Everyone starts on R1 000 of play money. Nothing you deposit, nothing you can lose, nothing to pay. What is at stake is your record: three boards, each measuring something different, each published and permanent.
101days left, the season closes on Sunday 15 November 2026, election week. Voting is 07 November.
Three ways to win
Three different things are ranked, because three different things need proving: whether the prices carry information, whether anyone can beat them, and whether the parts of the board nobody else offers actually get traded. You can top more than one.
The Forecaster
Accuracy, ranked on Brier score
The player whose prices were closest to what actually happened. Not who won the most, but who was least wrong, most often.
The formula
Brier = (1/n) × Σ (pᵢ − oᵢ)²
pᵢthe price you paid, as a probability that YES happens. Pay 30c for YES and p = 0.30. Pay 30c for NO and p = 0.70, because that is what your money says about YES.
oᵢwhat happened. 1 if the market resolved YES, 0 if NO.
nhow many of your positions have settled.
Lower is better. 0.000 is perfect foresight, 0.250 is a coin flip, 1.000 is being confidently wrong every single time.
Worked example
Say you settle three calls.
Your call
p
Result
o
(p − o)²
Vaal Dam over 100% YES at 80c
0.80
YES
1
0.04
PowerBall jackpot hit, YES at 25c
0.25
NO
0
0.0625
Boks beat the All Blacks, YES at 90c
0.90
NO
0
0.81
Brier = (0.04 + 0.0625 + 0.81) ÷ 3 = 0.304
The two sensible calls cost you 0.10 between them. The one confident mistake cost 0.81 on its own, eight times as much as both the others combined. That is the entire design: you are not punished for being wrong, you are punished for being sure and wrong.
Why this measure
Brier punishes confident wrongness far harder than cautious wrongness. Backing a 95c favourite that loses costs you roughly four times what a 55c call costs. That is deliberate: it is the only score here that cannot be farmed by buying longshots and getting lucky once.
Ties: Most settled positions. If still level, the earlier qualifying date.
Leading now, lowest wins
Nothing has settled yet. Standings appear as markets resolve, from the very first one.
The Sharpest
Skill, ranked on return per rand risked
The player who beat the price by the widest margin for every rand they put at risk.
The formula
Edge = (returned − staked) ÷ staked
returned100c for every contract you held on the winning side, across all your settled positions.
stakedeverything you paid for those positions, win or lose.
Higher is better, shown as a percentage. +0% means you exactly matched the market. +20% means every R100 you risked came back as R120.
Worked example
Say you settle three positions of 100 contracts each.
Your call
Staked
Result
Returned
YES at 40c × 100
R40.00
YES, won
R100.00
NO at 60c × 100
R60.00
YES, lost
R0.00
YES at 50c × 100
R50.00
YES, won
R100.00
Edge = (R200.00 − R150.00) ÷ R150.00 = +33.3%
You risked R150 and got R200 back, so every R100 at risk returned R133. Note that you lost one call out of three and still scored well, edge does not care how often you are right, only whether the prices you paid were too cheap.
Why this measure
Edge normalises for bankroll and for volume, so it measures how far you beat the price you traded at rather than how much you traded. It is the closest thing to a professional's number, and it ranks finding mispriced questions above backing favourites.
Ties: More rands staked. A wide edge on a large book beats the same edge on a small one.
Leading now, highest wins
Nothing has settled yet. Standings appear as markets resolve, from the very first one.
The Regular
Breadth, ranked on range across the catalogue
The player who traded across the widest slice of the board. This one counts range, not accuracy, the tie-break is what stops it rewarding being wrong everywhere.
The formula
Breadth = Σ min(nc, 10)
ncyour settled positions in category c, money, sport, the vote, utilities, the draw, screens.
min(..., 10)each category counts up to 10 positions and no more.
Higher is better. The cap means 10 positions in one category scores the same as 30, so spreading across six categories beats hammering one.
Worked example
Two players, thirty settled calls each.
Thandi
counts
Sipho
counts
money
6
6
0
0
sport
5
5
30
10
the vote
7
7
0
0
utilities
4
4
0
0
the draw
5
5
0
0
screens
3
3
0
0
Thandi = 30 across 6 categories · Sipho = 10 across 1
Identical effort, triple the score. Sipho's last twenty rugby calls earned nothing at all, because the cap stopped counting at ten, and he fails the four-category rule besides. This board is for range, and range is the thing this season exists to measure.
Why this measure
This is the board that exists for the platform rather than for the player. Everyone already knows South Africans will bet on rugby. Nobody knows whether they will price a dam level or an inflation print, and that question is the whole reason this season is running.
Ties: Better Brier score. Range should not pay for being wrong everywhere.
Also: Also needs settled positions in at least 4 different categories.
Leading now, highest wins
Nothing has settled yet. Standings appear as markets resolve, from the very first one.
Every formula above is defined in full, with its limitations, on the methodology page.
The rules
Free, and there is no prize money. Nothing to pay, no way to add money, and nothing to withdraw. The standings are the whole reward.
You appear from your first settled question. Under 15 settled you are marked provisional: still ranked, still visible, but the score is thin and cannot take a final standing. Thin scores are ranked closer to the middle on purpose, so one lucky call cannot top a board.
One account per person. Multiple accounts are disqualified, and on a calibration board they would not help you anyway.
No board ranks on raw profit. Where you finish does not depend on how much play money you end with, so there is no reason to gamble your balance on a longshot at the end.
Voided markets do not count toward anything. Every stake in them is returned.
Scores are computed from the same code that runs the leaderboard, on settled markets only, and recomputed every time a market resolves.
Final standings published within 14 days of the season closing, alongside the full results.
Why this exists
To find out whether South Africans will price questions nobody currently offers, dam levels, inflation prints, council coalitions, and whether the resulting prices are any good. Your calls are the experiment. The results get published at the end of the season.
Play money only. No deposits, no withdrawals, no prize money, no stake at risk. Nobody is ever asked for a cent.
Contact us
A person reads every one of these. Tell us what happened and we will come back to you.