Every revision to the grading method, dated, with what changed and why. A grade means little if the rubric moves silently underneath it, so each change is kept here in full rather than folded quietly into the next version of the page that describes it.
2026
25 September: what Preliminary means
Preliminary now has one published meaning, decided on 22 September 2026: the grade is based on the firm’s contracts, published rules and prices, read firsthand, and does not yet include a payout test. Before this, our pages described Preliminary as a grade computed from part of the rubric. The change is to the label’s wording, and no letter moved.
16 September: the cut points are ratified
The cut points that turn a score out of 100 into a letter were ratified on 16 September 2026. The table sets out the score each letter starts from, and the score below which a firm receives an F.
| Letter | Score | Ratified |
|---|---|---|
| A+ | From 95 | 16 September 2026 |
| A | From 87 | 16 September 2026 |
| A- | From 79 | 16 September 2026 |
| B+ | From 77 | 16 September 2026 |
| B | From 74 | 16 September 2026 |
| B- | From 73 | 16 September 2026 |
| C+ | From 67 | 16 September 2026 |
| C | From 57 | 16 September 2026 |
| C- | From 51 | 16 September 2026 |
| D+ | From 45 | 16 September 2026 |
| D | From 36 | 16 September 2026 |
| D- | From 27 | 16 September 2026 |
| F | Below 27 | 16 September 2026 |
They were set against the real distribution of all 36 scored firms rather than guessed in advance, and setting them moved no published letter. The full ladder is set out in what each letter means.
1 September: the six weights are settled
The rubric weights were ratified by Krishi Chowdhary, the external grading expert, and recorded as ruling R12. Payout reliability 25, rule fairness 20, trading conditions 20, company trust 15, challenge value 15, support and disputes 5.
Two of those moved. Trading conditions rose from 15 to 20, split ten to drawdown and ten to platform quality. The reasoning was not that the drawdown model matters less: drawdown already carries a separate recognition through rule fairness at 20, so weighting it twice as heavily again inside trading conditions would score one property twice. Execution quality was judged to carry economic weight of its own, because poor spreads and fills widen every loss and clip every win.
Support and disputes fell from 10 to 5, with contract-level dispute machinery scoring under company trust instead. Those are one decision rather than two. Fifty-seven of the 170 contract findings across the firms we cover are dispute or remedy machinery, and seven of those are critical. Halving support and disputes would have halved the weight of a third of the evidence, had it been scored there. It belongs to the question a trader is actually asking, which is whether a firm will seize on a term to avoid paying. Support and disputes at 5 is responsiveness and ticket quality only.
Nothing had been scored when the decision was taken: 44 assessment packages, 264 dimension rows, every score null. The change cost no rescoring and corrected no published grade.