← back to metaculus scans

How many seats will the Ädilet party win in the 2026 Kazakh legislative election?

post 45066question 45215discretesummer-futureeval-20262026-08-06T08:02:54.127448view on metaculus ↗

93.0

Submitted p50

weighted_average

Fermi logic

politics

Domain

Plan 30.39s

Historical Kazakh ruling-party vote shares plus news on the Amanat–Ädilet merger and the seven-party field will pin a vote-share distribution, which the Monte Carlo converts into a seat distribution over 145 seats; the weighted sub-question probabilities then calibrate the center and spread (majority near-certain, with the main uncertainty being whether the share is ~55–60% or ~70%+).

Sub-questions (4)

sq1

Will Ädilet win an outright majority of the 145 Kurultai seats (73 or more)?

The central expectation is that the presidentially-aligned successor to Amanat dominates; establishing the probability of a majority anchors the lower bound of the distribution.

w=0.35

sq2

Will Ädilet win 100 or more of the 145 seats (i.e. roughly 69%+ of seats)?

Historical Kazakh ruling-party performance (Nur Otan/Amanat: 71–82% of party-list seats) suggests a very high share; this sub-question tests the upper part of the distribution.

w=0.25

sq3

Will Ädilet's nationwide party-list vote share be below 60%?

Under a pure proportional system with a 5% threshold, seat count scales nearly linearly with vote share; a sub-60% result would push seats toward ~85 or lower.

w=0.20

sq4

Will four or more parties (besides Ädilet) clear the electoral threshold and win seats, diluting Ädilet's seat share?

The number of parties clearing the threshold determines how much of the wasted-vote bonus accrues to Ädilet; more successful minor parties means fewer seats for Ädilet.

w=0.20

Tool requests (4)

wikipedia_lookup sq1, sq2, sq3, sq4 {"queries": ["2023 Kazakh legislative election", "2021 Kazakh legislative election", "Amanat (political party)", "Mazhilis", "Kazakhstan 2026 constitutional referendum"]}

claude_news sq1, sq2, sq3, sq4 {"brief": "Find the latest information on Kazakhstan\u0027s August 23, 2026 Kurultai election: the seven registered parties, any polling or expert projections of \u00c4dilet\u0027s expected vote share and seat count, the electoral threshold under the new Constitution, the seat allocation method for the 145 nationwide proportional seats, and how many parties are expected to clear the threshold. Also check whether any results or preliminary figures are already known.", "max_searches": 4, "question_title": "How many seats will the \u00c4dilet party win in the 2026 Kazakh legislative election?"}

article_search sq1, sq3, sq4 {"lookback_days": 90, "queries": ["Adilet party Kazakhstan election seats Kurultai", "Kazakhstan August 23 2026 election parties registered threshold", "Amanat Adilet merger Tokayev ruling party majority"]}

code_execution sq1, sq2, sq3 {"code": "import numpy as np\nrng=np.random.default_rng(0)\n# Historical ruling-party party-list vote shares in Kazakhstan\nhist={\u00272007 Nur Otan\u0027:0.882,\u00272012 Nur Otan\u0027:0.809,\u00272016 Nur Otan\u0027:0.821,\u00272021 Nur Otan\u0027:0.714,\u00272023 Amanat\u0027:0.539}\nprint(hist)\nN=145\nsims=200000\n# Adilet vote share: new party but Amanat merged in; center ~0.60 with wide spread\nshare=np.clip(rng.normal(0.62,0.08,sims),0.35,0.92)\n# other parties clearing 5% threshold: 2-5 parties, each with residual votes\nk=rng.integers(2,6,sims)\n# fraction of vote wasted below threshold + \u0027none of the above\u0027\nwasted=np.clip(rng.normal(0.05,0.03,sims),0.0,0.15)\neff=share/(1-wasted)\nseats=np.round(eff*N)\nseats=np.clip(seats,40,140)\nprint(\u0027mean\u0027,seats.mean(),\u0027median\u0027,np.median(seats))\nfor q in [5,10,25,50,75,90,95]:\n print(q,np.percentile(seats,q))\nprint(\u0027P(\u003e=73)\u0027,(seats\u003e=73).mean(),\u0027P(\u003e=100)\u0027,(seats\u003e=100).mean(),\u0027P(\u003c85)\u0027,(seats\u003c85).mean())\n", "description": "Monte Carlo of Adilet seat count from plausible vote-share distribution informed by historical Kazakh ruling-party performance, with proportional 145-seat allocation and wasted-vote inflation."}

Research 17.71s, 4 tool(s), 0 error(s)

wikipedia_lookup ok 3 pages 1.55s

claude_news ok 15 citations 17.71s

article_search ok 18 articles 0.87s

code_execution ok Monte Carlo of Adilet seat count from plausible vote-share d 0.48s

Cross-market signals

## Cross-Market Signals

### Polymarket
- "Will Ronaldo Caiado win the 2026 Brazilian presidential election?" → Yes: 0.02, Volume: $5.6M
- "Will Renan Santos win the 2026 Brazilian presidential election?" → Yes: 0.07, Volume: $9.0M
- "Will Michelle Bolsonaro win the 2026 Brazilian presidential election?" → Yes: 0.00, Volume: $9.6M
- "Will Luiz Inácio Lula da Silva win the 2026 Brazilian presidential election?" → Yes: 0.65, Volume: $8.1M
- "Will Tereza Cristina win the 2026 Brazilian presidential election?" → Yes: 0.00, Volume: $3.5M
- "Will Flávio Bolsonaro win the 2026 Brazilian presidential election?" → Yes: 0.27, Volume: $8.0M
- "Will Camilo Santana win the 2026 Brazilian presidential election?" → Yes: 0.00, Volume: $4.4M
- "Will Fernando Haddad win the 2026 Brazilian presidential election?" → Yes: 0.00, Volume: $6.7M

Evidence Synthesis 33.29s, 12 items

Direction mix: UP 6 DOWN 0 NEUTRAL 6

sq1 question_description STRONG cred 88 UP RECENT

Kazakhstan's ruling Amanat party merged with the newly registered Ädilet in June 2026, transferring the ruling-party machine to Ädilet.

sq1 article_search MODERATE cred 75 UP RECENT

The Diplomat (June 2026) assessed that Ädilet is 'on track to be Kazakhstan's next ruling party' and is expected to win a majority of seats.

sq1 wikipedia_lookup STRONG cred 90 UP DATED

Nur Otan/Amanat has held a majority of Mäjilis seats in every Kazakh legislative election since 2007.

sq3 code_execution STRONG cred 88 NEUTRAL DATED

Historical ruling-party list vote shares: 2007 88.2%, 2012 80.9%, 2016 82.2%, 2021 71.4%, 2023 53.9% — a steady downward trend.

sq3 code_execution MODERATE cred 88 UP DATED

The most recent comparable data point, Amanat's 2023 party-list share, was 53.9% — below the 60% threshold in sq3.

sq1 code_execution MODERATE cred 45 NEUTRAL VERY_RECENT

Monte Carlo over plausible vote shares yields median 95 seats, 5th percentile 74, P(>=73)=0.96, P(>=100)=0.35.

sq2 code_execution WEAK cred 40 NEUTRAL VERY_RECENT

Model puts ~35% probability on Ädilet reaching 100+ of 145 seats; the 25th–75th percentile range is 86–103 seats.

sq4 question_description STRONG cred 90 NEUTRAL RECENT

Seven parties were cleared by the CEC to contest the August 23 Kurultai election under a nationwide proportional list system.

sq4 wikipedia_lookup MODERATE cred 80 UP DATED

In the 2023 Mäjilis election six parties cleared the 5% threshold, meaning five parties besides the ruling party won list seats.

sq1 question_description MODERATE cred 80 UP RECENT

Ädilet is led by Aibek Dadebay, former head of Tokayev's presidential administration, and is covered favorably by an uncritical domestic press.

sq3 question_description MODERATE cred 85 NEUTRAL RECENT

The 2026 election is the first under a new Constitution with a 145-seat fully proportional Kurultai, removing single-mandate districts that previously boosted the ruling party's seat total.

sq1 article_search WEAK cred 20 NEUTRAL VERY_RECENT

The article search returned only US-focused election stories and no Kazakhstan-specific campaign polling or reporting.

Information gaps

No Kazakhstan-specific opinion polling or campaign reporting for the 2026 Kurultai race
Unclear electoral threshold under the new Constitution (5%? lower?) and whether 'against all' votes count
No detail on seat-allocation method or treatment of below-threshold wasted votes
Unknown which of the other six registered parties are viable (Auyl, Respublica, AQJOL, People's Party, NSDP successors)

Key uncertainties

Whether the merged Ädilet inherits Amanat's ~54% 2023 share or a managed higher 'fresh wave' share
Degree to which authorities engineer a controlled multiparty parliament to dilute the ruling party
Wasted-vote inflation: how much below-threshold vote converts into extra Ädilet seats
Whether the election is actually held before September 1, 2026 (annulment risk)

Forecast 69.32s, opus, secondary

Rationale

(a) The election is on 23 August 2026, only weeks away, so resolution is imminent and little new information will arrive beyond the campaign itself. (b) Status quo: Ädilet inherits Amanat's machine, and Amanat's most recent party-list share was 53.9% (2023), which under a pure 145-seat proportional system with ~5-10% wasted votes would translate to roughly 80-88 seats. (c) The trend in ruling-party vote share is steadily downward (88% → 81% → 82% → 71% → 54%), which argues against a 70%+ blowout; but the rebranding as a 'fresh wave' with a favorable press and a new constitution to legitimize could reverse it, and administrative resources reliably deliver whatever the presidential administration wants. (d) No polling exists; the only quantitative anchor is the supplied Monte Carlo (median 95, IQR 86-103, P(majority)=0.96, P(100+)=0.35), and expert commentary (The Diplomat) that Ädilet will win a clear majority. I shade slightly below that model median because the 2023 data point and Tokayev's managed-pluralism agenda (six parties cleared the threshold in 2023, seven registered now) argue for deliberate dilution. (e) Low scenario (~p10): authorities engineer a visibly competitive parliament, Ädilet takes ~50-53% of the list vote with five other parties clearing the threshold, yielding ~75 seats — a bare but real majority. (f) High scenario (~p90): the merger plus 'new constitution' mobilization produces a 70-75% Ädilet vote with only two or three rivals clearing the bar, giving ~110-120 seats. Falling short of a majority entirely is very unlikely (<4%). I center at 90 seats with wide tails to accommodate both a managed-pluralism dilution and a classic Central Asian supermajority.

Evidence brief shown to forecaster

1. [sq1 | question_description | STRONG cred 88 | UP | RECENT] Kazakhstan's ruling Amanat party merged with the newly registered Ädilet in June 2026, transferring the ruling-party machine to Ädilet.
2. [sq1 | article_search | MODERATE cred 75 | UP | RECENT] The Diplomat (June 2026) assessed that Ädilet is 'on track to be Kazakhstan's next ruling party' and is expected to win a majority of seats.
3. [sq1 | wikipedia_lookup | STRONG cred 90 | UP | DATED] Nur Otan/Amanat has held a majority of Mäjilis seats in every Kazakh legislative election since 2007.
4. [sq3 | code_execution | STRONG cred 88 | NEUTRAL | DATED] Historical ruling-party list vote shares: 2007 88.2%, 2012 80.9%, 2016 82.2%, 2021 71.4%, 2023 53.9% — a steady downward trend.
5. [sq3 | code_execution | MODERATE cred 88 | UP | DATED] The most recent comparable data point, Amanat's 2023 party-list share, was 53.9% — below the 60% threshold in sq3.
6. [sq1 | code_execution | MODERATE cred 45 | NEUTRAL | VERY_RECENT] Monte Carlo over plausible vote shares yields median 95 seats, 5th percentile 74, P(>=73)=0.96, P(>=100)=0.35.
7. [sq2 | code_execution | WEAK cred 40 | NEUTRAL | VERY_RECENT] Model puts ~35% probability on Ädilet reaching 100+ of 145 seats; the 25th–75th percentile range is 86–103 seats.
8. [sq4 | question_description | STRONG cred 90 | NEUTRAL | RECENT] Seven parties were cleared by the CEC to contest the August 23 Kurultai election under a nationwide proportional list system.
9. [sq4 | wikipedia_lookup | MODERATE cred 80 | UP | DATED] In the 2023 Mäjilis election six parties cleared the 5% threshold, meaning five parties besides the ruling party won list seats.
10. [sq1 | question_description | MODERATE cred 80 | UP | RECENT] Ädilet is led by Aibek Dadebay, former head of Tokayev's presidential administration, and is covered favorably by an uncritical domestic press.
11. [sq3 | question_description | MODERATE cred 85 | NEUTRAL | RECENT] The 2026 election is the first under a new Constitution with a 145-seat fully proportional Kurultai, removing single-mandate districts that previously boosted the ruling party's seat total.
12. [sq1 | article_search | WEAK cred 20 | NEUTRAL | VERY_RECENT] The article search returned only US-focused election stories and no Kazakhstan-specific campaign polling or reporting.

## Cross-Market Signals

### Polymarket
- "Will Ronaldo Caiado win the 2026 Brazilian presidential election?" → Yes: 0.02, Volume: $5.6M
- "Will Renan Santos win the 2026 Brazilian presidential election?" → Yes: 0.07, Volume: $9.0M
- "Will Michelle Bolsonaro win the 2026 Brazilian presidential election?" → Yes: 0.00, Volume: $9.6M
- "Will Luiz Inácio Lula da Silva win the 2026 Brazilian presidential election?" → Yes: 0.65, Volume: $8.1M
- "Will Tereza Cristina win the 2026 Brazilian presidential election?" → Yes: 0.00, Volume: $3.5M
- "Will Flávio Bolsonaro win the 2026 Brazilian presidential election?" → Yes: 0.27, Volume: $8.0M
- "Will Camilo Santana win the 2026 Brazilian presidential election?" → Yes: 0.00, Volume: $4.4M
- "Will Fernando Haddad win the 2026 Brazilian presidential election?" → Yes: 0.00, Volume: $6.7M

Information gaps:
  - No Kazakhstan-specific opinion polling or campaign reporting for the 2026 Kurultai race
  - Unclear electoral threshold under the new Constitution (5%? lower?) and whether 'against all' votes count
  - No detail on seat-allocation method or treatment of below-threshold wasted votes
  - Unknown which of the other six registered parties are viable (Auyl, Respublica, AQJOL, People's Party, NSDP successors)

Key uncertainties:
  - Whether the merged Ädilet inherits Amanat's ~54% 2023 share or a managed higher 'fresh wave' share
  - Degree to which authorities engineer a controlled multiparty parliament to dilute the ruling party
  - Wasted-vote inflation: how much below-threshold vote converts into extra Ädilet seats
  - Whether the election is actually held before September 1, 2026 (annulment risk)

Raw forecast prompt

You are an elite superforecaster. Produce a probability distribution over the answer to this Metaculus numeric question.

## Question
How many seats will the Ädilet party win in the 2026 Kazakh legislative election?

## Description / Resolution Criteria
## Description
On July 1, 2026, Kazakh President Kassym-Jomart Tokayev signed a decree [scheduling the upcoming elections](https://astanatimes.com/2026/07/kazakhstan-sets-august-23-date-for-kurultai-elections/) for the Kurultai, the unicameral legsilature of Kazakhstan, for August 23. This will be the first elections after the country's [new Constitution took effect](https://astanatimes.com/2026/07/kazakhstans-new-constitution-takes-effect-heres-what-changes/). The Kurultai will consist of 145 deputies elected through a nationwide proportional party-list system.

The [Ädilet ](https://thediplomat.com/2026/05/what-to-make-of-adilet-kazakhstans-newest-political-party/)[party is newly registered](https://thediplomat.com/2026/05/what-to-make-of-adilet-kazakhstans-newest-political-party/) and led by Aibek Dadebay, former head of President Tokayev's administration. The party's name means "justice" in Kazakh. In June, Kazakhstan's ruling party, [Amanat, merged with Ädilet](https://astanatimes.com/2026/06/kazakhstans-amanat-party-approves-merger-with-adilet/). Although [seven parties have registered](https://astanatimes.com/2026/07/kazakhstan-approves-election-timeline-for-kurultai-vote-seven-parties-cleared-to-participate/), it is expected that [the Ädilet party will gain a majority of seats](https://thediplomat.com/2026/06/with-amanat-merger-adilet-is-on-track-to-be-kazakhstans-next-ruling-party/), being seen as a "fresh political wave by an uncritical press", and as the one that supports the president.

`{"format": "metac_reveal_and_close_in_period", "info": {"post_id": 44920, "question_id": 45069}}`

## Resolution Criteria
This question will resolve to the total number of seats the Ädilet party win in the 2026 Kazakh legislative election, according to [The Central Election Commission of the Republic of Kazakhstan](https://www.election.gov.kz/eng/).

## Fine Print
If the 2026 Kazakh legislative election is not held before September 1, 2026, this question will be annulled.

***
This question's information (resolution criteria, fine print, background info, etc) is synced with an [original identical question](https://www.metaculus.com/questions/44920) which opened on 2026-08-04 10:48:31. This question will resolve based on the resolution criteria and fine print of the linked original question. However, if this question would resolve differently than the original question, then this question will be annulled. Additionally, if the original question's resolution could have been known before this question opened, then this question will be annulled.

## Range
The answer must be a number in [49.5, 145.5] (units: Seats).

## Sub-question decomposition (planner)
- (w=0.35) Will Ädilet win an outright majority of the 145 Kurultai seats (73 or more)?  — The central expectation is that the presidentially-aligned successor to Amanat dominates; establishing the probability o
- (w=0.25) Will Ädilet win 100 or more of the 145 seats (i.e. roughly 69%+ of seats)?  — Historical Kazakh ruling-party performance (Nur Otan/Amanat: 71–82% of party-list seats) suggests a very high share; thi
- (w=0.20) Will Ädilet's nationwide party-list vote share be below 60%?  — Under a pure proportional system with a 5% threshold, seat count scales nearly linearly with vote share; a sub-60% resul
- (w=0.20) Will four or more parties (besides Ädilet) clear the electoral threshold and win seats, diluting Ädilet's seat share?  — The number of parties clearing the threshold determines how much of the wasted-vote bonus accrues to Ädilet; more succes

## Synthesized evidence
1. [sq1 | question_description | STRONG cred 88 | UP | RECENT] Kazakhstan's ruling Amanat party merged with the newly registered Ädilet in June 2026, transferring the ruling-party machine to Ädilet.
2. [sq1 | article_search | MODERATE cred 75 | UP | RECENT] The Diplomat (June 2026) assessed that Ädilet is 'on track to be Kazakhstan's next ruling party' and is expected to win a majority of seats.
3. [sq1 | wikipedia_lookup | STRONG cred 90 | UP | DATED] Nur Otan/Amanat has held a majority of Mäjilis seats in every Kazakh legislative election since 2007.
4. [sq3 | code_execution | STRONG cred 88 | NEUTRAL | DATED] Historical ruling-party list vote shares: 2007 88.2%, 2012 80.9%, 2016 82.2%, 2021 71.4%, 2023 53.9% — a steady downward trend.
5. [sq3 | code_execution | MODERATE cred 88 | UP | DATED] The most recent comparable data point, Amanat's 2023 party-list share, was 53.9% — below the 60% threshold in sq3.
6. [sq1 | code_execution | MODERATE cred 45 | NEUTRAL | VERY_RECENT] Monte Carlo over plausible vote shares yields median 95 seats, 5th percentile 74, P(>=73)=0.96, P(>=100)=0.35.
7. [sq2 | code_execution | WEAK cred 40 | NEUTRAL | VERY_RECENT] Model puts ~35% probability on Ädilet reaching 100+ of 145 seats; the 25th–75th percentile range is 86–103 seats.
8. [sq4 | question_description | STRONG cred 90 | NEUTRAL | RECENT] Seven parties were cleared by the CEC to contest the August 23 Kurultai election under a nationwide proportional list system.
9. [sq4 | wikipedia_lookup | MODERATE cred 80 | UP | DATED] In the 2023 Mäjilis election six parties cleared the 5% threshold, meaning five parties besides the ruling party won list seats.
10. [sq1 | question_description | MODERATE cred 80 | UP | RECENT] Ädilet is led by Aibek Dadebay, former head of Tokayev's presidential administration, and is covered favorably by an uncritical domestic press.
11. [sq3 | question_description | MODERATE cred 85 | NEUTRAL | RECENT] The 2026 election is the first under a new Constitution with a 145-seat fully proportional Kurultai, removing single-mandate districts that previously boosted the ruling party's seat total.
12. [sq1 | article_search | WEAK cred 20 | NEUTRAL | VERY_RECENT] The article search returned only US-focused election stories and no Kazakhstan-specific campaign polling or reporting.

## Cross-Market Signals

### Polymarket
- "Will Ronaldo Caiado win the 2026 Brazilian presidential election?" → Yes: 0.02, Volume: $5.6M
- "Will Renan Santos win the 2026 Brazilian presidential election?" → Yes: 0.07, Volume: $9.0M
- "Will Michelle Bolsonaro win the 2026 Brazilian presidential election?" → Yes: 0.00, Volume: $9.6M
- "Will Luiz Inácio Lula da Silva win the 2026 Brazilian presidential election?" → Yes: 0.65, Volume: $8.1M
- "Will Tereza Cristina win the 2026 Brazilian presidential election?" → Yes: 0.00, Volume: $3.5M
- "Will Flávio Bolsonaro win the 2026 Brazilian presidential election?" → Yes: 0.27, Volume: $8.0M
- "Will Camilo Santana win the 2026 Brazilian presidential election?" → Yes: 0.00, Volume: $4.4M
- "Will Fernando Haddad win the 2026 Brazilian presidential election?" → Yes: 0.00, Volume: $6.7M

Information gaps:
  - No Kazakhstan-specific opinion polling or campaign reporting for the 2026 Kurultai race
  - Unclear electoral threshold under the new Constitution (5%? lower?) and whether 'against all' votes count
  - No detail on seat-allocation method or treatment of below-threshold wasted votes
  - Unknown which of the other six registered parties are viable (Auyl, Respublica, AQJOL, People's Party, NSDP successors)

Key uncertainties:
  - Whether the merged Ädilet inherits Amanat's ~54% 2023 share or a managed higher 'fresh wave' share
  - Degree to which authorities engineer a controlled multiparty parliament to dilute the ruling party
  - Wasted-vote inflation: how much below-threshold vote converts into extra Ädilet seats
  - Whether the election is actually held before September 1, 2026 (annulment risk)

## Required pre-forecast walkthrough

Before giving percentiles, address these explicitly in your rationale:
  (a) The time left until the question resolves.
  (b) The outcome if NOTHING changes from today (the status quo value).
  (c) The outcome if the CURRENT TREND continues.
  (d) The expectations of experts / markets / base rates.
  (e) A plausible scenario that produces a LOW outcome (near p10).
  (f) A plausible scenario that produces a HIGH outcome (near p90).

## Calibration guidance

- **Be humble about tails.** Good forecasters set WIDE 90/10 intervals to account for unknown unknowns. Narrow tails get punished by the log score far more than slightly-biased medians.
- **Status quo anchoring.** The p50 should be close to the status quo value unless you have strong evidence of a trend.
- Don't pile mass at one value — if you're tempted, widen the spread by 20-50%.
- **Anchor on markets/experts.** If liquid market prices, analyst forecasts, or community percentiles appear in the evidence, center your distribution on them and widen — don't override a liquid market without specific evidence it lacks.
- **Relative-return / spread questions ("how much will X's return exceed Y's").** A near-zero median is usually right, but size the TAILS to the more VOLATILE leg, not to a generic 2-3pp spread. Two broad equity indices (e.g. Nasdaq-100 vs S&P 500) do stay within roughly ±2-3pp over a two-week window. But when one leg is a commodity (crude oil, gold) or a single high-beta stock (e.g. Nvidia), the two-week realized spread regularly reaches ±10pp or more — crude-vs-equity spreads have hit ±18pp. In those cases set p10/p90 to at least ±10pp (±15pp or wider for oil) while keeping the median near the status-quo spread; the 2-3pp intuition applies only to index-vs-index.
- **Heavy-tailed one-off magnitudes (opening-weekend box office, product-launch or debut numbers).** Pre-event tracking for these routinely misses by ±40% or more, and results land both far above and far below the consensus. Do NOT treat the displayed [min, max] range as the plausible range or clamp p10/p90 to its edges: the outcome can fall outside the range in either direction (recent opening weekends resolved ~7% above a displayed ceiling in one case and roughly half the displayed floor in another). Set p10/p90 wide enough that a 2x-or-0.5x surprise versus your central estimate still lands inside the interval, and when concrete current tracking disagrees with historical/franchise base rates, weight the tracking.

## Output

Return ONLY valid JSON, no markdown fences:

{
  "rationale": "<address (a) through (f) above, then state your distribution — 6-10 sentences total>",
  "p10": <number>,
  "p25": <number>,
  "p50": <number>,
  "p75": <number>,
  "p90": <number>
}

Percentiles MUST be in non-decreasing order (p10 <= p25 <= p50 <= p75 <= p90).