Method leaderboard

Every prediction method we run, scored on real draws — including a pure-random control arm. This is the table no tipster site would ever publish.

How to read this: 47 draw nights scored so far, averaging 6.0 operator boards a night. A number with four different digits should collect about 0.32 box hits a night by pure chance; a number with a repeated digit has fewer arrangements, so it collects fewer. Each method is therefore measured against the chance line of the numbers it actually picked — that is the "vs chance" column. Straight = exact match; box = any other arrangement of the digits.

Scoring change, 3 Oct 2026: until now every method was measured against the chance line of a number with four different digits. Numbers with a repeated digit have fewer arrangements and get fewer box hits by luck, so methods that pick them — including the random control — were made to look worse than chance. Each method is now measured against the numbers it actually picked. Scoring is recomputed from the stored picks, so the whole record above changed at once.

The table is ordered by the "vs chance" column — how far each method sits from the chance line of its own picks — not by raw box hits per pick, which would rank a method higher just for picking numbers with more arrangements. A discontinued method no longer locks picks: its record is kept, listed last and left out of the method counts.

When the way a method picks its numbers changes, its record starts again from the first draw picked the new way — a verdict should describe one method, not two. The earlier version keeps its own row at the bottom, with no verdict, and is left out of the method counts.

With 24 methods each tested at a 5% false-alarm rate, about 1.2 would show "Edge" by luck alone. A single Edge is what luck predicts; treat it as a lead to watch, not proof.

MethodPicksStraightBoxBox/pickvs chanceReturn1B 1S iBox1B 1S straight
AnniversaryCollecting1142350.31+0.0896%86%9%
Cold endingsCollecting1140350.31+0.0773%76%0%
Weighted ensembleCollecting1201450.38+0.0688%92%4%
Own-cycleCollecting852230.27+0.0591%89%165%
PositionalCollecting1293230.18+0.0593%99%395%
FollowerCollecting1202330.28+0.0494%95%25%
AI (Claude)Collecting1142340.30+0.0457%48%18%
Profilesince 7 Oct 2026Collecting570200.35+0.0358%50%0%
Jiransince 7 Oct 2026Collecting570140.25+0.0333%17%0%
Trend-builtCollecting1354260.19+0.02102%108%756%
DueCollecting1413390.28+0.0138%26%11%
Near missCollecting1200300.25+0.0181%83%0%
In formChance1141280.25-0.0159%54%4%
RhythmChance1202270.23-0.0175%77%25%
Lucky 8Collecting57060.11-0.0229%27%0%
CompanionsChance1355290.21-0.02105%109%400%
Random controlcontrolChance1200240.20-0.03107%117%0%
HotChance1414300.21-0.0380%88%379%
EndingChance954200.21-0.0469%68%168%
Head pairsChance1142210.18-0.0543%46%224%
CoolingChance1142220.19-0.0673%79%26%
ML modelChance1201240.20-0.0830%27%4%
Cold numbersChance571100.18-0.1051%47%26%
ConsensusCollecting18020.11-0.1121%11%0%
Repeat watchsince 7 Oct 2026Chance57170.12-0.1122%11%26%
Profilebefore 7 Oct 2026Earlier version781400.51+0.20145%155%6%
Jiranbefore 7 Oct 2026Earlier version690150.22-0.0127%19%0%
DateDiscontinued11020.18-0.059%5%0%
Repeat watchbefore 7 Oct 2026Earlier version780130.17-0.0517%9%0%

The return column

What RM1 iBox Big on every pick, on every board, would actually have returned — priced with the real dividend table. Counting box hits treats a consolation and a 1st prize as the same event; this does not. A fair draw returns about 67%, and 100% is break-even.

The two 1B 1S columns price RM1 Big + RM1 Small on every pick, on every board. Straight only pays the exact number, so most methods sit at 0%. A single straight 1st prize pays RM6,000 (RM2,500 Big + RM3,500 Small) against roughly RM900 staked over this record — that one night alone shows as about 650%. The fair long-run return for this bet is about 64%; treat any straight figure far above it as one lucky night, not a method that works.

Read this column last. A single RM1 bet has an average return of 67 sen but a standard deviation of RM5.75, because 38% of all expected return sits in the 1st prize alone — an event with a 0.24% chance. That makes return roughly 5 times noisier than the box-hit rate and in need of about 25 times more data to mean anything, which is why the verdict is decided on hits and the table is not sorted on money. For scale: the pure-random control arm is currently returning 107%.

Read this before reading the table. With only 47 nights, every gap in this table is statistical noise — a method "leading" today will regress. If any method were consistently and significantly ahead of the random control over hundreds of nights, that would be extraordinary (and we would say so loudly). Until then: this page exists to show they are all the same, honestly. The random control is currently collecting 0.20 box hits per pick — any method below it is doing worse than throwing darts.

From 20 Aug 2026 every row here is scored on the Malaysian-licensed boards only (Magnum, Toto, Da Ma Cai, Sabah 88, STC, CashSweep) — the same six the picks are built from. Fewer boards per night means fewer chances to hit, so both the methods and the random control sit lower than they did before, and the whole history is recounted on this basis.

How we scoreThe statistics behind each verdict, and the rules for retiring a method›

How the verdict is decided (and why it isn't decided yet)

Re-checking a leaderboard every night and calling the best row a winner is how false discoveries are made — repeated peeking inflates the error rate. The verdict column instead runs Wald's sequential probability ratio test per method: chance-level box rate vs a 1.5× edge, with 5% error caps that hold no matter how often we look. A method is only called an Edge or confirmed as Chance when the accumulated evidence crosses the threshold; until then it honestly says Collecting.

For scale: a method performing exactly at chance needs roughly 135 picks before the test can confirm it — verdicts here are earned slowly by design.

Method retirement policy

Pre-committed

A method is retired from the nightly picks when its SPRT verdict is "chance" AND it has at least 300 scored picks. Retired methods keep their full record on this leaderboard, marked retired. Any change to the line-up is announced here with its date. The random control is never retired — a control removed for behaving like a control is not a control.

Status right now: no method meets this rule. That is the whole point of publishing it today — a rule written after a method fails is not a rule, it is an excuse.

Why a pick floor as well as a verdict: the SPRT can confirm "chance" fairly early on a method that simply had a quiet run. The floor makes sure nothing is benched on a short record, and keeping the retired rows visible means nothing quietly disappears from the scoreboard once it stops being flattering.

Changed 12 Nov 2026: the pick floor was lowered from 1,000 to 300. No method met either number on that day — the longest record was 135 picks — so this is still a rule set before anything reached it. The reason: 1,000 picks is more than two years of draws, and a method that really is at chance is usually confirmed within 100 to 160 picks.

Relegation & promotion

The softer stage before retirement: a MAIN method (hot, overdue, own-cycle, ending) whose verdict is "chance" with at least 150 scored picks loses its main-pick slot and continues in the deep set — still locked, still scored, still on this board. Its slot goes to the deep method with the highest ensemble trust weight and at least 150 picks of its own. The random control and the ensemble itself are never eligible. Swaps happen only when this rule fires, and are announced here with their date.

Status right now: no main method meets the relegation rule. Published today, before any does.

Are “hot numbers” real? The shrinkage test

Empirical Bayes on the last 365 days of top-3 hits, fitted across all 10,000 numbers.

7.2% of the spread between numbers exceeds what Poisson noise explains — worth watching, but see the fairness audit before reading anything into it.
NumberHits (365d)Honest estimateExcess kept
643761.087%
767561.087%
237551.017%
315351.017%
326451.017%
425151.017%
762351.017%
787451.017%
804351.017%
813451.017%

“Honest estimate” is the posterior expected hit count for the NEXT 365 days, after shrinking toward the all-number average of 0.70. “Excess kept” shows how much of a number's above-average performance the math believes; near 0% means its hot streak was luck.