Research Register
Making a backtest look strong is the easiest exercise in trading.
You start at the edge of what has already been tested, instead of spending months on a direction that closed here in a week.
This page and the MCP tool get_research_verdicts read the same table, so your assistant answers with exactly what stands here.
Basket 1
Both readings point the same way on gold. If I take only the M5 entries where a candle strength signal lines up with a support or resistance level, does the confluence lift my hit rate?
One entry stays open for everyone. This is what a full one looks like.
When one currency is strong across the board and another is weak across the board, that should be the cleanest trend there is. Can I trade that spread mechanically, and if the extremes revert, can I fade them?
What we measured, the verdict, what it means for your own testing and what stays open: VIP opens all four.
Does the grid logic from a gold system carry over to an index like US500?
What we measured, the verdict, what it means for your own testing and what stays open: VIP opens all four.
Range bound crosses like AUDCAD, AUDNZD and NZDCAD look made for mean reversion. Does a basket of them diversify a grid?
What we measured, the verdict, what it means for your own testing and what stays open: VIP opens all four.
CFDs hide the real market. If I read the futures curve and exchange volume, do I get a timing edge my CFD chart cannot show me?
What we measured, the verdict, what it means for your own testing and what stays open: VIP opens all four.
My broker prints tick volume. If I compare it against exchange volume, can I see where the real flow is?
What we measured, the verdict, what it means for your own testing and what stays open: VIP opens all four.
Basket 2
Every entry here started as a plausible improvement to a shipped system. Each was measured against the shipped version on identical data with identical costs. The pattern in this basket is the point: most improvements that sound obvious cost money once they are measured.
The system misses vertical moves because it waits for confirmation. Why does it skip the market entry right at the breakout, or a hybrid of both?
What we measured, the verdict, what it means for your own testing and what stays open: VIP opens all four.
A swing pivot tool finds the real structural pivot. Why does the system place its initial stop somewhere else?
What we measured, the verdict, what it means for your own testing and what stays open: VIP opens all four.
The indicator lets me choose how much history it reads. Is a longer history the safer setting, and does agreement between two history lengths mark the better signals?
What we measured, the verdict, what it means for your own testing and what stays open: VIP opens all four.
The basket gives profit back during the day. Should it lock in a share of the day's gain and stop for the day?
What we measured, the verdict, what it means for your own testing and what stays open: VIP opens all four.
If the give back is real, the losing trades should look different at entry. Can a marker sort them out in advance?
What we measured, the verdict, what it means for your own testing and what stays open: VIP opens all four.
The live account trails the backtest. Where does the loss actually sit?
What we measured, the verdict, what it means for your own testing and what stays open: VIP opens all four.
When a release hits, some pairs move much harder than others in the first minutes. Should the basket put more risk on those legs?
What we measured, the verdict, what it means for your own testing and what stays open: VIP opens all four.
A second release turns the currency around while the first basket is still open. Should the old basket be closed, or at least its weakest legs cut?
What we measured, the verdict, what it means for your own testing and what stays open: VIP opens all four.
The standard
Every verdict here was produced the same way. A Python simulation only sorts out what is hopeless. The word tradeable is earned in the MT5 Strategy Tester on real ticks, with the spread, commission and swap of a live symbol. The rule is fitted on one half of the data and applied unchanged to the other. Sample size, a bootstrap interval, an outlier check and a stability check all have to pass before a result counts.
Stop and target inside the same bar, where the simulation has to guess which came first.
A value that was still forming at the moment of the decision.
Results computed before spread, commission and swap.
A fixed forward window that silently drops every trade resolving later.
One sample, no bootstrap interval, no out of sample half, no stability check.
The best excursion seen so far, read back as the price at that moment.
Reused inputs from the last run, and a run on thin history.
One bright cell whose neighbours are poor. A finding is a plateau, a spike is noise.
Two runs of the same system carrying very different cost per trade.
Where a study searches through many candidates, the hypothesis, the cells and the kill criteria are written down before the first number is computed, and the best candidate is compared against the best of an equally large set of random ones. All of this establishes what held on the tested instrument, in the tested window, under real trading costs. Live results decide the rest, and they are the last stage of every cascade we run.
Ask your assistant
Point your MCP client at our endpoint and call get_research_verdicts. Ask by topic, by symbol or by entry, and you get what stands on this page, with the same cut: the question and the scope for everyone, the verdict for VIP keys.
https://stein.investments/mcp Tool: get_research_verdicts · How to connect →
Notes
All figures are results measured on the data and in the windows named in each entry. They are past observations and carry no forecast and no recommendation to trade. Trading carries risk of loss. Parameter values, thresholds, filter logic and engine internals of our products stay with us. What we publish here is the question, the method, the result and its boundary.