800–5000 m: a running notebook

Updated 2 Nov 2026

My larger project is human agency: the ability to act, and to keep acting throughout life. In The Optimization Nobody Is Running, I describe capability as something whose level matters throughout that life. Running is one concrete way to work on the physical part of the ambition.

I want to build elite physical capability and preserve it for life. Running is my foundation, from 100 m to the marathon, with strength and other physical skills alongside it. I think about the body through an engine, a structure and a driver: supplying the work, carrying it and coordinating the movement. Running appeals to me because it is such a simple, natural way to put those capacities to work. Its full range asks for speed as well as endurance. Strength and other skills extend what I can do beyond the abilities a running time tests.

Health and a longer functional life are part of what I hope that capability will support; event performances alone cannot establish those outcomes. Staying able to train, without repeatedly losing continuity to injury, is already part of the project.

For now, I am concentrating on 800+ points on the World Athletics scoring tables across the whole 800–5000 m range: 800, 1000, 1500, mile, 2000, 3000, two miles and 5000. I want that performance profile in one runner. The common scale makes my strengths and gaps visible; the training has to turn the ambition into something I can actually do.

Living project · Current training · Programme · Personal bests · Running record

This version includes completed training through 2 November 2026. The upcoming sessions are plans. The current record keeps the chronology, splits and conditions behind the summaries here.

Starting point · Goals · Programme · Observations · Next checkpoint · Model · Scoring notes

Where I am starting

Two kilometres has a particular place in my running. It was a childhood event distance and the length of a neighbourhood loop on which I accumulated a great deal of running. I remember 8:33 on that loop in sixth grade. Separate Kids Run results were 8:31 at age 12 and 8:13 at age 14. I no longer live near that loop, but it remains a personal anchor. Two kilometres also sits at the geometric midpoint of 800 and 5000: 800:2000 = 2000:5000. That is pleasing; it does not make the distance a uniquely revealing physiological test.

By 2022 I was racing on the track, with 2:03.62 for 800 m and 4:23.90 for 1500 m at the June national championships. That season also included repeated injury interruptions. The 2022 account follows the workouts, pacing decisions and races through the July U20 national championships, beyond the June peak. It gives me both a level to return toward and a reason to take continuity seriously.

The years since were uneven. I ran intermittently in 2023 and 2024, including substantial long runs, interval work in early 2024 and the Orange County Marathon on 5 May 2024. The large gap was 2025: my Strava has one run recorded for the year. Earlier 2026 self-timed 5Ks were around 20:30. What I am rebuilding now is regular, repeatable middle-distance training, with some experience still available and plenty of fitness to recover.

The first event benchmark of this return was 7:16.67 for 2000 m on 27 September. It is a 181-point table reference, beside the 855-point 5:30 goal. The trial was self-timed, with its method and pacing retained in the record. After it, I wrote, “At least we are on the scale.” The gap is real, and I now have a starting point from which later training and tests can mean something.

The 11 October 5000 m trial now adds a continuous longer-distance reference: 19:59 reported, 3:59.8/km, 161 points, beside the 810-point 15:00 goal. The phone failed near the finish; the dated account keeps that estimate with the splits. These two different events locate the current return within the larger project.

My personal bests and recorded best efforts have one table of original times, paces, World Athletics points, dates and race or training context, from 300 m to the marathon. It shows a stronger short end. The marks come from different years and settings, so they do not form a simultaneous fitness test. Nonstandard distances retain their original marks beside a labelled same-pace comparison in a named discipline; the mile reference scales the documented 1600 m pace to the full mile.

The runner I am trying to build

I began with an equal 810-point curve from an IAAF calculator: 1:58.03 for 800 m, 4:03.85 for 1500 m and 15:00.02 for 5000 m were examples of the same score at different distances. The tables are now published by World Athletics. That common scale is the backbone of my comparisons: where does my performance fall away as the event gets longer, and is the whole curve rising?

I later chose round goal times that put the strongest part of the intended profile around 800–2000 m. The central goals span 810–862 points, with 800+ as the shared floor. The scores below use the men’s outdoor columns of the 2025 revised World Athletics tables:

For orientation, exactly 800 points corresponds to 1:58.43 for 800 m, 4:04.73 for 1500 m and 15:03.36 for 5000 m in this edition. The rounded goals sit above that floor by different amounts. The official lookup awards the lower score when a mark falls between two listed thresholds; a calculator that interpolates or rounds can differ by a point.

The wider picture includes 12.00 for 100 m (615 points), 24.00 for 200 m (671), 52.00 for 400 m (744), 1:24 for 600 m (798), and 33:00 for 10,000 m (717). The lower scores at the edges define the shape I want: strongest in the middle distances, with speed below them and useful endurance above them. They do not turn the present programme into simultaneous preparation for every distance. These are ambitions, with no completion date attached.

When I discussed the profile, I kept returning to Jakob Ingebrigtsen and Josh Kerr as references for its shape and for how high the ceiling could be. I was imagining my own curve at a much lower point level, with particular interest in 1500 m and the mile. That comparison keeps the ambition open; it does not give me their training requirements or establish that my eventual curve will have the same shape.

Getting the whole range above 800 points means closing the gap toward the longer events as well as improving the 800. A promising interval session helps me choose the next workout. An event performance is what moves the scored curve.

Why the programme changed

I initially imagined a long tempo-first build. I then chose to start with a small mixture of sustainable work (LT), work intended to approach maximal oxygen uptake (VO₂) and fast running (F), with easy days between them. That let me observe more than one part of the returning runner without starting with large sessions.

There was another change of mind. At first, I liked varying workouts: it could keep every week from becoming a judgment against last week’s stopwatch. Once I had a few actual sessions, I wanted to see what would happen if I repeated them. If the same 3×1200 was becoming quicker and more manageable, why replace it with something that changed both the distance and the comparison?

Some of the work already feels familiar. The 900s bring back memories, and I described the three-minute efforts of 3×800 as feeling like home. Recognisable work makes pace, effort and reserve easier to compare. I can improve within a workout, then earn a change of structure, instead of changing everything at once.

That is also a durability question. I want regular track work and the freedom to run around the city; adding both quickly could bring back the interruptions I am trying to leave behind. A programme is useful only if I can keep doing it. Faster repetitions, more volume and more ordinary movement all have to fit into the same life and the same body.

The experimental work changed too. A fixed anaerobic-capacity A slot became X, a place to try a candidate session. My October 6 attempt at 10 seconds hard / 10 seconds recovery was broken into two sets by a phone failure. Its approximate distance included recovery movement, so turning that distance into a performance score would have rewarded the wrong thing. An experiment can still help me decide whether a format belongs in the recurring programme; it need not show a training improvement each time.

A programme that earns its next step

The original template uses 18 days: six blocks of two training days followed by a chill day. It contains four LT sessions, four VO₂ sessions, one stable F session, one mixed-rest FA, one event time trial and one X experiment. The programme page keeps the exact order, templates and progression tables.

The cycle began on 15 October. Its original eleven specified workouts and flexible X remain identifiable. On November 2 I chose an extra VO₂ repeat after a session disrupted by stomach discomfort, moving LT to the following day. That adds a thirteenth planned workout. I have not yet completed the cycle and shown that I can sustain it. Its cost also extends beyond the listed workouts: running to the gym, carrying a backpack and a hard bike commute still count as movement, even on a day whose programme says chill.

LT and VO₂ advance independently. The first LT gate was 3×1200 m averaging 4:48, with another repetition plausibly available. Passing it on October 8 unlocked 3×1600 m. The longer workout’s eventual 6:08 gate is a later target, rather than a compulsory pace on the first attempt.

On 15 October, VO₂ independently passed its 3×800 m, 2:40 average gate, without a major final collapse. That earns 3×900 m next. Its 2:55.5 average is the next graduation target, rather than a compulsory entry pace. Recovery remains nominally 5:30; feeling ready earlier is not an automatic reason to shorten it. The point is to give each repetition a chance to produce the intended work.

The later rows describe ambitious directions, not completed fitness or proof that the exact steps are optimal. Keeping the workout stable long enough to understand it makes a change of structure meaningful.

What the observations now tell me

Both interval progressions have now earned longer repetitions. On October 8, I ran three 1200 m repetitions with one minute between them: 4:42.5, 4:42.7, 4:41.3. They averaged 4:42.17 (3:55.1/km), with a 1.4-second spread. I finished feeling that a fourth repetition was definitely available, possibly a fifth. September 19’s same nominal workout had averaged 4:56.33 (4:06.9/km), and I had been spent after the third. The later session passed the 4:48 gate in both time and intended character.

Before the September session, a hard hilly bike ride had already left my legs burning. The improvement is an observed change in performance and feel; attributing all 14 seconds per repetition to training adaptation would ignore that starting condition. The intervening closed-track session used a longer asphalt loop with elevation changes, so its times cannot be dropped into the same comparison as measured track 1200s.

On the VO₂ side, 3×800 with the same 5:30 recovery moved from 2:48 average (3:30/km) on September 22 to 2:44 (3:25/km) on October 1. The later session felt more limited by breathing and less by the legs. That change in experience is useful, although it does not measure oxygen uptake or identify the cause of the improvement.

On October 9, I completed ~2:43, 2:39.4, 2:39.3. The first is a conservative estimate after the phone died at the finish. Another shutdown added a partial 250–300 m effort and a 60–90-second walk before restarting the second full 800. All three full repetitions were done, and the last two were tightly matched despite a very hard finish without a sprint. Using the first estimate gives about 2:40.6 (3:20.7/km) on average; it may have been quicker. This was a good completed session. I left wanting another 3×800 to cement the 2:40 gate with reliable timing.

The fast work needs a different comparison. September 24’s six 200s averaged 33.64, relaxed and closely grouped. On October 5, after an overly relaxed first repetition, the last five averaged 31.82. I was making more deliberate effort to maintain knee lift, quick return and posture within each repetition. The effort changed along with the times. Those faster 200s tell me something about fast running; an 800 time trial still has to answer the endurance question.

On October 12, the same 6×200 m / 3 min structure averaged 30.76. The first four averaged 31.25; I then pushed the last two to be more sprinty, running 30.22 and 29.30. It felt good. One repetition is now below thirty seconds in this series, with the extra intent recorded beside it. I want to see how repeatable that speed becomes with the movement I am trying to preserve.

After October 8’s LT session, I thought the earlier fast work had helped make it more relaxed, even while I was still sore from F and A. My fast stride was opening and my knees were coming higher; slower running around 4:00–4:05/km seemed to need very little push. Whether the returning movement has lowered the cost of that running, and which work helped it return, are questions to investigate.

The 11 October 5000 m made the continuous-running question concrete. I report 19:59 (3:59.8/km), worth 161 points. The first four kilometres were 3:58, 4:00, 4:02, 4:02, reaching 4000 m in 16:02; the final 3:57 is estimated from the accepted finish after the phone shut down in the last 50 m. This was one of the most unpleasant efforts of my life. My legs were heavy, breathing felt spent, and posture and knee lift faded after about 2500–3000 m. I broadly maintained the pace, but could no longer maintain the way I wanted to run.

I later clarified that I brought the posture and knee lift back together over the final two laps, especially the last. Holding good movement through the difficult middle and finding it again for the finish are different observations.

The controlled LT intervals had been about 3:55/km; this continuous effort was about 4:00/km, with a much greater felt cost. The trial adds the duration and continuity that the repetitions could not establish. The longer LT structure was already earned; VO₂ still needed its own confirmation. The movement I can maintain belongs beside the time when I compare efforts.

That confirmation came on 15 October: 2:35.26, 2:35.11, 2:36.27, averaging 2:35.55 (3:14.4/km) with a 1.16-second spread. The nominal recovery was 5:30; COROS recorded 5:25.67 and 5:20.82 between the efforts. The three-repetition average is 3.80 seconds quicker than October 9’s last two, although that earlier session’s interruptions make it an imperfect comparison.

It was extremely hard, but easier than October 9: “hard, not disgusting.” Breathing felt maximal in the last 200 m, yet I stayed composed, felt a kick was available and deliberately did not sprint. Finishing naturally felt like home. A brief loss of coordination early in the third repetition settled before the next curve. The gate is passed through the whole set and the way I completed it; the next question is how that control carries into 900s.

The feverish feeling had gone before bed on 14 October; my left Achilles was quiet during and after the workout.

The first 3×1600 m on 16 October averaged 6:23.88 (3:59.93/km), with 1:11.93 and 1:15.99 for the two 100 m walks. Early laps felt comfortable, but direct sun and mounting strain made the third repetition survival work. Its posture and knee-lift loss felt like the difficult middle of the 5000 m. I recovered good form for a roughly 95-second final lap, finishing the last 1600 in 6:27.07.

The nearly even times concealed how much harder the effort felt. Heat complicates the fitness comparison; it does not supply a numerical correction. The session makes preserving good movement through the middle of a hard effort a concrete question to follow.

On 19 October, the first 900s were 2:53.74, 2:56.22 and 3:01.05: 2:57.00 average (3:16.67/km), with recorded rests of 5:03.34 and 5:14.31. The first felt smooth and controlled; heavy legs persisted through the recoveries, and a strong headwind made holding pace harder. In the third rep I tried to recover form and find a finishing gear, but could not. The inner ankle was quiet before, during and after. The next aim is more even execution and a lower average at the same format.

On 21 October, the first 4×200 m / 3 min + 4×200 m / 90 seconds averaged ~30.31 then 30.50, with a 4:56.05 break between sets. One longer repetition is converted to a 200 m comparison; the record preserves its original time and distance calculation, alongside an interrupted start before repetition seven.

The opening repetitions felt controlled, about 0.94 seconds quicker than October 12’s first four. The second set stayed close in speed while the effort changed dramatically. The track was wet and my shoes slipped. My legs were already stiff after the first set; severe stiffness arrived in the last 50 m of repetition seven and earlier in repetition eight. I tried to drive harder into the ground and use my arms more to hold the speed. I finished completely spent, with burning legs for about ten minutes.

I think holding that speed was impressive, and I want to repeat the format. The nearly even times also conceal the late breakdown. It is a concrete reference for producing fast running under fatigue; whether that practice helps me preserve movement and find a finish in the longer reps is still a question. Right-ankle tension afterwards changed my walking for a couple of hours despite no ankle pain; it had settled by the discussion. I planned rest next.

The second 3×1600 m on 23 October was 6:20.51, 6:18.81, 6:16.05: 6:18.46 average (3:56.54/km), with 1:02.68 and 1:11.25 recovery. The mean improved by 5.43 seconds per repetition. More importantly for how I experienced it, I felt comfortable through the laps and kept good posture. It was still hot, with a difficult fourth lap in rep two, but I avoided the previous breakdown. Quicker final laps were part of the pacing; the near-even whole reps conceal that variation.

I still had mild glute and thigh soreness from the 200s. Both ankles felt under tension afterwards, without changing my walking, and were already easing by the discussion. The result is progress towards 6:08 per 1600 (3:50/km), with conditions and recovery kept alongside the times.

After three days without running while unwell, the second 3×900 m on 27 October was 2:57.25, 2:57.58, 3:01.57: 2:58.80 average (3:18.67/km). Actual recoveries were 5:00.20 and 5:04.03. I felt composed and finished strongly, with my ankles feeling fine. What was missing was access to faster running from the start. Opening in 2:57 felt slow, yet I did not feel I could choose to run 2:50. October 19’s quicker opening had felt like flying before the later breakdown; today’s experience was different.

The average was 1.80 seconds per rep slower, with slightly shorter rests; the third rep differed by just 0.52 seconds. I interpret it as a return after the short illness and expect the next efforts to feel sharper. That is an expectation, not an identified cause or a demonstrated loss of fitness.

The third 3×1600 m on 28 October was 6:19.82, 6:20.62, 6:15.86: 6:18.77 average (3:56.73/km), with 57.42 seconds and 1:06.01 recovery. That almost matched October 23, but I had deliberately paced conservatively. After struggling in the final 900s, I was unsure whether my legs would hold and wanted to feel good in each rep. I kept my posture and composure, then was less cautious in the quicker final rep. I think I might have used that pace earlier; repeating it across all three remains untested.

I also lightly felt tension along the inner right lower leg, just above the ankle bump, during and after the run. It had settled by the later discussion, but I decided to rest on Thursday, October 29, then do X = 4+4×200 m on Friday, October 30. The remaining sequence moves forward by one day. The programme keeps the remaining provisional dates, including the November 4 800 m trial.

On 30 October, I completed X by repeating the 4+4×200 m. Recorded set averages were 30.36 and 30.40. Late button presses in the first and third reps make the opening set uncertain; my more likely delay estimates give ~30.13, with the original times preserved in the record. The 3:03.23 set break was almost two minutes shorter than October 21’s. In rep seven I tried to stay relaxed on the straight; in rep eight I committed everything and ran 30.11, finishing completely spent. It is an encouraging repeat at similar speed with less recovery and a maximal finishing effort.

The 1 November 900s were 3:03.63, 2:58.33 and 3:06.34: 3:02.77 average (3:23.07/km), with 5:11.14 and 10:48.17 recorded recovery. I had eaten too much at lunch and felt nauseated, urgently needed the toilet, and stopped around the second repetition. I ran at night in the cold on an unlit track and felt awful while running. This differs from October 27’s missing faster gear: the slower times came with severe stomach discomfort and an interrupted session. I chose to repeat VO₂ on November 2 and move LT to November 3. The repeat was subsequently completed on November 2.

The repeat on 2 November was 2:56.82, 2:55.43, 2:54.84: 2:55.70 average (3:15.22/km), with 4:59.86 and 6:03.16 recorded recovery. Crowds prompted the longer wait before the final rep, which still required many outside passes, including towards lane three on the bends. I felt good through that fastest rep and afterwards. The mean improved on October 19 by 1.31 seconds per rep; the finish improved by 6.21 seconds. This time the quicker average, progressively faster reps and my experience line up positively.

I then joined another group for three 200s with walk-back recovery, estimated around 30 seconds each by feel. The group timed them, but I did not obtain the splits. My ankles felt good. The recorded 900 average is 0.20 seconds per rep above 2:55.5: close to the target, not graduated. It is a good result. I aim to cement the graduation target next cycle; 3×900 m remains the current format until the gate is earned.

The returning speed makes the shape of the performance curve interesting. I hope to challenge 2:20 in the 800 m trial, and possibly 2:16 as the remaining training develops. In the same 2025 men’s outdoor tables, 2:16.00 would score 418 points, equal in score to 4:43.60 for 1500 m, 6:29.56 for 2000 m and 17:32.09 for 5000 m. That would be 257 points above the October 11 5000’s 161. The 800 is still an ambition; differently dated tests during a comeback do not establish a simultaneous fitness profile.

Goals, plan and log belong together because each constrains the interpretation of the others. The goals explain why these observations matter. The plan makes the next comparison possible. The log preserves what actually happened, including the details that a clean average would hide.

Scores and rarity

The score gives me an equivalent performance at another event. The same point total at 800 and 5000 may still require very different work from me. My ambition is to bring both ends, and the events between them, into the same performance band.

Rarity needs a population as well as a score. The official tables use athletics statistics to compare performances, but they do not provide a percentile among all men or all runners. An annual best-by-athlete list can place a mark among recorded competitors; a race’s results can place it among that race’s finishers. Neither denominator describes everybody who could run the time.

One concrete reference is World Athletics’ 2025 men’s senior 800 m list, checked in October 2026 with “Best by Athlete,” world-wide coverage, electronic timing and both outdoor and indoor performances. It lists 7241 athletes strictly faster than my 1:56 goal, and 10,973 strictly faster than 1:58.43, the outdoor 800-point threshold. These are counts against times in that published list, whose reporting limit is 2:00.00; they are not a percentile among all runners or a count of equal-scoring indoor and outdoor results. The tied boundary rows are on page 73 and page 110.

Rarity-based scoring systems do exist. A 2007 paper in the IAAF’s research journal discusses Dale Harder’s separate scale, in which a further 100 points represents a tenfold reduction in the qualifying fraction. That rule belongs to a different scoring system; applying it directly to my World Athletics points would invent a conversion the current tables do not establish.

The rarity of the entire 800–5000 profile would require linked performances from the same athletes across the range. The events are related, so multiplying separate event rarities would not answer that question. An athlete who has no listed result at one distance may simply never have raced it.

The working model

The programme’s labels describe intended kinds of work. They give me a way to ask what a session is for before judging it by how hard it felt.

F — fast running. Produce a useful fast velocity with good movement and enough recovery to repeat it. How fast I can run and how long I can preserve that movement are separate questions. The current benchmark is 6×200 m with three minutes between repetitions; it is not an all-out 200 m test.

VO₂ — work near maximal oxygen uptake. The quantity I want is actual time near that state. Minutes at an externally chosen pace do not directly measure it. How quickly oxygen uptake rises, how long the effort remains tolerable and what recovery restores all enter the question.

LT — sustainable high lactate turnover. I am interested in what can be sustained while lactate appearance and disposal remain balanced at a high rate. Blood concentration alone cannot tell me those rates. LT is also a working label for my hard, controlled repetitions in support of 3000–5000 m. I have not established their relation to maximal lactate steady state or critical speed.

A — anaerobic capacity. I am exploring the idea that time spent with a finite reservoir deeply depleted might matter, rather than merely recording the highest instantaneous output.

My present measurements are field times, felt effort and event results; I have no direct measurements of oxygen uptake, lactate flux or anaerobic capacity. The LT and VO₂ labels name the intended work, rather than certify a measured internal state. Faster LT-labelled running might reflect changed lactate handling, improved movement, different pacing, better recovery or several changes together.

The fuller argument belongs in What should a running workout maximize?. Here, the model has to produce decisions that I can carry onto the track and observations that can inform the next decision.

The next checkpoint

Ten of thirteen intended workouts are complete, through November 2. That is nine of the original twelve plus the extra 900 session. The repeated 900s are now the best reference at that format; the additional three 200s belong to the same session.

LT on November 3 remains planned at 3×1600 m. After 6:18.77 average (3:56.73/km) last time, I want to move towards the low 6:10s: open a little more confidently, aiming under 6:15, then try to accelerate across the set. Around 1:33 per lap means 4:39 through 1200 m and 6:12 for 1600 m (3:52.5/km). The 6:08 average LT gate is unchanged.

The 900s’ 2:55.70 remains close to the 2:55.5 target, without graduation. I aim to cement that next cycle; 3×900 m stays current, with 4×900 m waiting until the gate is earned.

The earlier November 4 800 m TT, November 5 6×200 m and November 6 closing chill day remain provisional. I am treating the trial as a checkpoint within normal training, without a formal taper.

One LT slot remains before the time trial. Five LT sessions and three VO₂ sessions remains a possible reallocation for the next cycle, to assess after the remaining work and actual trial. The added repeat this cycle is a separate decision.

The next efforts ask three immediate questions:

• How does repeatability develop at the longer distances? The first 1600s and 900s supply references with different constraints. Each next gate is a target to earn rather than an entry requirement.
• How do pace and movement hold up together? The 5000 supplies a continuous event reference. I want to follow the ability to maintain my stride alongside the splits as the efforts become longer.
• How do speed and its cost change when I repeat the 200s? The October 30 repeat held similar speed with a shorter set break, timing uncertainty and an all-out finish. Whether this helps the longer efforts remains untested. The separate F benchmark stays at six reps with three-minute rests.

I had hoped for a 5000 around 19:30. The reported 19:59 is the result of testing that hope. Even 3:55/km continuously would give 19:35; the controlled interval session did not guarantee that continuous performance. I now have the effort’s time, pacing and movement to place beside the returning ease of the 1200s.

The track is one part of the capability I want. Precise sessions can coexist with running by feel around a neighbourhood, exploring on a Sunday, or casually crossing town on foot. Running makes the city smaller develops that ordinary freedom. I want the performance curve to rise, and I want to keep being able to use what I build.

This notebook holds the current choices as they evolve. The log retains the observations and their conditions. When a result changes the plan, the useful record is both the new decision and the evidence that earned it.