racing. Both are relatively small club races, so there is no crowd to interfere with running "your" race. They are also fairly hilly, which should make them somewhat "conservative" predictors for the relatively flat courses of MCM, NYC and Chicago....which is good. If runners in the mid-Atlantic area aren't taking advantage of them to prepare for fall marathons, they should consider doing so.)

Back to the subject at hand. Why are longer races better predictors of marathon performance? Simply because they more closely replicate the use of the physiological systems/processes on which the body will depend in a marathon. For instance, I would NEVER use a 5k race as a primary predictor for a marathon. Why? Because a 5k race is run at or near aerobic capacity (VO2max) which, next to percent of fast twitch fibers, is, arguably, the least important physiological factor for a marathon. A 5k race, much like a set of Yasso 10x800's, predicts the marathon time that your aerobic capacity will support. However, neither tells you much about your lactic threshold (LT), endurance, running strength, or running efficiency.....all of which have a greater effect on marathon performance than aerobic capacity. A 5k race really isn't a heck of a lot more relevant to a marathon than is a 1-mile race or a 100 yard dash....both of which can theoretically be used in race calculators to "predict" a marathon time. But, who would?

Twenty years ago, 10k was the most common race distance in the US road running community. In most American running communities, runners could find a 10k race at least every month....frequently, every weekend during prime spring and fall training/racing seasons. Longer distance races, such as 15k and half marathon, were often hard to find and 5k's were rare. Thus, most marathoners used a 10k race as a basis for marathon race planning. Today, however, 5k has replaced the 10k as the most popular shorter race distance. The running community has become largely based on the two extremes of 5k's and marathons. 10k's have become much harder to find. Longer races, although more popular today, still remain somewhat the exceptions. It's 5k's that are available most weekends in many communities. As a result, more marathoners are using them as a marathon prediction baseline. I think that is a significant step backwards in the last 20 years. It's much better to go out of your way, if necessary, to find a race of 10k or longer on which to base your marathon plan.

There is a school of thought that says that a 10k race is the best choice of a baseline race. Personally, I like it as a marathon time predictor. In fact, it's the only race distance I have ever used as a baseline race. Why 10k? After all, it's just 3.1 miles longer than a 5k; it's less than half the distance of a half marathon; and it's less than a fourth the distance of a marathon. It is an excellent baseline race for marathon prediction simply because it's essentially a threshold race and LT is one of the most important physiological factors in determining marathon ability....there is a direct correlation between LT and MP. But, what about the other important factors for marathoning....endurance, running strength and running economy? Although a 10k provides a much better test of them than a 5k, aren't they even better tested by a longer race, like half marathon? Perhaps. But a 10k is a better LT challenge than the longer races, and I think there is another way to incorporate endurance, strength and running economy into predicting a marathon time. What is it? Read on.....;-)

OK, now that you have a baseline race under your belt, how do you use the results? The most common method, probably because it's easiest and the most publicized, is to simply plug your baseline race result into a race calculator or table to determine what your theoretical marathon time should be, then base a race goal and plan on that prediction. How accurate is that method? Well, it varies. However, the general trend, as indicated by the responses to TrackMom's post concerning using a 5k to predict a marathon time and Steve in CT's post about half marathon vs. marathon pace, is that the prediction is often optimistic. Other forum threads in the past have reached the same conclusion. Why is that? Are the running gurus who created the calculators and tables so wrong....or stupid? Well......let's talk about that.

A race calculator is nothing more than a mathematical algorithm that multiplies a race time by a fixed factor (ratio) to get a projected time for another race. For instance, most race calculators use a ratio in the range of 4.6-4.7 to predict a marathon time from a 10k time, or vice versa. (The calculators' algorithms result in other fixed ratios for other race distances.) RW calculator's algorithm produces a ratio of 4.6 to relate 10k and marathon distances. Merv's calculator uses 4.67. Glover's tables in his book, "The Competitive Runner's Handbook", use a ratio of about 4.7. The tables in Daniels' book, "Daniels Running Formula", use about 4.6. Thus, I think that it's clear that 4.6-4.7 is the range used by most "experts" for marathon/10k relationship. In fact, in the 80's....before the advent of the internet helped to make the use of running calculators so readily available and popular....a RW article suggested that runners simply use a factor of 4.65 to predict a marathon time from a 10k race. That's the guidance I used throughout my first running life.

OK, so just how "accurate" are race calculators at predicting marathon times for today's marathoners? If they aren't terribly accurate, as so many of this forum's participants indicate, why aren't they? And, most importantly, can we still use them to arrive at a reasonably accurate marathon time prediction?

We all know that, for race calculators to "work", it is necessary to be "adequately trained" for the distance that you are predicting. But, what is "adequately trained"? Although all of the many cookbook training programs available to us agree on certain fundamentals of marathon training, they don't completely agree on many of the specifics. Also, within a family of training programs from a given source, there is a variance of intensity depending on a marathoner's level of experience, such as Glover's categories of beginner, novice, basic, advanced and champion levels. For instance, a runner striving to move from the basic to advanced level might have fully developed aerobic capacity, mostly developed LT, and still be in relatively early stages of running strength and economy development, which take a lot longer than the first two. As an example, with further consistent training, my marathon times continued to improve for 3-4 years after my 10k times had essentially peaked. So, "adequately trained" for the marathon distance can take on different meanings to different runners.....but the calculators don't recognize those differences.

The algorithms used in running calculators don't know which training program you are using or what your marathon objective is. They don't know what you have incorporated into your training program....base training, tempo runs, intervals, Yasso's, hill repeats, high or low mileage, number of 20+ mile LSD's, etc. They don't know if you have incurred an injury that interrupted the continuity of your long term training regimen. They are very impersonal devices that simply say, "Well, if you could run x miles in y time, then, based on a data base that I have of what some others have done, then I assume that you can run 26.2 miles in z time". They have no idea of your running history, how well you have recently prepared for the target race distance, or when and under what conditions you ran your baseline race. One thing they do, however, is assume that you are an advanced, competitive runner who has trained very hard and well to prepare for the target race distance.

Another factor that affects the accuracy of race calculators.....and this opinion of mine might be a bit controversial....the algorithms used in them are based on actual data collected from marathoners more than 25 years ago. (For instance, RW says that their calculator was developed in 1977.) Marathoning, at least in America, has changed since then....and not for the better, as far as competitiveness is concerned. Unfortunately, the change has been in the direction to make race calculators less accurate in predicting marathon times for the "average" runner. If that is the case, then should the algorithms be adjusted to reflect today's more typical marathoner in order to make them more "accurate"? I certainly hope not! That would simply contribute to the downward spiral of the competitiveness of the American marathoner....but I digress by climbing back up on another soapbox.......;-)

As we discussed above, the veracity of the calculator depends on the length and timing of the baseline race. However, you can run a race of 10k-half marathon at the optimum time (3-4 weeks before the marathon) and still get an optimistic marathon time projection from a race calculator. Why? Because a race of half marathon or shorter simply cannot fully replicate all of the variables that go into a marathon performance.....especially the combination of the most significant factors of LT, endurance, strength and running economy. Once you go beyond 15-20 miles in a race environment, you enter a different running world. No race short of that can precisely predict how intense and balanced your training and preparation have been to support a full 26.2 mile race. However, I think there are a couple of prediction methods that will better account for all of these factors than simply plugging a baseline race result into a calculator.

I think the best method is to develop your personal ratio, based on your favorite baseline race distance, for predicting marathon times. (Of course, you could develop multiple ratios for different baseline race distances.) Your personal ratio would reflect the uniqueness of your physiological traits and your specific long term training and racing regimen and intensity. However, arriving at a reasonably accurate personal ratio takes time and experience. It is necessary to complete several marathon cycles to zero in on your personal ratio.

What do you do during the interim time while you are developing your personal ratio? That leads us to an alternate solution, which, I think, can be used by marathoners of any level of experience. It involves the use of a combination of a baseline race and race calculator, plus an "adjustment criteria" to incorporate as many of the physiological factors that significantly affect marathon performance as possible into your prediction.

Firstly, make the baseline race 10k-10miles. That range embraces LT for most runners. Thus, your baseline race will be a good measure of LT, which, as I said previously, is a primary key to marathon performance.

OK. Now that LT has been accounted for, how do we factor in the other significant variables of endurance, strength and running economy? The answer is by making the marathon prediction dependent on training mileage, which is a key training factor that affects all three of the other significant marathon ingredients. But, the race calculators/tables don't do that. So, how do we do it without simply guessing at a fudge factor, as was suggested by a couple of folks in response to TrackMom's post? Well, we can thank a past fellow forumite for the answer to that question.

This subject was discussed extensively on the old RWOL Competitive Forum 6-7 years ago. An early forumite, BrianW, along with several others of us, pondered the same quandary of how can we more accurately predict marathon time from a shorter race.

(Background.....Brian is a 2:40 marathoner and an aeronautical engineer who suggested the creation of the RWOL Competitive Forum in late 1996, which has been recently reincarnated in the form of the High Performance forum. The old Competitive Forum later spun off into Merv as a result of one of the RWOL system format changes. Brian also developed Merv's running calculator that is used by so many of us today.)

It was apparent to those of us discussing the subject in 1996/97 that the missing variable was the intensity of marathon training.....in other words, how "adequately trained" one was for the marathon distance. We postulated that mileage was a significant measure of the "intensity" of a marathoner's training program. In other words, the higher the mileage one was running, the greater the probable overall intensity the program.

Given that 10k racing was the traditional basis for predicting a marathon time and assuming that mileage is a measure of intensity in a marathon program, Brian conducted a survey of marathoners to determine their actual experience in 10k times, marathon times and training mileage. He surveyed RWOL participants and members of the San Diego Track Club. Thus, his survey was limited. However, it did clearly indicate a very direct relationship between mileage and the marathon/10k ratio. He reported the results on the old Competitive Forum in a post on 12/16/97. I saved it for future reference. The results were as follows:

Mileage                Ratio

30-35 mpw            5.5

40 mpw              5.0-5.3

55 mpw                 4.9

60 mpw            4.75-4.85

70 mpw              4.7-4.8

80-100 mpw     4.55-4.65

As you can see, the surveyed runners whose actual ratios were 4.6-4.7, which is the range on which race calculators/tables are based, ran more than 70 miles per week. That is an advanced stage of competitive marathoning and just might be the level of the typical runner on which the original calculator algorithm was based 25 years ago, since most marathoners in those days were of the "hardcore" type. For the most part, the responders to Brian's survey reported 10k and marathon PR's, not the results of 10k baseline races run shortly before specific marathons. Thus, the resultant ratios might be skewed to the high side since it is unlikely that a 10k PR would be run late in a marathon program. (The relationship of 10k to marathon PR's vs. current races is another subject for another thread.) However, the results still indicate that the relationship between marathon and 10k times, regardless of the nature of the data base, is a direct function of training intensity. Even if you adjust within Brian's tables by a category or two to account for using PR times vs. actual baseline races, most runners running less than 60 mpw would still have a ratio higher than that of the race calculators.

Another data point....a few years ago an RW survey of 2000 runners, which should represent a reasonable cross section of America's more current running community, indicated that their average marathon/10k ratio was 5.0. If you assume that the "average" marathoner averages somewhere in the 40-50 mpw range, then RW's survey would be consistent with Brian's table.

How much difference does the specific ratio make? Well, let's calibrate that. For a 45 minute 10k runner, RW's calculator predicts a 3:27:00 marathon, whereas Brian's table agrees with that only if he runs more than 80 miles/week. OTOH, Brian's table predicts a time of 3:40:30 if he runs 55 mpw or 3:45:00-3:58:30 if he runs 40 mpw. RW's average survey result of 5.0 predicts a marathon of 3:45:00. Overall, that is a wide spread of as much as 30 minutes! If Brian's tables and RW's survey have any validity at all, then that's certainly more than sufficient to make the calculator prediction overly aggressive for someone running low to moderate mileage, which could easily result in a race plan that is spelled T-R-O-U-B-L-E! And the scenario would probably be an even worse if 5k had been used as a baseline race!!

There can certainly be exceptions to this pessimistic assessment of the reliability of race calculators. In fact, I am one exception. As I said previously, I have used 10k's exclusively to "predict" a marathon time in the 10 different years in which I have run marathons. I used the published RW factor of 4.65 in the '80s and the Merv race calculator in the late '90s. (I took a "break" from running in 1990-96.) My resultant actual ratio experience is as follows:

Year     Ratio     Mileage

1983     4.83       32.1

1984     4.59       38.0

1985     4.6         46.8

1986     4.76       44.7

1987     4.78       45.6

1988     4.67       47.5

1989     4.57       47.1

1997     4.61       45.0

1998     4.57       40.6

1999     4.73       38.2

My 1984-89 data is based on running the same 10k baseline race (the Columbus Chase 10k) and marathon (MCM), which were 4 weeks apart, each year. Most years, my Columbus Chase 10k was within a minute of 10k PR time....one year (1984) it was a PR. "Mileage" is the average weekly mileage I ran during a 16 week period preceding the marathons, which generally was in the 40-47 mile range. My peak mileage weeks were fairly consistently 60-62 miles with a couple of years a little higher and lower than that.

The variation in my ratio from year to year mostly reflects differing race day weather conditions. My highest ratio of 4.83 in 1983 was because it occurred during my first year of running, less than 8 months after my very first race of any distance. In other words, I was still quite underdeveloped for the marathon distance compared to the shorter 10k distance. Plus, unlike those of you who have lots of information on the internet, I had no guidance for setting a marathon goal and developing a race plan. Thus, I was overly aggressive in my first marathon.....and paid for it in the last few miles! The ratios of 4.76 in 1986 and 4.73 in 1999 reflect hot weather conditions on marathon day, which cost me approximately 15 minutes each year. Otherwise, the ratios those years would have been around 4.6. The ratio of 4.78 in 1987 was the result of a poor training year because of a work related project that required 70-80m hour weeks in Jan-May, while I still tried to maintain a full training and racing program, and resulted in a 10% decrease in annual mileage. All of my race results were down in the second half of that year, including my 10k baseline race simply because I was tired, but my marathon performance that year suffered the most.

The net of this data is that I developed what I consider to be my personal ratio....approximately 4.6-4.67. Thus, the RW and Merv calculators are reasonably accurate for me, even though I haven't reached the mileage that Brian's survey indicates is needed to be in the 4.6-4.7 ratio range. So, I guess that makes me an exception to RW's and Brian's data bases. However, I still think they have value for many people in predicting marathon times.

Why am I an exception? I really don't know. Perhaps it's simply due to my particular physiological makeup. However, I like to think that it's because of my training philosophy. Namely, I have always alternated 10k and marathon seasons....a spring 10k program and a fall marathon program. Thus, I always trained for both distances each year. Maybe that has made me a reasonably "balanced" runner.

So, what is the solution to the quandary concerning improving the accuracy of marathon time prediction for other runners, especially less experienced marathoners? Ideally, I would like to see a race calculator that incorporates a measure of training intensity. But, that would require that someone spend a lot amount of time and effort to gather a very extensive data base and translate it into a new generation of race calculator algorithms. That isn't likely to happen in my lifetime.....or yours! So, given all of this confusion about the veracity of race calculators, how should you go about increasing the accuracy of predicting a marathon time with the tools that are available today? I think that there are a couple of approaches.

As I said previously, build a data base over several marathon cycles that will enable you to develop your personal ratio(s) based on your preferred baseline distance(s). You can then use subsequent marathon experiences to keep your personal ratio(s) updated and fine tuned. That will free you from the ambiguities of calculator results. Meanwhile, use one of the following two approaches:

(1) Run a 10k race 3-4 weeks before a marathon. Use both a race calculator and Brian's table to predict marathon times. Then, base your marathon goal and race plan on the slower of the two methods.....remember, a little conservatism is better than being overly aggressive. Chances are that Brian's table will yield the slower prediction. If you feel strongly....and honestly....that Brian's table is overly conservative for you, then shift a category or two within the table. But, you should probably not go all the way to the calculator prediction, unless you are training at a very high mileage level or already have marathoning experience that clearly indicates that it would be realistic to do so.

(2) If a 10k race isn't available 3-4 weeks before your marathon or you prefer to run a different race distance, then do the following:

----- (A) Run a race of any distance between 5k and metric marathon 3-4 weeks before the marathon.

----- (B) Use a race calculator to convert the results of your baseline race to a predicted marathon time.

----- (C) Use the same race calculator to also convert the results of your baseline race to an equivalent 10k time.

----- (D) Use Brian's tables to convert the equivalent 10k time from (C) above to a predicted marathon time. Use the guidelines in (1) above to further adjust this predicted time if you think that Brian's tables are overly conservative for you.

----- (E) Base your marathon race plan of the slower of (B) or (D) above.

This approach might put you on the conservative side of where you really are....unless you adjust the predictions of Brian's tables too much. However, I can guarantee you that being on the conservative side is much better than being on the aggressive side of your true ability, which is where simply plugging a 5k race time in a calculator will put many....maybe most....people! And it offers a bit more structure to marathon goal setting and race planning than just a "Kentucky windage" adjustment to a calculator result.

Jim2