The new Boston Cutoff Tracker was launched with results through November 15, 2026. For a detailed breakdown of the data, see the deep dive here.
The Tableau dashboard below collects data from marathons, tracks the number of finishers who meet their Boston qualifying time, and projects an estimated cutoff time for the 2028 Boston Marathon.
I will make regular updates to this dashboard, typically on a weekly basis, through the registration period in September 2027. For more details on the data, the assumptions, and other factors, scroll down below the dashboard. I’ll follow up with a more complete prediction towards the end of the qualifying period. Follow me on Threads for the latest updates.
Want to Stay Up to Date with the Latest Data?
I will continue to update this dashboard throughout the qualifying period as additional races take place.
Every month or two, I will write up an analysis of the recent races and what effect they’ve had on the projected cutoff time. Use the form to subscribe to my weekly newsletter, and I’ll let you know when those are published.
More frequently, I’ll just be adding data to the dashboard. In the top right corner you’ll see the date of the latest results, and I’ll also post something on Threads. Follow me there if you want to be the first to know when new data is added.
Updates and Analyses
Throughout the qualifying period, I will include links here as I publish more in depth analysis related to the cutoff time:
How Accurate Was This In the Past?
This is everyone’s favorite question – cut to the chase and tell us how well this worked in the past!
For 2027, the final projection on the tracker (5:13) was almost spot on. The actual cut off was 5:17.
That’s not to say that I’m perfect. This required a little bit of luck. I had actually assumed that a few things would push the cutoff slightly lower than the projection, and my final prediction was between 4:50 and 5:10.
The number of applicants (33,649) ended up being very close to the final projection (33,811). The number of accepted applicants was slightly higher than expected, though, and if this had been known ahead of time the final projection would have been 4:59 (slightly low).
You can read more reflections about the 2027 Boston Marathon projection here.
The previous year (2026) was the first year that I tracked the results throughout the year, and over the course of the year the projection ranged from as low as ~5:00 to as high as ~6:30. By the end of the qualifying period, the tracker projected a cutoff time of 5:36. The actual cutoff time was 4:34.
This prediction essentially includes two parts. First, we’re predicting the number of applicants. Based on the tracker’s projection, we expected to see 34,118 applicants. There were actually 33,249. That’s a difference of 2.5%, which is not perfect but pretty close.
Then, we’re predicting the actual cutoff time. Since the prediction of the number of applicants was slightly high, so was the prediction of the actual cutoff time. But if we knew the actual number of applicants, some simple math would have yielded a prediction of 5:16. And if we had known the number of accepted applicants would increase by a few hundred, the prediction would have been 5:02. A little high, but not off by much in the grand scheme of things.
You can read some more reflections on the 2026 Boston Marathon projection here.
The previous year, I used similar methodology and came up with a prediction of 7:03. This was just slightly higher than the actual cutoff time for 2025 (6:51). But it also assumed there would only be about 22,000 accepted applicants.
They actually accepted 24,000 applicants. If we had known that ahead of time, the prediction would have been closer to 7:45 to 8:00. So in this case, the modeling came in lower than the actual cutoff time – but by less than a minute.
You can read more about that prediction here.
Ultimately, it’s very difficult to come up with a precise and accurate prediction of exactly what the cutoff time will be. There are a lot of unknowns that simply can’t be controlled for. But we do know enough to identify the correct ballpark, and I’d be very confident that the final projection will be within a minute or less of the actual cutoff time. And it’s quite likely that it ends up being within 20 to 30 seconds.
Keep in mind, however, that the projection will change throughout the year. It’s a roller coaster that goes up and down with each weekend’s results, especially early on in the qualifying period. While there are still results coming in, there are far more unknowns. Until the end of April, this is more like a weather vane to understand which way the winds are blowing. By the end of April, though, it’s unlikely there will be any further big swings.
What Data Is the Dashboard Built On?
For this dashboard, I’ve collected the results of marathons with 200 or more finishers in the United States and Canada. I’ve also included the results of London and Berlin, because those are part of the Abbott World Marathon Majors. Tokyo and Sydney are not currently included, because such a small percentage of their runners actually apply to run Boston, but I may change that in the future.
Based on its date, each race is classified as being in the 2027 or 2028 qualifying period. Unlike previous years, there is no overlapping double dipping period in which one race is eligible for qualification for both Boston Marathons.
Although I have the full dataset for the 2027 qualifying period, I’ve excluded the races that haven’t yet been run in the 2028 qualifying period. As the results become available for the 2028 qualifying period, I’ll update the dataset and include results for both periods.
There are some races that have zero finishers in one of the qualifying periods. If that race appears in the dataset, that means the race did not take place in that qualifying period – either because it was discontinued, it is new, or it was temporarily cancelled due to extenuating circumstances. Two of these races are in September, and the timing worked out that the September 2026 fell in the qualifying period for the 2027 Boston Marathon – leaving the September 2025 race without a corresponding race.
The dataset includes a finish time, age group, and gender for each runner. The same qualifying times are in place for each of the two qualifying periods. Both qualifying periods are subject to the new downhill results index adjustment, and times are adjusted accordingly. Each runner is identified as a qualifier or a non-qualifier and their individual buffer time is calculated.
The runner’s age when they ran their marathon is used to determine the appropriate qualifying time. It’s possible they age up before Boston, so the actual number of qualifiers is higher than the calculated number. But, we can assume that this difference is similar from year to year and that the difference washes out in the aggregate.
Runners are identified and have their results grouped according to a fuzzy match on their name, their age, and their gender if they’ve run more than one race in a given qualifying period. This method may miss some possible matches, but it does a good job of making most potential matches. By setting the “Use Fastest Time” option to “True,” the calculations will be based only on a runner’s fastest time. This reduces the impact of a single runner achieving multiple qualifying times in the same qualifying period.
The majority of the data comes directly from each race’s results. In some cases, the data is collected from another source, like Athlinks or Marathon Guide. I maintain a copy of the full dataset on Kaggle, if you are interested in doing your own analysis.
What Assumptions Are the Dashboard Based On?
This dashboard is based on several assumptions. These assumptions were first tested when I attempted to predict the cutoff time for the 2025 Boston Marathon, and they were refined while I tracked the results and attempted to predict the cutoff time for the 2026 Boston Marathon. The methodology is not perfect, but it holds up fairly well and is a good starting point for understanding how things are (or aren’t) changing.
The Boston Marathon cutoff time is based on three things: a) the total number of applicants, b) the number of accepted applicants, and c) the distribution of applicants’ finish times.
The biggest variable is the total number of qualified applicants. There are a lot of factors that go into this, and it’s impossible to fully account for all of them. But the core assumption that I’m starting from is that the number of applicants is directly related to the number of runners who meet their qualifying times.
Not every runner who qualifies for Boston will apply. But the percentage of runners who do apply should be fairly stable. Publicly available results allow us to track how many runners qualify. We can then calculate how that number changes. Finally, we can apply that change to the number of applicants from 2026 to project the number of applicants for 2027.
More complex modeling would take into account the runner’s gender and age, the race at which they qualified, and their individual buffer. I did take a stab at this more complex modeling previously, but it may have been a bit over engineered. This simpler approach ended up being more accurate.
The next variable is the number of accepted applicants. We can’t know exactly how many runners the Boston Athletic Association will accept. However, this has been between 22,000 and 24,000 in recent years. The three two years, it has been slightly over 24,000. There is no indication the field size will increase significantly, and therefore it’s unlikely the number of accepted applicants will grow beyond last year’s levels (24,630).
Therefore, I’m using that as the second assumption.
The third variable is the distribution of the actual qualifying times, which determines how many applicants need to be rejected to balance the field. Based on the data released by BAA when it announced previous cutoff times, I’m assuming this to be between 1,800 and 2,000 runners per minute.
Instead of simply using third variable to calculate the projected time, though, we can use a linear regression from the previous results. Over the past decade, there has been a stable, linear relationship between how many applicants are rejected and what the cutoff time is. We can simply plug the number of projected applicants into the equation and get a projected cutoff time.
With these assumptions – and the underlying data of results – it’s possible to make a simple prediction of what the likely cutoff time will be.
An Example of How the Math Works
Here is an example of how the math behind the projection works. These example numbers are based on the results available through early December.
First, take the number of qualifiers in last year’s qualifying period: 27,165.
Then, determine how much this increased or decreased in this year’s qualifying period. It increased 1,990 (7.33%) to 29,155.
Next, adjust the number of applicants in last year’s qualifying period (33,649) by the percentage difference in qualifiers (7.33%) to get the projected number of applicants in this year’s qualifying period (36,115).
Now, use the number of accepted applicants in last year’s qualifying period (24,630) to project the number of rejected applicants: 11,485. Convert this into a fraction of the total number of applicants: 11,485 / 36,115 = 0.31801.
Finally, plug this into the regression equation: (1237.32128 * 0.31801) – 30.292453 = 363 seconds. Or approximately 6:03.
Going back through 2015, this regression equation has only been off by more than 30 seconds once. It was 56 seconds high in 2024. In every other year, it was within 30 seconds – and in most years it was less than 20 seconds.
Always remember that there’s an element of chance to this, so it’s quite likely that the final outcome could be 10 to 20 seconds higher or lower. But it’s very unlikely that it’s more than a minute from the projection.
Can I Play With the Data Directly?
The individual results used for this dashboard is a large, raw dataset. The dashboard is designed to make it easier for your average runner to understand. I’ve invested a considerable amount of time into cleaning and processing that data to produce a simple set of visuals and a single projection.
But if you’re a data professional, you might be interested in doing your own analysis of this data – either to make your own Boston cutoff prediction or for some other purpose. I’ve published the full dataset, and you can read more about that here. This dataset includes races from earlier qualfying periods, as well.
Note that this dataset does include some additional race results that I have not factored into the tracker. You can make your own decisions of whether and how to incorporate them.
You can download the dataset as a set of csv files to work on locally. Or you can create notebooks on Kaggle to work with it in the cloud.
Future Enhancements and Further Analysis
The dashboard is a starting point, but I may make some enhancements over the course of the year to improve its ability to project and predict the Boston Marathon cutoff time.
First, I’m not currently including Sydney and Tokyo. I’ve also included Berlin and London as equally weighted races. I may revisit that decision or I may apply some weights to the Majors to account for differences in conversion rates.
Second, I’m not currently attempting to age runners up and calculate their age when they run Boston. Technically, this would be a more accurate reflection of the number of qualifiers. It’s impossible to identify everyone who will age up, but it’s possible to identify some of them. I may update the tracker with this at a later date.
Third, I don’t currently provide a way for you to adjust the assumptions. I plan to include some controls for you to change this like how many accepted applicants you expect. That’ll help with some of the “What if?” questions.
But in the past, the more I tweaked with things and tried to account for every variable … the less reliable the outcome was. So I may just stick with what I’ve got. We’ll see.
If you have other questions or suggestions, leave a comment below. I’m not trying to overcomplicate things, but I am open to suggestions to make the tool more useful.
