How an idea earns a place in our lineup

From an observation to a written hypothesis, controls that can kill it, the most recent months locked away as out-of-sample, and a running fight against backfitting. Most ideas die on the way. That is the process working.

Most backtests we see answer the wrong question. They ask “what settings made the most money?” We ask “is there anything here at all?”

Here is how a new idea moves through our process before an automated NinjaTrader® strategy is allowed anywhere near an account.

1. It starts as an observation, not a strategy

Something we keep seeing on the chart, often from watching good discretionary traders work. We write it down as a plain sentence that could be wrong: “after a fast drop and a recovery of a key low, price reaches the next level more often than chance.” If it cannot be written as a sentence that could be proven false, it is not ready to test.

2. We write down how it could die, before we test it

The kill rule comes first. If the idea cannot beat simple controls — the same levels without the setup, a naive entry, random lows at the same time of day — it is not an edge, however good the chart looks. Writing that down in advance is what stops us from moving the goalposts once the results come in.

3. We lock away the most recent months

The last few months of data are set aside untouched. All the work happens on the older data. The hidden months get opened once, at the end, only to confirm. If we peek, they are no longer out-of-sample, and the test is worth nothing.

4. We fight our own backfitting

Levels come only from bars that already existed at the time. No future information. When a single bar touches both the stop and the target, we count it as a stop. Every rule is logged on its own, so we can switch it off and see whether it was doing anything. And every extra parameter has to earn its place, because three more knobs can make any idea look brilliant on the past.

5. Only then does it get dollars

Counting comes first. P&L comes later, from the trading platform’s own simulator, with commission and slippage on every trade.

Most ideas die somewhere in steps 2 to 4. That is the process working, not failing. A strategy that survives all five was not tuned until it looked good; it was given every chance to fail and did not.

Then it trades in public

When something survives, it trades forward where anyone can check it. Our two micro engines trade in public as Orthant Duo on Collective2, where every closed trade is listed after costs: collective2.com/details/157794444.