Learn / Backtest
Is this even tradable?
Test whether that pattern would have made you money
So could you actually have traded it?
Suppose you have found a group of days that leaned upward — around a particular conjunction, say. Knowing they leaned is not the same as knowing you could have made money on them, and this is where you find out.
A few tenths of a percent vanishes under a realistic stop. And an average return tells you nothing about whether you would still have been holding after the fourth loss in a row. So this turns the pattern into an actual rule and runs it over real daily bars. You pick which events to trade around, say how the trade opens and how it closes, and you get back every trade that rule would have taken: where each one filled, why each one ended, how the outcomes spread out, and the worst losing run you would have sat through.
It is deliberately unflattering about all of it. Fills take the worse side of anything ambiguous, the holding period is always capped, and when an exit filled worse than its level because the market gapped overnight, the results say so instead of quietly smoothing it over.
When is a pattern ready for this?
The moment it starts to look real. Up to this point nothing you have found can fail — a lean is just a description, and descriptions are safe. Give one an entry, a stop and an exit and it becomes a claim that can be wrong.
Most patterns turn out not to survive that. Finding out here costs you nothing, which is the whole argument for doing it in this order.
How do you set up a trade rule?
- 1
Pick the events you want to trade around. Event type, sub-type and planet — the same guided selection the Explorer uses, so a group you found there comes back here as exactly the same set of events. Holding (days) sets the backstop that always applies.
- 2
Say how the trade opens. Fixed bar enters on a schedule — long or short, on the event day or one or two bars later, at the open or the close. On level break waits for price to break the event day’s high or low instead, and the side that breaks decides the direction — so in that mode the direction setting is ignored.
- 3
Say how it closes. A stop-loss, a trailing stop, a take-profit, and always a holding-period backstop. Whichever comes first ends the trade. Within a single day, a stop is assumed to hit before a target — the conservative reading, not the flattering one.
- 4
Watch the diagram, not just the fields. The diagram redraws on every change, and it mirrors what the engine will actually do. So you can see a rule before you spend a run on it.
- 5
Read the headline tiles. Occurrences, average and median return, win rate, and a long/short split. These describe the trades your strategy actually took — sign-adjusted, so a profitable short counts as a gain.
- 6
Check every single occurrence on the chart. One dated marker per event across the whole price history, so you can see whether a result rests on a few clustered years or on the entire span. Below the chart — off the bottom of this shot — the drill-down opens Advanced stats for the distribution and drawdown, List for every fill and exit reason, and Chart to step through trades one candle window at a time.
Where could this flatter you?
Why a loss can be bigger than your stop
On daily bars, a stop does not cap your loss. If the market opens straight past your stop level, the trade fills at that open — the worse price — because on a daily bar there was never a chance to get out in between. So a 2% stop can hand you a loss bigger than 2%. Those exits are tagged (gap) in the results, and the tag is there for exactly this reason: so a loss under your stop reads as an overnight gap rather than a broken engine.
The market context (the panel measuring the raw market move around the event day) panel under Advanced stats deliberately ignores which way your trade was pointing. On a short, it can show you a rising market in green while the trade lost money. That is intended — it is describing the market, not your P&L. For what your strategy actually did, read the tiles and the trade-outcome panel.
And the occurrence count does not have to match the Explorer group you came from. Backtest counts each event once and drops any event with no forward price data to trade into. Explorer counts trading days. Same events, tallied for different jobs.
The likeliest way to misread this
Turning knobs until the numbers look good. Every setting you change while watching the result is a decision made with hindsight, and a rule you found by nudging stops until the win rate rose has been fitted to the past, not tested against it. The win rate’s confidence interval and the occurrence count are your honest guides here. A strong result over a dozen trades is still a small sample.
Nothing here charges you anything. No brokerage, no slippage beyond the gap handling, no tax, no financing. So any rule on this screen would perform worse in real life than it does here — treat every result as the optimistic end of what was available.
Which leaves the honest summary of this whole method. A rule that survives all of that is not a prediction — it is one claim that has not fallen over yet. That is a genuinely different thing from a reel telling you Mercury retrograde will reverse the market, and it is the most anyone here can honestly offer you. If you find one, we would very much like to see it.
And when you want another one to test, you need somewhere for the next question to come from. That is the Dashboard: every planetary condition active on a day you pick, each carrying what it has done before. It is the one screen that hands you candidates instead of asking you to bring one.
The standing caveat
Celestial Market Lens measures what markets did around historical planetary events. That is a description of the past. It is not a prediction, not a recommendation, and not investment, financial, legal or tax advice. Past results do not tell you what will happen next, and you are responsible for anything you do with money.