Datasets:

harshbhx
/

gavel-sim-30d

hold_id string	lot_id string	bidder string	ceiling_cents int64	opened_day float64	captured_cents int64	released_cents int64	authorizations_in_chain int64
hold-soak01-b054	soak01	b054	27,500	13.8663	0	27,500	2
hold-soak01-b066	soak01	b066	44,500	14.1262	0	44,500	2
hold-soak01-b087	soak01	b087	44,000	14.3783	0	44,000	2
hold-soak01-b091	soak01	b091	59,500	16.0512	0	59,500	1
hold-soak01-b073	soak01	b073	76,500	16.1625	0	76,500	1
hold-soak01-b028	soak01	b028	98,000	16.2076	97,500	500	1
hold-soak01-b042	soak01	b042	38,000	16.7883	0	38,000	1
hold-soak01-b064	soak01	b064	30,000	18.0378	0	30,000	1
hold-soak01-b071	soak01	b071	38,000	18.0378	0	38,000	1
hold-soak02-b083	soak02	b083	89,500	1.2496	89,000	500	2
hold-soak02-b008	soak02	b008	21,500	2.1513	0	21,500	2
hold-soak02-b048	soak02	b048	26,000	2.2716	0	26,000	2
hold-soak02-b040	soak02	b040	45,000	2.3155	0	45,000	2
hold-soak02-b036	soak02	b036	60,000	3.0992	0	60,000	1
hold-soak02-b096	soak02	b096	46,500	3.682	0	46,500	1
hold-soak02-b058	soak02	b058	15,500	3.8812	0	15,500	1
hold-soak02-b003	soak02	b003	66,000	4.8436	0	66,000	1
hold-soak02-b070	soak02	b070	86,000	4.8436	0	86,000	1
hold-soak03-b043	soak03	b043	11,500	21.8442	0	11,500	2
hold-soak03-b096	soak03	b096	41,000	22.0877	0	41,000	2
hold-soak03-b037	soak03	b037	18,500	22.2527	0	18,500	2
hold-soak03-b063	soak03	b063	65,000	22.2993	0	65,000	2
hold-soak03-b008	soak03	b008	79,000	22.5863	78,500	500	2
hold-soak03-b028	soak03	b028	63,500	22.6552	0	63,500	2
hold-soak03-b019	soak03	b019	52,000	23.5095	0	52,000	1
hold-soak03-b002	soak03	b002	67,500	23.7345	0	67,500	1
hold-soak04-b085	soak04	b085	96,000	19.8477	0	96,000	1
hold-soak04-b047	soak04	b047	98,000	21.0606	97,500	500	1
hold-soak04-b031	soak04	b031	20,000	21.4619	0	20,000	1
hold-soak04-b020	soak04	b020	34,000	21.563	0	34,000	1
hold-soak05-b012	soak05	b012	23,500	14.904	0	23,500	2
hold-soak05-b031	soak05	b031	73,500	14.904	0	73,500	2
hold-soak05-b040	soak05	b040	77,500	16.3029	77,000	500	2
hold-soak05-b029	soak05	b029	73,000	17.096	0	73,000	1
hold-soak05-b075	soak05	b075	28,500	18.0378	0	28,500	1
hold-soak05-b065	soak05	b065	13,000	18.5138	0	13,000	1
hold-soak06-b081	soak06	b081	66,500	1.1128	0	66,500	1
hold-soak06-b024	soak06	b024	37,500	2.4286	37,500	0	1
hold-soak06-b050	soak06	b050	34,000	2.7348	0	34,000	1
hold-soak06-b021	soak06	b021	32,500	2.9833	0	32,500	1
hold-soak07-b058	soak07	b058	47,000	12.7314	0	47,000	1
hold-soak07-b059	soak07	b059	22,000	13.7143	0	22,000	1
hold-soak07-b026	soak07	b026	39,000	13.9041	0	39,000	1
hold-soak07-b075	soak07	b075	28,000	13.9602	0	28,000	1
hold-soak07-b088	soak07	b088	33,500	14.3103	0	33,500	1
hold-soak07-b053	soak07	b053	79,500	14.904	79,000	500	1
hold-soak07-b087	soak07	b087	48,500	14.904	0	48,500	1
hold-soak08-b016	soak08	b016	71,000	9.7306	0	71,000	2
hold-soak08-b044	soak08	b044	95,000	10.6391	94,500	500	2
hold-soak08-b045	soak08	b045	23,000	11.1466	0	23,000	2
hold-soak08-b049	soak08	b049	47,500	11.4236	0	47,500	2
hold-soak08-b075	soak08	b075	34,000	11.4958	0	34,000	1
hold-soak08-b086	soak08	b086	89,000	11.5605	0	89,000	1
hold-soak08-b003	soak08	b003	65,000	11.7425	0	65,000	1
hold-soak08-b080	soak08	b080	29,000	12.46	0	29,000	1
hold-soak08-b072	soak08	b072	11,500	13.3331	0	11,500	1
hold-soak09-b015	soak09	b015	50,500	18.4244	0	50,500	1
hold-soak09-b037	soak09	b037	52,000	18.4807	0	52,000	1
hold-soak09-b018	soak09	b018	71,000	18.6001	0	71,000	1
hold-soak09-b040	soak09	b040	72,000	19.0341	71,500	500	1
hold-soak09-b085	soak09	b085	48,500	19.4612	0	48,500	1
hold-soak09-b036	soak09	b036	70,000	20.1441	0	70,000	1
hold-soak10-b011	soak10	b011	11,500	18.3857	0	11,500	1
hold-soak10-b094	soak10	b094	51,000	18.5289	0	51,000	1
hold-soak10-b054	soak10	b054	44,000	18.8438	0	44,000	1
hold-soak10-b071	soak10	b071	85,000	18.844	0	85,000	1
hold-soak10-b057	soak10	b057	35,500	19.4612	0	35,500	1
hold-soak10-b043	soak10	b043	91,000	19.4612	90,500	500	1
hold-soak10-b004	soak10	b004	34,000	19.4612	0	34,000	1
hold-soak11-b012	soak11	b012	94,000	8.0547	0	94,000	2
hold-soak11-b090	soak11	b090	49,500	8.3094	0	49,500	2
hold-soak11-b013	soak11	b013	75,500	8.4353	0	75,500	2
hold-soak11-b015	soak11	b015	60,500	9.1997	0	60,500	2
hold-soak11-b025	soak11	b025	77,500	9.6174	0	77,500	2
hold-soak11-b010	soak11	b010	71,000	10.0912	0	71,000	1
hold-soak11-b086	soak11	b086	30,500	10.1209	0	30,500	1
hold-soak11-b038	soak11	b038	22,500	10.5764	0	22,500	1
hold-soak11-b073	soak11	b073	98,500	11.1466	98,000	500	1
hold-soak11-b082	soak11	b082	96,000	11.1613	0	96,000	1
hold-soak12-b089	soak12	b089	86,000	11.6715	85,500	500	2
hold-soak12-b005	soak12	b005	26,500	12.6498	0	26,500	1
hold-soak12-b083	soak12	b083	40,000	12.7257	0	40,000	1
hold-soak12-b016	soak12	b016	50,500	12.7326	0	50,500	1
hold-soak12-b099	soak12	b099	46,000	12.8534	0	46,000	1
hold-soak12-b082	soak12	b082	41,000	13.0594	0	41,000	1
hold-soak12-b032	soak12	b032	20,000	13.6196	0	20,000	1
hold-soak12-b047	soak12	b047	41,000	14.1147	0	41,000	1
hold-soak12-b021	soak12	b021	74,000	14.3106	0	74,000	1
hold-soak12-b100	soak12	b100	71,500	14.904	0	71,500	1
hold-soak13-b062	soak13	b062	96,500	15.9567	0	96,500	1
hold-soak13-b012	soak13	b012	23,500	16.0277	0	23,500	1
hold-soak13-b039	soak13	b039	55,500	18.1256	55,500	0	1
hold-soak13-b095	soak13	b095	48,500	18.3531	0	48,500	1
hold-soak13-b037	soak13	b037	19,000	18.6029	0	19,000	1
hold-soak14-b083	soak14	b083	36,000	8.485	0	36,000	1
hold-soak14-b015	soak14	b015	34,500	8.6076	0	34,500	1
hold-soak14-b075	soak14	b075	68,500	8.6654	0	68,500	1
hold-soak14-b069	soak14	b069	68,500	8.6686	0	68,500	1
hold-soak14-b087	soak14	b087	92,000	9.0624	91,500	500	1
hold-soak14-b008	soak14	b008	78,500	9.4813	0	78,500	1

End of preview. Expand in Data Studio

gavel-sim-30d

Honest labels: simulated, synthetic. Every row comes from a simulation with a fake clock against a model of PayPal, not from PayPal and not from real buyers.

What it is

The settlement of 40 simulated auction lots over 30 simulated days, with 100 fake bidders, random process crashes and injected payment faults. Per lot: how many holds were placed, who won and at what price, whether it settled or went to review and why, and where every cent of every hold ended up (captured, released or still open). One row per hold in a second table, and a table of the eight simulation scenarios with their check counts.

It shows that a card-hold-based auction settlement keeps its books balanced (no double capture, no lost hold, held equals captured plus released plus open) under crashes, restarts, outages and rate limiting.

How it was made

Produced by the sim-multiday suite (packages/payments/test/simdays, scenario S8, seed 20261107) which runs the real settlement code (saga, reauthorization pass, journals, webhook merge) against the fake PayPal in packages/payments/src/fake. A fake clock starts at 2026-11-01T00:00:00.000Z; days are simulated days. Nothing waits in real time.
The fake follows these assumptions: authorization honor period 3 days, validity 29 days, reauthorization allowed only after the honor period. They are assumptions about PayPal, not measurements; the repository documents which ones the sandbox has confirmed.
Every fifth lot ("storm") gets two injected HTTP 429 answers on its first capture and a restart after a 2-hour outage. Other lots get random faults (declined capture, timeouts, 5xx, connection resets), a 3 percent chance of a crash at each saga boundary and, in 5 percent of crashes, an outage of 1 to 4 days.
The export is test-only code in the suite (SIM_EXPORT_DIR). The same run is reproducible: two runs give identical rows.

Provenance

Built from commit a72767848d84b0d166f5b48f4ffdbce4efef49e6 of https://github.com/harshbhx/gavel-auction by https://github.com/harshbhx/gavel-auction/actions/runs/42724763623. The same values are in MANIFEST.json.

MANIFEST.json lists the sha256 and size of every file and the row count of every table. python verify.py recomputes the hashes and fails on any difference.

Re-run workflow sim-multiday (.github/workflows/sim-multiday.yml) with the repository at the commit above and SIM_EXPORT_DIR set; the suite is deterministic (it asserts that two runs are equal), so lots.jsonl and holds.jsonl come out byte for byte the same.
docs/evidence/SIM/SIM-MULTIDAY.md in the repository is the human-readable report of the same suite and is checked by CI against the generated one.

Fields

lots.jsonl / .csv

One row per lot (40 rows).

Field	Type	Meaning
`scenario`	string	Always `S8`, the 30-day soak scenario
`lot_id`	string	Fake lot id
`close_day`	float	Simulated day (since day 0) when the auction closed
`holds`	integer	Holds placed on the lot
`storm`	boolean	Lot got the injected 429 storm on its first capture
`status`	string	`settled` or `review` (ended in the review state)
`journal_state`	string	Final state of the settlement journal
`sold`	boolean	A winner was charged
`winner_bidder`	string or null	Fake bidder id of the winner
`winner_cents`	integer or null	Winning price in integer cents
`runner_up_fallback`	boolean	The first winner failed and a runner-up was tried
`crashes`	integer	Simulated process crashes survived
`passes`	integer	Saga passes needed
`down_days`	float	Total simulated downtime in days
`review_cause`	string or null	Why the lot is in review; null when settled
`faults`	string	Injected faults as `call:kind`, separated by semicolons
`held_cents`	integer	Sum of the hold ceilings
`captured_cents`	integer	Cents captured
`released_cents`	integer	Cents released back (voided or expired)
`open_cents`	integer	Cents still held after the final expiry sweep

holds.jsonl / .csv

One row per hold.

Field	Type	Meaning
`hold_id`	string	Fake hold id
`lot_id`	string	Lot of the hold
`bidder`	string	Fake bidder id, `b001` to `b100`
`ceiling_cents`	integer	Amount held, in cents
`opened_day`	float	Simulated day the hold was opened
`captured_cents`	integer	Cents captured from this hold
`released_cents`	integer	Cents released
`open_cents`	integer	Cents still held
`authorizations_in_chain`	integer	Authorizations in the reauthorization chain (1 means never reauthorized)

scenarios.jsonl / .csv

One row per simulation scenario S1 to S8.

Field	Type	Meaning
`scenario`	string	Scenario id
`title`	string	What the scenario does
`checks`	integer	Checks in the scenario
`checks_passed`	integer	Checks that passed

run.json

Run constants and totals.

Field	Type	Meaning
`seed, start_iso and totals`	mixed	Seed, fake-clock start and the counters the suite reports

Intended use

Studying settlement and reconciliation behaviour, reviewing failure causes, testing reconciliation or anomaly-detection code, and as a worked example of an invariant-checked payment saga.

Limitations

Simulated. The payment provider is a model written from assumptions; real PayPal may behave differently, which is why the repository also runs a real-sandbox probe.
40 lots and one seed; no real buyers, no real money, no real names. Bidder and lot ids are fake labels.
Faults and crash rates are chosen by the author, not measured from production, so the share of lots in review is not a failure rate of any real system.
The clock is simulated; durations in days say nothing about real latency.

License

CC-BY-4.0 (Creative Commons Attribution 4.0 International). You may copy, adapt and redistribute the data, including commercially, if you credit the project (see Citation). The licence covers the data files in this dataset only; it grants no right to the marks or services of PayPal, Deepgram, Groq or any other third party named in it.

Simulated or real

Labels: simulated, synthetic. Read the label note at the top of this card before using any number. Nothing here is production data, and no real person or real money is involved.

Citation

@dataset{gavel_gavel_sim_30d_2026,
  title  = {gavel-sim-30d},
  author = {Bhanushali, Harsh},
  year   = {2026},
  url    = {https://huggingface.co/datasets/harshbhx/gavel-sim-30d},
  note   = {Generated from https://github.com/harshbhx/gavel-auction}
}

Contact

Open an issue at https://github.com/harshbhx/gavel-auction/issues. The code that builds this dataset is in the scripts/hf folder of that repository.