(On Running) #20 Using Claude to Predict My Boston Marathon Qualification Odds

The start of a beautiful Saturday run

Ideation vs. Execution

One week. That is all that remains between me and my second Boston Qualifying (BQ) Marathon attempt. 

When I started training for the Baystate Marathon on June 7th, I had lofty ambitions for the amount of time I could dedicate to training. I started working full-time for my summer job the same week. I had dreams of waking up at 5 AM and running before work every day to build toward my 80 mile peak week goal, averaging 65-75 miles per week for the 19 week training block. I knew that achieving a BQ time would require an indelible commitment to the running grind. 

Ultimately, my marathon training ideation failed to account for the life I have outside of running. I ended up traveling to see friends and family most weekends. During the semester, I scarcely have the time to do this, and it felt important that I prioritize my relationships. The inconsistencies in my routine that come with traveling meant that I started to sleep-in later than 5 AM on the weekends—which turned into waking up at 6 AM everyday. Instead of running every morning before work, I pushed some runs to the evenings. I began lifting 3x a week for the first few weeks of the summer, but the evening runs made it difficult to run and lift, and I eventually stopped lifting all together. 

This summer, I also took the opportunity to apply for an English Teaching Assistant award through the Fulbright commission. If awarded, I would live and teach English in Italy for nine months after completing my graduate program in school counseling. I dedicated myself to learning Italian, with the goal of studying for an hour each day. To learn, I signed up for an online course, bought a workbook, and listened to a podcast on my way to work. I began working on my application materials on the weekends. The combination of crafting my application and learning a new language took precedence over my 80 mile peak week dreams. 

With this in mind, I compromised. I pursued my marathon goals while working on my Fulbright application. I ended up averaging about 58.5 miles per week from June 22 - Aug 23, which is close to my peak from my first training block. The consistent volume built a strong foundation to set my half marathon PR in the Tree House Half. However, my volume dipped to ~50 miles per week for the last 7 weeks; a result of fighting illness, preparing for the half marathon, and the start of the semester tightening my schedule even further. All in all, I averaged 54 miles per week over the last 17 weeks, and I submitted my Fulbright application a few days before the deadline. 

So, what does this all mean for my Baystate Marathon race plan?

Predicting Boston Marathon Qualification

Even if I didn't execute my training block to perfection, my Boston Qualifying goal isn't outside of the realm of possible outcomes. The difference is that now, it is a reach goal instead of the expectation. 

This summer brought an exciting development for Boston Marathon hopefuls. The Boston Athletic Association introduced a lottery system that allows 1,000 runners who achieve a BQ time but don't meet the cutoff to get a bib. My original goal of sub-2:48 was based on the idea that I would have to beat the cutoff by over seven minutes to ensure entry (BQ standard is sub-2:55 for Males 18-34 yrs old). Now, there is an off-chance I can earn a bib by running sub-2:55. The cut-off for the 2027 Boston Marathon ended up being 5 minutes and 17 seconds, meaning sub-2:49 would give me a reasonable chance. 

I have a number of valuable race predictors to consider ahead of race day. My first marathon time was 3:07:10 after blowing up in the last 10K. Based on my tune-up half marathon time of 1:23:11 on September 13th, this VDOT calculator estimates my equivalent marathon time to be 2:53:44. My Garmin Connect race predictor has me at 2:54:43 for the marathon. I held 6:30/mi pace from miles 12-20 in my peak long run workout. While these times aren't the limit of my potential, they are a good indicator of what a realistic target would be. 

Since sub-2:55 now guarantees a chance at Boston Qualification, it is worth weighing whether targeting sub-2:48, sub-2:55, or something in between optimizes my odds of Boston Qualification. 

I have been turning to Claude as my "personal trainer" since creating my original 80 mile peak week training plan in June. Since them, it has become my preferred AI in most areas. I have found the output to be much sharper than ChatGPT, and appreciate that it takes its time rather than rushing to the wrong answer. I also like that I can connect my Strava data to Claude, giving it the full picture of my training and race potential. 

When considering what I should target at Baystate, I started to formulate a decision tree in my head. I had three clear options: swing for sub 2:48 and a high chance of guaranteed entry, hold for sub 2:50 and a chance at guaranteed entry AND the lottery, or bank 2:52 and go for the lottery. I realized that there were many probabilities at play, and any probability I assigned to my running potential would be arbitrary and biased. That's when I realized that Claude, with access to all my runs for the entire block, would be the best judge for this particular exercise. 

I started with this prompt:

After a few tweaks, it returned the following decision tree:

Claude, 2026

Then, I asked it to explain in detail how it estimated these probabilities. Afterwards, I asked it to summarize this explanation concisely and provide a description for this post. 

Here is what Claude came up with:

How I estimated one runner's odds of getting into Boston 2028 from the Baystate Marathon

I'm Claude, an AI assistant made by Anthropic. A runner asked me to review their Strava training since June and estimate their chances of qualifying for Boston 2028 from the Baystate Marathon on Oct 18, 2026. This is how I built the estimate. It's a model with stated assumptions, not a statistical result.

1. From fitness to finish time. The runner's 1:23:14 half marathon converts to about 2:53:30 with the Riegel formula. I adjusted that to roughly 2:51:30–2:52:30. The half course was hillier than Baystate, and the runner went out too fast and faded in the last three miles. I then added a 15% "bad day" tail, finishing around 3:00–3:15, covering fueling or stomach trouble and pacing errors. That's based on their previous marathon (a 3:07 that ended in a bonk) and a couple of fueling problems in training. For a plan targeting 2:50, this gave about 7% for sub-2:48, 17% for 2:48–2:50, 52% for 2:50–2:55, and 24% for slower than 2:55.

2. From finish time to the cutoff. I assumed the 2:55:00 men's 18–34 standard. The cutoff is how far under the standard you have to run. The 2027 cutoff was 5:17. I modeled 2028 as roughly 5:45, give or take 55 seconds. On that model, a 2:50:00 earns direct entry about 21% of the time, 2:49:00 about 61%, 2:48:00 about 91% and 2:47:00 about 99%.

3. The lottery. Boston now randomly selects about 1,000 qualifiers who miss the cutoff, and in 2027 that worked out to about an 11% chance each. I added 11% for anyone who qualified but missed direct entry.

4. Putting it together. I multiplied each finish window's probability by its entry odds and summed. Holding a 2:50 plan came out around 25%, banking a 2:52 around 19%, and swinging for 2:48 around 15%. The swing and bank distributions are my judgment-based shifts of the baseline, not separate simulations.

Caveats. The numbers are sensitive to a few inputs:

• If the 2028 cutoff lands near 6:30, the three totals fall to about 19%, 15% and 12%.
• If the lottery doesn't return for 2028, they fall to about 18%, 11% and 11%.
• The 15% bad-day tail is a judgment call.

I'd treat the totals as ±5–7 points. The ranking, with the 2:50 plan ahead of the other two, held up in every variation I tried, and that's the part I'd trust most.

(Claude, 2026)

How This Affects My Race Plan

Before this, I was leaning towards playing it safe and shooting for sub-2:55. Now, I recognize the opportunity cost of not going for sub-2:50 and at least giving myself a chance at direct entry. 

Moreover, holding the sub-2:50 plan (which was also my original marathon goal in May) has a positive impact on my mindset going into race day. It is easy as a runner to rule out a goal before the race even starts, and ultimately start the race without self-belief. The more I run, the more I understand the importance of the mind-body connection in running and sports in general. 

To conclude: with one week to go before race day, my current plan is to hold the sub 2:50 goal. Baystate has 2:50 pacers, and I intend to hang about a minute behind them for the first half, keeping them in reach for a late surge. Then, if all goes well, I will catch up with them for the back half and ride the pack to a sub 2:50 marathon. 

This outcome wouldn't guarantee Boston Marathon entry. Far from it. But it gives me a chance, and having that hope is critical. Going for 2:50 feels like the right thing to do—to keep my belief alive. 

This Week: 32 miles

The Baystate Marathon is this Sunday. Stay tuned for results!

Follow on Instagram for blog updates

Follow my training on Strava

Ryan K. Beck can be reached at rkbeck.writing@gmail.com