2021: the 900s and the rest of the season

Updated 10 Oct 2026

Running notebook · Running record · Personal bests · Race history

2015–2019 · 2020 · 2021 · 2022 · 2023–2025 · 2026

The workout I keep returning to from 2021 is 5×900 m with three minutes of recovery. On 30 April I ran 2:48, 2:48, 2:47, 2:49, 2:49. I recorded it as a personal best: a rounded 2:48 average, or 2:48.2 from the written splits, about 3:07/km over the nominal 900 m. Five repetitions within two seconds of each other. In my present return to training, getting back to those 900s is still something I look forward to.

But that one good evening is an incomplete picture of the season. I repeated the workout twelve times between March and May. Some days I started too fast. Some needed extra recovery. A fortnight after the best set, I wrote that I was completely burned out after the third repetition.

The wider record matters too. There was an 18:16 South Park 5 km race, longer city runs, shorter combinations on the track, and an autumn Sofia Marathon short-distance race. The 900s are a useful way into that year, rather than its entire content.

February: longer efforts

On 8 February, I recorded two efforts titled 5 km, in 19:18 and 19:16. On 19 February, the titles were 5 km in 18:46, then 2.5 km in 9:14. I have since confirmed that the 18:46 was on the track; it is my recovered 5000 m training best. The native timer corroborates the written duration but records no distance or location. These are the intended distances and written times; the recorded GPS distances are preserved separately in the ledger below. Their nominal effort paces were 3:52/km and 3:51/km on 8 February, then 3:45/km and 3:42/km on 19 February. The recovery between the efforts is not established.

These were part of the period I now remember as having comfortable longer running alongside the faster work. That recollection helps explain what I want back: being able to run through the city by feel, as well as going to the track for a measured session.

28 March: 15 km long run

The long run on 28 March contains fifteen consecutive kilometre laps in the watch record. Their elapsed times total 1:03:16.68, about 4:13/km, with no intervening timer pause. The full upload is 15.01 km in 1:03:19. This is a GPS-distance training reference, without a measured-course claim. It is the quickest recovered 15 km reference and belongs alongside the faster track work from that spring.

March–May: the repeated 900s

The table preserves the written repetitions, including the question mark in the 19 March description. The averages are the rounded figures I recorded at the time; 30 April’s exact arithmetic mean is 2:48.2. The added work paces are arithmetic means of the written splits over the nominal 900 m, rounded to the nearest second. They exclude the recoveries.

There is a distance question worth keeping with these records. A March 2021 post in my export asks whether the small track at NSA was really 300 m, or closer to 293–294 m. A later comment suggested 295 m seemed plausible, but the exchange did not establish a measured length. I retain the original 900 m workout label here without converting it into a certified distance or rescaling the times from that estimate.

The early 2:43 was faster than any repetition in my best set, but it came with fading splits and a seven-minute recovery. The 30 April result is satisfying for a different reason: I could repeat the pace with the intended recovery.

Some comparisons remain imperfect. On 16 April, the fourth and fifth recorded work laps are 2:55 and 2:49, against 2:52 and 2:47 in my description; two consecutive short-distance laps also follow the first repetition. On 25 April, the last recorded lap is 2:53, rather than the written 2:52. I keep those differences with the comparison.

The slower May sessions also belong in the story. On 23 May, I recorded 3×900 m in 2:59, 2:58, 2:48 — nominal repetition paces 3:19/km, 3:18/km, 3:07/km — with the note: “I think I am overtraining, so I will take a rest week.” That note preserves my concern and the rest week I intended.

10 April: the South Park personal best

Between those interval sessions, I ran 18:16 at 5kmrun in South Park, finishing second overall and first in the male 15–19 group. The average race pace was 3:39/km. I was 17, in eleventh grade.

For years I remembered the result as about 18:20. Recovering the organizer’s record settled it: these were the same race, and 18:16 is my recorded 5 km personal best. It is a separate kind of evidence from the interval averages: one continuous organized race.

Other sessions, and a run home

18 April: longer repetitions

The 900s were not the only format. On 18 April, I ran 3×2 km with two minutes of recovery, in 7:26, 7:23, 7:23, or 3:43.0/km, 3:41.5/km, 3:41.5/km. My note says “City run; Wet surface.”

18 May: ten seconds sprint, ten seconds rest

The 18 May recording is titled 16×(10 seconds sprint / 10 seconds rest). A separately recorded warm-up and cooldown bracket it, and the recording contains programmed ten-second work/rest steps. When I brought this workout back into the conversation in 2026, I remembered inventing it and how brutal it felt. The archive supplies the date; it does not settle which adaptation it produced. Its present place as an experiment is described in the running notebook.

16 June: three efforts

On 16 June, I recorded three efforts in order: 3000 m — 10:17, 1500 m — 4:53, then 1200 m — 3:58. Their nominal effort paces were 3:26/km, 3:15/km, 3:18/km, respectively. The final description is “Totally burnt out.” Their nearby start times establish the sequence, but I have not recovered a description saying whether they were planned as a combined workout or separate tests. I therefore leave the 3000 m effort’s classification open.

Summer: combinations and running home

The summer 900s also continued inside paired blocks of 900 m + 3×300 m. I recorded two numbered blocks on 8 June, 20 June and 5 July. Their opening 900s were 2:43 / 2:43, 2:40 / 2:40, then 2:38 / 2:40; the following 300s were around 49–53 seconds. The nominal 900 m paces were 3:01/km in June’s first pair, 2:58/km in the second, then 2:56/km / 2:58/km in July. The recorded gaps were roughly three minutes after each 900 and one minute between the 300s. The numbered titles keep the two uploads connected, though they do not specify the intended rest between blocks. July’s 2:38 belongs with the 300s and recoveries that followed it; April’s 2:48.2 averages five 900s.

On 8 July, 1200 m + 2×300 m gave 3:32, 50, 48. The 1200 m effort was about 2:57/km. I still remember that as another high point. The record also includes subsequent 900 m/300 m and 300 m/100 m combinations that morning; isolating the first upload would miss part of the day’s work.

Two days earlier, the run had a simpler purpose. The 6 July activity is called “33 back home”: 7.59 km in 34:04 of moving time, about 4:29/km. I remember running home across Sofia after a night out, through the empty boulevards. That kind of freedom is the subject of Running makes the city smaller.

Autumn: longer running and the Sofia race

The later record includes 15.00 km in 1:08:55 of moving time on 27 September, about 4:36/km, followed by 14.10 km in 1:03:16 on 4 October, about 4:29/km. These are recorded distances and moving times, with exported elevation-gain estimates of about 58 m and 54 m. They give the autumn race some surrounding context, rather than leaving it as a single isolated result.

10 October: the Sofia short-distance race

On 10 October, I ran the Sofia Marathon’s short-distance event. The official results give 42:04 chip time, 42:06 gun time, and 18th among the men, under a 10.5 km heading. I remember trying to stay with the leading woman, holding on longer than I had in 2020, then falling behind and finishing strongly, completely spent.

Using that 10.5 km result label gives a chip-time pace of about 4:00/km. The contemporary short-course map advertises 10 km, so the exact course length remains unresolved. The race-results record keeps that distinction with the source links. I do not treat the time as a measured 10,000 m personal best.

Looking back, the repeated sessions preserve something a race time alone cannot: how uneven training can be around a peak. The April set remains a benchmark I want to return to. Its neighboring sessions make it a more useful one.