MRE ships the same object to two structurally different worlds. A reversal setup on a 15m timeframe, classified by type, with a bracketed entry carrying a pre-set stop and target. On BTC, ETH and SOL that object behaves the way it reads. On an equity index, one of its assumptions quietly stops holding, and the failure is not a matter of degree.
The assumption is that the next bar exists.
A bracketed scalp is a bet on continuity, not just direction
Every visible row in the capture is stamped 15m in the TF column and SCALP in the Type column. That combination encodes a holding period measured in a small number of bars and a risk bounded by the stop distance. The bound is what makes the position sizeable, and the bound is delivered by an assumption nobody writes down: that between the entry price and the stop price, every intervening price will trade, so an order resting at the stop gets touched on the way through.
On a continuously traded crypto pair that assumption is close enough to true that the residual is a slippage estimate. Depth at the signal bar varies, spreads widen at unfriendly hours, and a fast move can print through a resting order, but the price path is continuous in the sense that matters. Your realised loss is the stop distance plus a slippage term you can measure from your own exit prints.
On a cash index instrument the assumption fails once a day, on purpose. The session closes. Order flow stops. The next price is set by an auction after hours of accumulated information, and it can be anywhere. A stop level below the close is not a price you will get. It is a price at which an order becomes marketable, and it will fill at whatever the reopen prints. The stop stops being a bound and becomes a trigger.
Pin the clock before you map a timestamp to a session
Take the timestamps in the capture at face value first. SOL/USD at Aug 25 2026, 08:09:05. BTC/USD at Aug 24, 21:11:04 and again at 05:30:06. Stablecoin dominance at Aug 23, 22:21:06 and 22:06:06. ETH/USD at Aug 23, 22:18:03.
The panel does not display a timezone next to those stamps. That is the first item on the list, not a footnote, because the entire session analysis is a mapping from these strings onto trading hours and the mapping is undefined until you know the clock. Establish whether the column renders in the viewer's local time, in exchange time, or in UTC, and record the answer in the same document as the rest of the specification.
Once it is pinned, the shape of the problem is visible without any further work. Four of the six visible stamps sit in the evening and one sits before dawn. On a 24/7 book those are ordinary fills. On an index book, on any plausible mapping of that clock, the majority of this flow arrives when the cash market is shut, which means the signal is computed on a price series that is not currently tradeable and delivered to a desk that cannot act on it. Scoring updates on every supported timeframe close and new signals reach the dashboard, email, Discord and Telegram within seconds of computation. The delivery is prompt. The market is closed. Those two facts coexist and only one of them is a product characteristic.
The overnight is not a wider stop, it is a different distribution
The tempting fix is to treat the gap as extra volatility and widen the stop to cover it. That misstates the problem in a way that will show up later as a surprise loss.
Intraday risk on a continuous instrument accumulates through a path you can observe and, in principle, exit at any point along. Overnight risk on a closed instrument is a single draw from a different distribution, taken all at once, with no opportunity to act between the draw and its resolution. The two are not the same quantity measured over different intervals. Widening a stop to a level that would have covered a typical overnight move does nothing at all, because the stop was never going to be honoured inside the gap. You end up with a wider bound that is equally unenforceable and a smaller position for no compensating reduction in risk.
Measure the overnight distribution separately, on your own history, for the specific instruments you would trade. Close to next open, in volatility units, with the tail characterised rather than summarised. I am not going to quote you a gap statistic and neither should the memo, because the number depends on the index, the period and whether you include the days everyone excludes.
Why the obvious hedge usually costs more than the setup is worth
The desk answer is to carry an overnight hedge in the corresponding future and lift it at the reopen. It works, in the sense that the risk changes character. Whether it works economically on this signal is a separate question and it usually does not.
A scalp-class reversal has an expected edge measured per bar over a small number of bars. A hedge has a cost measured per night: the basis move between the cash instrument and the contract, the round trip on the hedge itself, the contract granularity that leaves you over or under hedged on any position not sized to a multiple of the contract, and the operational cost of a lift at the reopen that has to happen whether or not anyone is watching. Those costs are fixed with respect to the size of the edge. Hold a 15m scalp across one night with a full hedge and you have converted a directional trade with a defined target into a basis position whose carry can exceed the target.
Whether the hedging instrument is even available in the size and hours you need is a question for your venue list rather than for the module, and it should be answered before the strategy is approved rather than at the first close it matters.
Two books, two sizing quantities
The cleanest way to run both is to stop pretending one risk limit covers them.
Size the crypto book off the stop. The binding constraint is stop distance plus measured exit slippage at the depth available on the signal bar, and the limit is a per-trade loss in basis points of book. That is the normal calculation and the bracket supports it.
Size the index book off the gap, not the stop. Set the notional so that an adverse overnight move at a stated percentile of your measured distribution costs no more than the per-trade loss you allow on the crypto side. The stop still exists and still works during the session. It is simply not the quantity that determines the size, because it is not the constraint that binds on the risk you cannot exit.
Then add the rule that saves the most trouble for the least effort: a cutoff before the close, stated in minutes, after which a scalp-class index signal is not taken. Signals that arrive inside the cutoff are logged and skipped rather than downsized, because a downsized position still holds the gap and still needs an owner overnight. Put the cutoff in the strategy document with the reasoning attached, so the question does not get relitigated by whoever is covering the desk on a quiet Friday.
The coverage claim and the flow are not the same statement
One last thing to reconcile before any of this reaches an approval memo. The module describes coverage as BTC, ETH and SOL with multi-timeframe reversal detection, plus major stock indices including the S&P 500, Nasdaq 100 and Dow Jones, plus blue-chip stocks across major sectors.
The panel reports TICKERS COVERED as 4 over the last thirty days, and the instruments actually appearing in the feed are BTC/USD, ETH/USD, SOL/USD and a stablecoin dominance composite. Every one of them trades continuously. So in this window the index reversal book is a coverage claim rather than an observed signal stream, and the gap analysis above is a design exercise for a book you would be starting, not a diagnosis of a book you are running. That is a materially different thing to put in front of a committee, and it is the sort of distinction that is much cheaper to make yourself than to have made for you.