Much of the track record you are shown for a QIS index is simulated. The useful question is not how the simulation performed. It is whether the simulation was measuring the same thing the live index measures.
In one rulebook we read, it was not. The strategy earns its return from the gap between volatility at two different times of day, so how those two prices are sourced is the entire economics of it. Both definitions changed inside the simulated period: the source moved from official settlement to a bid-ask midpoint, and an early-close rule was added several years later. Ten and a half years of that history is simulated; under one year is live.
The simulation is therefore measuring a different index, and a performance screen will not catch that, because the series looks continuous and the statistics look fine. It is visible only in the document.
We read index rules for a living, and we score them on seven things in two groups: what the sponsor can still change after you buy, and how well the thing was built and disclosed. We never average the two, because a document can be excellent on one and poor on the other.
Documents argue with themselves
Risk factors routinely assert that the sponsor has no discretion, and then the operative clauses grant it.
One rulebook requires the administrator to publish a concise explanation of every use of expert judgement. Elsewhere, the same document says it has no obligation to inform any person of the result of any action taken on the occurrence of the very events in which that judgement is exercised.
Another manages the trick inside a single sentence, permitting calculations "in its sole, absolute and unfettered discretion, but in good faith and in a commercially reasonable manner". Those are not the same standard. One of them is doing no work.
Neither finding needs legal training. It needs reading the definitions section as carefully as the methodology section, which is where discretion usually lives and where a single pass almost never goes.
Backtests get calibrated on their own history
A third document sets the strategy's leverage from a volatility figure defined as the in-sample standard deviation of sub-index returns, computed up to a date sitting well inside the backtest.
It is stated openly, in a table, as a numeral, so nothing is hidden. It is simply on nobody's checklist. The effect is that the leverage the simulated track record ran at was chosen with knowledge of the returns it was applied to.
Good drafting exists, and it looks like this
The reading only earns its keep if it discriminates, so the tight documents get the same scrutiny.
One names its optimiser down to the version number, states its two non-default parameters, and publishes a twenty-five step constraint relaxation ladder with every relaxed value and a determinate outcome at the end of it. There is no room left for interpretation anywhere in it.
Another requires the effective date of every amendment to be stated in the new version. That one clause makes it the only index in our files that carries its own change history on its face, and it does more for a monitoring process than any amount of narrative assurance.
Index rules do not show you the fee
Swap confirmations and note terms are a different document set, and we do not hold them when we do this work. Fee and spread mechanics, early termination rights, hedging disruption remedies and what happens to a holder when an index is cancelled usually sit there rather than in the rules.
That produces an asymmetry worth carrying around. A wide amendment right is real evidence, because it is there in the document you read. A narrow one tells you almost nothing, because a provider can look constrained when its latitude has simply been relocated to a document you were not sent. A rulebook that is unusually quiet on amendment, fees or notice is a reason to ask for the confirmation rather than to award marks.
The same relocation happens inside the index documentation, and that version you can see. Several rulebooks incorporate a master document or a calculation module by reference, amendable from time to time, sitting one tier above the client's copy. An amendment upstream changes the index without a single word changing in the document the buyer holds, and in every instance we have found, the sponsor owed no notice.
Four things to check in the next document you read
Read the definitions section as carefully as the methodology. Discretion is granted inside defined terms far more often than under a heading that announces it.
Look for the document arguing with itself. Where risk factors and operative clauses disagree about the same right, the operative clause governs.
Check what the backtest was measuring. Not how it performed, but what quantity it computed, and whether the conventions producing that quantity survived into the live period.
Then ask for the version history. Clause reading tells you what a bank could do. It says nothing about what any bank has done, and nothing here should be read as suggesting otherwise. Version history answers that directly, most desks will supply it, and almost nobody asks.
We are running this across the wider universe now. The question we are most curious about is whether the scores cluster by provider. If amendment language is house convention, reused across a bank's whole shelf, then the score is not the signal. The deviation from a bank's own template is, because it marks the indices where something specific was engineered or negotiated.
