Liquidity Arena 2026 Track A, Phase I (41 teams)

0 of 41 reach t = 22026-10-12Sources public, fetched on the stated date

Finding

No team reaches t = 2; the highest is 1.63. The format is closed-book; the window is too short.

The subject is the whole leaderboard, not any one team.

Subject

Phase I of LTP's AI quant trading competition: a sandbox elimination round, 2026-07-20 to 08-21, 41 teams, official rankings and equity curves.

Three inputs

Input	Value
Knowledge cutoff	Not applicable. Live, forward-looking trading is closed-book by construction; that is a strength of the format.
Window	25 to 31 return days per team (median 30).
Attempts	41 (teams in the round); the winner is the best of 41.

Finding

None of the 41 results is statistically distinguishable from luck.

Method: daily closing equity from each team's equityCurve → daily returns → annualised Sharpe (× √365) → t = Sharpe × √(days / 365).
Teams with t ≥ 2: 0 of 41. The highest t is 1.63 (rank 11). The top three are about 1.42, 1.40 and 1.20.
Picking the best of 41 raises the bar from t = 2 to t ≈ 3.26 (Bonferroni, one-sided; selection_penalty(41)).

Rank	Team	Sharpe (daily equity)	Official Sharpe	Days	t
1	Little J	4.87	3.28	31	1.42
2	X-Explore	5.23	3.20	26	1.40
3	Quantech Services Group	4.13	3.76	31	1.20
11	Kinance (highest t)	5.58	2.50	31	1.63

The official Sharpe's method is not stated, so it differs from ours. The finding holds under either: plugging the official values into 33 days gives a highest t of about 1.46.

What it would take

True Sharpe	Evidence needed, one attempt	Evidence needed, best of 41
5	about 2 months	about 5 months
2	about 12 months	about 32 months

What this does and does not say

It says a one-month ranking mostly reflects luck and risk appetite.
It does not say any team's strategy fails. The window is too short to show either.

Source

Official ranking endpoint (fetched 2026-10-11; snapshot at the end of Phase I). A suggestion to show t-statistics on the leaderboard was sent to the organiser on 2026-10-12.